Seconded. I have no clear idea what this project is even about.
Anki exists. Better Anki exists in the form of SuperMemo, which is a UIX nightmare.
Khan Academy also already exists.
"AnkiHub will empower students by democratizing accelerated learning and potentiating the ever growing wealth of quality, free, educational resources. "
I agree with you that I (medical student, Anki user for 10+ years, 30k+ cards) am not clear on what this proposal is suggesting, but would just like to vouch for the team behind them- it looks from the website linked that its the guys behind the AnKing deck, which is a very well established/respected deck in the medschoolanki community.
It looks like the key offering is this platform to make real-time collaborative edits to Anki collections, which is a cool idea insofar as you could get a small group of people on your course to build a useful Anki collection together in a slightly smoother way than is possible now. Other than that I'm not clear what the rest of it (Anki Mastery course etc) adds that isn't already widely available.
I heard about that deck and I don't even study medicine! Good to hear, that there's something legit behind that. Collaborative real-time editing does sound useful, too. Maybe that's what it takes for a whole class to use the same Anki deck? Or just a couple classmates building it together, which then gets the teacher's attention. And once a whole class adopts it and they outperform, it might become very obvious that using SRS is low-hanging fruit and should be a deeply integrated into any curriculum. But currently there is not a great platform to make Anki work in a class-environment.
I kinda get, why it's so vague now.
There's almost too many interesting possibilities on how that could be developed.
And there's no real limit to the ambition behind it. Once you've got a beachhead in some uni in medicine, you spread it to other unis.
Other fields. Once you have a working platform, where teachers can quantify their student's progress, why not go for high schools?
Of course, this might not work out, since the SRS revolution could have happened 20 years ago, but it's great someone taking a shot at that. Still think, they could have done a better job communicating. I would have sold this way differently to this audience, cause the way it's phrased just triggers everyone's bs detectors here. Almost tempted to try.
I'm all ears if you have any suggestions for how I can avoid sounding like a maniac when I say, "trust me, this could revolutionize education." I feel like this is something people aren't going to really grok the potential of until they can actually try it out. "Make it really easy to curate and study a comprehensive set of highly individualized digital flashcards for anything you want," just doesn't quite convey the spirit of the thing... But I thought it was worth a try!
Although I'm on board with the idea for collaborative editing, I'm always wary of people talking about SRS having the potential to "revolutionise" education, which I feel like kind of oversimplifies a) what SRS is good and bad at and b) what education/learning/knowledge/competence really is.
Sure, I'll take a whack at elaborating. Collaboration is just one of many things in Anki that is super tedious. Utilizing pre-made decks from Ankiweb or elsewhere introduces another major problem for students: selecting precisely the cards they need, at precisely the moment they need them. This problem actually makes pre-made decks very difficult to use effectively. The gold standard set of Anki flashcards for medical students has over 40K cards. The current workflow med students use to take advantage of these pre-made cards is a nightmare. So one of the first features we would likely work on after collaboration is streamlining the scheduling of cards based on information about what students are currently studying. For example, students could collaborate on lecture notes in a Google Doc, an Anki add-on could take that as input and use it to search a database of all prompts, and schedule the best matches for that group of students.
Although Anki has a plugin-based architecture, add-on authors are extremely limited by the Python environment shipped with Anki. This means no NLP libraries, no statistical computing libraries, etc. So AnkiHub will also enable a web-based plugin model with far less limited computing environments that have access to a massive database of all the cards on AnkiHub.
Here is a list of other things that the platform can facilitate once we make using Anki more joyful:
- Curating a database of review stats for thousands of students
- open sourcing such a database could lead to interesting discoveries in SRS, such as optimizations for the scheduling algorithms
- Enabling smarter scheduling algorithms
- e.g., an algorithm that was aware not only of your individual reviews/ratings but of all reviews from all other students studying the same prompts
- Automated flashcard generation using ML/NLP
- A deadline-aware scheduling algorithm
- To ensure students study their prompts as much as needed before a test, for example
- Automated image occlusion
- Input an image with text labels and output prompts by occluding the text
- A suggestion engine for discovering new prompts to add to your queue
- This could be based on inputs like youtube history, articles ready in Pocket, etc.
- automatically identify duplicate/similar prompts in a users collection and merge them into away in order to avoid redundancy/studying more than needed
- automatically suggested tweaks for a more quality card
- Generate mindmaps/zettelkastens from a collection of cards
- This would enable intuitive exploration of a collection of cards, addressing the common complaint of losing sight of the bigger picture.
- Auto discover and link to external sources like wikimedia, etc
- cards can be powered up by a bot that autolinks to relevant resources
Regarding the course: we have also published tons of free Anki learning resources, not to mention that we give out scholarships for the course for anybody who can't afford it. It's no secret that anybody can learn what they need for free from our YouTube channel and elsewhere. People buy the course because it offers a guided path to learn Anki quickly. We also provide individual support and a forum for people to ask for help as well as an add-on that configures Anki settings automatically.
Thanks so much for fleshing out some of the Anki palace ideas. I think some of them sound really cool, although I'm instinctively kind of wary that there's so many of massively varying complexity/difficulty.
Broadly, I think massing stats from large numbers of learners could feasibly tell us something about optimal scheduling, although the style of card, threshold for hitting different values of remembering and even differences between individual memory makes this very fuzzy data.
Some easy workflow for incremental reading->Ankifying is the holy grail that I and a lot of people have been searching for for a long time, and I've never seen anything come really close to it yet- off the top of my head, getpolarized.io looked to be doing interesting things, but I suspect its a very tiny market.
Some of your other suggestions, like generating mindmaps from cards/automated flashcard generation, I feel like underestimate the significant import of a) the process of producing, rather than just reviewing cards and b) the quality of the cards themselves- very very few people make cards of sufficiently quality to even check off all the 20 rules, never mind genuinely enable the kind of real long-term knowledge we're all looking for. Until we're better at that, I'm not sure how easy it is to automate the process- although of course the payoff could be substantial.
I don't want to sound negative, because I'm really pleased that people are doing things in the SRS space. I think used right it can be a really powerful tool, but that an understanding of Anki's limitations and correct use is both vitally important and pretty rare amongst the people I see using it day-to-day.
I'm going to try commenting to these one at a time.
#1: A Movement To Fight Attention Hijacking
Oh, God, no. People's attention is being hijacked enough, How about a project to teach people to protect their attention and think about whether what they're doing makes sense. Remember that some of your benevolent uses might be getting things wrong.
Sounds cool. I don't if you want to expand your subject, but one vagueness about a stabbing pain in the shoulder is that people mean different things by shoulder.
There are a lot of good ideas here. There are also a few bad ones; though I'm not going to point them out.
If I were a billionaire, I would give each of them $10k each, no questions asked. (Signaling risk is real.) Of course, since I'm not a billionaire, that's fairly easy for me to say.
I expect I will be emailing half-a-dozen of these people, a mix of "how can I help" and "would you like an investment -- and I do want my money back".
The total: 9 emails sent, one (#41) I would have emailed if they asked for anything, and about 5 where the "if this works, somebody else will give you better terms than I can, so I won't bother" principle applies.
I did get one reply to an email where I offered money, asking "I am not sure if [direct payment] is the right thing to do, or the money has to come from ACX." To be clear, I assume Scott doesn't want to be in the middle of any transactions here?
Hey Alex, I’m the person behind #41. Thanks for your interest! I’m not looking for any money, since that isn’t the limiting factor to what I’m doing right now.
But if you have any advice or criticism of my project, you can reach me at mikest@udel.edu.
Specifically, I’m trying to figure out what topics are the most useful to cover, and how to reach more people with my videos. So please let me know if you have any advice about that. Thanks!
I don't know what everyone else is thinking, but I was probably influenced by an NPR piece about sound and battlefields. Until amplified sound and radio, the military was limited to what it could quite literally see and hear.
Re. the one on the texture of pain: I like the idea. I just want to highlight how helpful it would be in medical practice to have a device that can measure pain. We still have no practical (eg non-fMRI) way of determining when unconscious or uncommunicative people are in pain, nor of measuring pain levels. So after they characterize the relation between textures in functional imaging, and perceived pain, it would be useful to search for correlations of fMRI-detected pain with some signal that can be found in blood samples (say, detection of endorphins), galvanic response, pulse, or skin temperature, or other simple assays.
An objective measurement of pain would help a lot with physical therapy studies, which currently rely on patient-reported subjective pain, which can change a lot from day to day because perceived pain is relative to recent events and state of mind. (BTW, the texture study might try to factor that kind of context out of its pain measurements. I mean factors like time of day, season of the year, current weather, and stress hormone levels.)
An objective measurement of pain could hinder actually getting physical therapy, though, if insurance companies decree that patients must stop receiving PT when their objectively-measured pain stops decreasing. Pro tip: When you start receiving PT, say your perceived pain is at least a 5 on the 1-10 scale. Every time you get re-evaluated to determine whether you still need PT, you need to report a lower pain level (to show the PT is working)... but this boxes you into a corner if you tell them your pain is a 3 at the start (which people often do because the level of pain that brings people in for PT is much less than intense pain from acute injuries); after 2 re-evals, you can't qualify anymore no matter what answers you give.
The fact that the only way to currently measure pain is to ask someone to subjectively provide a number (or point at a sad face!) bothers me constantly. If we could objectively measure pain it would solve so many problems. Just think of all the people who suffer chronic pain but don't have their problem taken seriously because there's no obvious cause! Or, alternatively, we could prove those people are actually in no more pain than normal and they can go see a psychologist. I guess that would depend on how it works.
I guess one way to go would be to define a set of calibration pain sources, but that has obvious ethical limitations as well as dual use concerns. Of course, even assuming that one could induce similar neurological signals, it could be that different brains react very differently to them.
Or it could even be that pain is not scalar at all, e.g. that there is no inter-human fixed preference ordering between nth degree burns, kidney stones, etc. And this is before we even get to emotional pain.
I think this would be relevant not just with pain, but with other sorts of subjective symptoms. Over the past two years I feel like I've grown more aware of the differences between cough caused by allergies and by generic dust or liquid stimulation (but I've also managed not to have any viral coughs, so I don't know as well whether those feel different). It seems like it could be useful for some people to be trained to know the difference between symptoms, even if most people don't pay attention.
This sounds almost identical to the old Book Genome Project (circa 2013) on Booklamp.com, which aimed to use semantic analysis to categorize books the way Pandora categorized music and use those to build recommendations. That project and company got quietly bought out by Apple Books as soon as their recommendation engine got good (it wasn't announced until months later), and the website was shuttered; I've never heard a word about it or the Book Genome Project again.
I'd be curious to see if Shepherd.com a) gets successful, and b) gets bought out silently the same way.
Good recommendation systems would wreck the financial planning of the publishing industry, by flattening out the distribution of sales per book. Advertising has an S-shaped ROI curve; very small and very large marketing campaigns are both low ROI. But "very small" still means very large relative to the profits on most books. One rule-of-thumb is that 50% of all books lose money, and 5% make money. This article on Forbes claims over 99% of titles published lose money, but it's counting specialty publications like agency reports: https://www.forbes.com/sites/quora/2014/05/28/what-is-the-percentage-of-books-published-on-which-publishers-actually-lose-money/?sh=1592aef36d58
The purpose of recommendation systems is to destroy the power-law distribution of sales per title (a miniscule fraction of titles account for most profits). But that power-law distribution is profitable for sellers, because it makes advertising more cost-effective. So big sellers like Amazon probably don't want recommendation systems. Notice that Amazon book recommendations are usually of best-sellers.
Remember the Netflix movie-recommendation competition? They paid out a million dollars to the winning team, but ended up not using it at all, because they realized that effective personalized recommendations weren't as profitable as strategic marketing recommendations.
> Good recommendation systems would wreck the financial planning of the publishing industry, by flattening out the distribution of sales per book.
Aren't you assuming that all books are equally interesting here? You're also assuming that being able to talk about books with other people easily is not a factor when getting a book recommandation.
Good questions. I'm not assuming those things, because this is not a Boolean decision between having the current power-law distribution, and a completely flat distribution. The goal of recommendation systems should be to flatten that power-law distribution somewhat, so that it gets closer to the normal distribution which book quality should have. Not all books are equally interesting, but the current distribution of book sales matches the distribution predicted by increasing-returns models in which quality has no effect on sales, and cannot be reconciled with any sane normal distribution of book quality.
You're right that there are increasing returns to the reader wrt how many people they can talk to about a book. But I think (but can't prove) that's only a small part of what produces the power-law distribution of sales.
Do we know they didn't use the techniques? Surely they heard about them while the contest was running?
The story I heard is that the big change was the move to streaming. This changed everything. The data was now in a very different shape than it had been before. There was a lot more data, but also it was revealed preference. And, the catalogue contracted.
It was in an article about the competition, written shortly after it was done. What I wrote about Netflix deciding not to use it was based on a quote from a Netflix employee, but that person didn't come right out and say "it's more profitable to do X". They said something a little vague, which I interpreted at the time as meaning the winning recommendation system wasn't profitable. But I don't remember where I read it, so I can't check that now.
Regarding being bought out, or more generally, pivoting to providing a shittier (but more profitable) service, similar comments apply to Aella's dating app (#56) and the 2 social networks (#40, #47). It would be really good if there was a social / legal / whatever mechanism such that websites were incentivized to provide a good service for their users and not chase continual growth/profits.
In defense of the free market, the perfect is the enemy of the good. There /are/ mechanisms that incentivize good service; but they can't work perfectly without perfect competition, which we certainly don't have with Apple music, Amazon books, Netflix streaming services, or the Google Maps API.
I partly blame legislation that big companies have been busy getting passed these past few decades, which make start-ups expensive or even impossible. Especially recent Internet regulations on taxes, tax reporting, observance of laws in the state and county being delivered to (eg selling liquor on the Internet), observance of European Union laws (notably the ones to "protect" minors), and so on. Also, the interpretation of patent law, the switch to first-to-file patent priority, and excessive regulation of biotech, to name a few.
But ultimately, perfection is impossible. These companies provide services that make some compromise between the customer's interests, and their own. Most of the solutions proposed for these problems require decisions to be made by government, without considering the customer's interests at all.
Hi, I'm the author of #40/Haven. I'm trying really hard to structure this differently. I'm not raising any money. The software is open source and the protocols are dead-simple (it's just RSS). The emphasis really is on some form of self-hosting. I expect this to generally take the form of paid hosting. My hosting offering puts each Haven on it's own dedicated VPS--and there is no barrier to other entities taking the software I write and offering their own hosting, or whole different implementations that speak the same protocol. My biggest fear is if Haven catches on enough for some company to start offering a "free" version with all the tracking/ads/censorship/invasions-of-privacy that are the status quo today.
There's a whole world of non-corporate possibility that opens up if we can find some way to make self-hosting (in some form) feasible to the broader population. Haven is my attempt to make progress on that front.
Thanks for including my project here Scott. If anyone has any questions or advice about mine (making free online tutorials for Bioinformatics and Computational Biology) please let me know! I don't need any money for this, but definitely appreciate advice and constructive criticism.
Hi Mike, I think your videos are great and I've actually used a few to help me learn things. I don't have any concrete advice, but I thought you'd appreciate having at least one supporter.
Haha, losing line breaks was a bit unexpected %) I'll take the liberty to post a version of #6 above, but with line breaks as-submitted :D
--
I'm Oleksandr Nikitin, and I want to build a system dynamics simulator.
Enable independent researchers to simulate, forecast, and visualize metabolic pathways, epidemic spread, mass transit, ecology, macroeconomics, etc. Show, don't tell. Without code.
Think Airtable+Vensim+Roam+Kumu, integrated and working offline.
Why offline? Why simulate? Why a new tool?
- Complex systems must be simulated. You miss emergent phenomena if you analyze parts separately or simplify the details.
- Offline sets you free from distractions and groupthink. Free to make your own breakthroughs.
- Take your references, notes and data with you, dive deep, then return with the verified, reproducible, interactive model.
- Research can take years. Tools should outlast devices and app stores.
- And it must be fast. Isn’t it insane for a productivity tool to make you wait?
I spent years on prototypes, tested in companies since 2013, now I want to put these experiments together. Not as a startup, but as a tool accessible to everyone.
The plan:
- Create a community of curious, inquisitive makers.
- Empower them with small, fast and robust core app.
- Iterate and grow together, augment the human intelligence even more.
- Understand the world.
I seek funding to focus on this project full-time, for two years.
To launch and to guide people to the finished research.
Sounds inspiring? Worth the funding? Want more details?
My immediate observation is that IMHO a reasonable model for funding something like this is how many other tools have been made, namely, when someone gets a grant to do research where this tool would be useful (e.g. employing a few grad students/postdocs to do system dynamics modeling for a few years) then they also use part of that funding/manpower to develop such a tool or improve an existing one. However, for that to work there needs to be some basis of core functionality and structure, and that basis needs to be open source (otherwise we just get a new iteration of Vensim), so the licensing policy choice - which is not discussed in the description - is quite relevant to how the project is be positioned; developing a startup product is quite different from developing an open source community.
I'm happy to answer questions anyone might have about #57, "Advocate Against Subsidies And Tax Breaks For Local Corporations."
I'm the president of the Center for Economic Accountability, the 501(c)(3) that's putting together the "Skeptical Reporter's Guide to Covering Economic Development." Our board chair and I are both regular SSC/ACX readers, and we're grateful to Scott for the opportunity to spread the word about our work to an audience that's predisposed to actually pay attention to research findings.
I'm naturally skeptical of opportunities like these. I oppose many (though not all) tax breaks and subsidies for large corporations. However, I'm not convinced this will be effective. Most corporations' governmental benefits seem to flow in because of the company's ability to directly pay for influence, nit because of independent reporters spreading the wrong story. As an example, during early stages of the pandemic, I felt like every news article I read was outraged over commercial airlines' tax breaks — yet despite all of the bad press, they (and companies like them) continue to get big breaks. I am willing to acknowledge that:
1) I don't read the news much, so maybe most news is super in favor of big business' tax breaks. This doesn't match my personal experience, but it's definitely possible.
2) I almost never read strongly conservative news sources for opinions about taxes / subsidies. Perhaps this praise of government tax cuts is common, it's just tied to conservative news sources.
If it's the first, I'd be interested in seeing examples of major news sources praising subsidies and etc. If it's the second, I'd really doubt that telling reporters to stop praising tax breaks would overcome whatever political purpose praising tax breaks serves, though I'd be interested in at least hearing the theory.
The problem isn’t that any one newspaper story is cheerleading for any one subsidy deal. The problem is that as a matter of practice, media coverage of these deals consistently report as fact things that we know to be at least unlikely and at worst untrue about the impact that subsidies have on corporate site selection decisions. Since local media tends to be the only source of information most people will encounter about these deals, the result is a climate of public opinion in which misinformed voters reward politicians who take credit for “creating jobs” through subsidies, despite evidence that these economic development programs are at best ineffective and at worst can be extremely harmful to the communities they’re supposed to be helping. (For more on this: https://economicaccountability.org/research-resources/)
Reporters, by and large, want to get the story right. (I used to be one, decades ago.) But overwhelmingly, what we’ve found from years of reading story after story after story about the most recent deal is that they’re *not* getting the story right, because they’re not asking the right questions and including the right context.
Consider the recent headline “Boosted by state incentives, General Motors making $7B investment in Michigan, creating 4,000 jobs.” The story begins, “General Motors, building toward an all-electric future, plans to invest $7 billion and create 4,000 jobs across four manufacturing sites in Michigan. Supported by $824.1 million in state incentives, the Detroit automaker will expand an Orion Township plant, build a Lansing electric battery manufacturing facility and make upgrades to two Lansing area plants.” (https://www.mlive.com/public-interest/2022/01/boosted-by-state-incentives-general-motors-making-7b-investment-in-michigan-creating-4000-jobs.html)
The clear implication of "boosted" and "supported by" here is that the incentives changed GM’s decision about where to invest $7 billion in new and upgraded plants. But if you look at the actual dollar amounts in context, that idea’s laughable. In reality, the $824 million in subsidies is less than three days’ worth of revenues to GM, and it’s being spread out over a decade or more. It's an insignificant blip on the company's bottom line. We saw the same thing in North Carolina, where $830 million in subsidies to Apple spread over 30 years were supposed to have sealed the deal for a company that was booking more than a billion dollars a day in revenues.
The idea that a major corporation is going to base a decision about where to locate mission-critical production facilities on the equivalent of less than a couple hours’ worth of additional corporate revenues a year, rather than the best place based on actual business-critical factors such as workforce availability and cost, is just ridiculous. It's not even borne out by the site selection industry's own surveys. But that’s what gets reported as fact, over and over again, which is a big reason why the average person believes that it’s true.
Reporters also rarely include context about historical performance of these programs. If a particular tax only brought in a third as much every year as the state was budgeting for it to collect, that relevant context would be included in stories about the tax. But while it’s extremely common for official audits to find economic development agencies rife with mismanagement and only reporting half to a third of contractually agreed-upon jobs -- and that's if you give them credit for every single job being the result of a subsidy. For instance, a 2017 audit of the Wisconsin Economic Development Corporation found that the WEDC's recordkeeping and oversight was so bad that it "cannot be certain about the numbers of jobs created or retained as a result of its awards." (https://legis.wisconsin.gov/lab/media/2627/17-9full.pdf) Later that year, the WEDC was entrusted with overseeing what would become one of the greatest economic development boondoggles of the modern era, the Foxconn project. if media coverage of WEDC deals at the time had included the context that the state's own auditors were screaming from the rooftops about the agency's pervasive inability to do something as simple as get companies to report whether they'd created jobs or not, maybe there would have been less support for the disastrous Foxconn deal.
(In 2019, the next round of audits found that the WEDC had improved to the point where it could actually report a job creation number; the bad news is that the number was only 34.9%: https://legis.wisconsin.gov/lab/media/2861/19-6full.pdf)
I spent about a year building and running a bespoke dating site with all kinds of interesting ways to match people. None of that matters - there's really only one important problem in online dating, and that's "how do I bootstrap a critical mass of attractive women on the site?" Personality tests (what drove OKCupid's early success) won't work anymore because Facebook has dialed the virality of these back to near-zero, and most people are burnt out on that kind of thing anyway.
Not that I think it's impossible to start a new dating site, but you have to start with the chicken-egg problem and ignore _everything else_ until you solve that. Most people who want to build dating sites spend all their time thinking about the matching stuff, because that's more fun. But it's the wrong problem.
better question is "how do I get a critical mass of women who will match with my male users". there's definitely untapped potential in the "dating app specifically for not-hot people" market for whoever can figure out how to make that sound not degrading
From what I gather, the major problem for dating sites today is trying to sort out the "only looking for casual sex and ideally a range of partners" from "want to date and maybe turn this into a long term thing".
The second problem, as you say, is "how do we get enough women on here?"
And the third problem is "how do we match up the women who want something more than 'fuck me tonight' with the guys who are the same, and avoid all the 750k horny guys just looking for some action?' because I think that drives women off.
> the major problem for dating sites today is trying to sort out the "only looking for casual sex and ideally a range of partners" from "want to date and maybe turn this into a long term thing".
The biggest problem is that individuals in your userbase don't even have very accurate guesses as to which of those two categories they're in. There are plenty of people who say they're interested in one but are actually interested in the other, in both directions. There's also people whose preferences go back and forth and they don't know which mood they're in tonight.
I still think personality tests have a good shot; personality test websites (i have a friend who runs one) gets hella traffic still; people share screenshots, and do it over platforms that aren't facebook. I also have some experience doing organic marketing over female-heavy platforms too (like tiktok).
I don't think 'critical mass' is 50%; most dating sites are functional with an imbalance of male/female users. It's hard to get around the "fewer matches will be made" with the imbalance, but there's lots of ways to get around the imbalance in in-app attention.
How many unique users are in hella traffic? What seems like a lot of users in a national or worldwide context ends up being fairly thin when you look at "who can I date in my local area?"
I also worked on Let's Date, an early Tinder competitor started by the guy behind suicidegirls. Seeded by that community and a million bucks in facebook advertising (aimed only at women), it reached maybe 100k users? But you still ran out of matches pretty quickly, even in urban areas. And the ad spend was not sustainable.
I agree that the balance ratio doesn't really matter... in as much that the men are irrelevant and you can ignore them in our your outreach. If your site has women, there will always be plenty of men. It sounds crude, but it's true.
I don't want to dishearten you - I'd love to see someone to take down IAC.
I ran out of matches pretty quickly on okcupid, but this seemed fine to me? As in, there were like maybe 10-15 eligible males in the area I wanted to date, and i would go message all of them and maybe go on dates with a few, or something. And I'd check back occasionally to see if anybody new was in the area, or answered more questions to update the match score, or of course if I was traveling for anything. This seems like the way most people used okc, and it seemed to function; I think the point isn't that you should get a lot of matches, but that you should get a few really high-potential matches. Of course I agree getting more women is better, and getting more women is hard. I do intend this to be 'niche targeted', at least at first; as in, I want to target communities where women will be organically interested in this sort of thing, and more likely to be compatible with the men who sign up.
OKCupid had a critical mass, definitely. In my mind it was the peak of online dating, though that may be an age-related judgement. I just think the number of people you need in the system to have 10-15 eligible candidates is pretty large, especially if you're filtering them by personality characteristics and not surface appearance. And getting that critical mass is the core problem because the viral mechanisms that existed in the mid-2000s don't really work anymore (or at least, don't work anywhere near as well).
Another complication is that back in the mid-2000s the free dating sites like OKC and Plenty of Fish supported themselves by advertising paid dating sites like Match.com. Now IAC owns everything and doesn't need to advertise, so that business model is probably over. Hosting costs have come down but programmers are more expensive, if anything.
I feel like okcupid's emphasis on personality tests worked great in the late 00s when online dating was more stigmatised, because you could lie to yourself and say that you were "only there for the tests".
I really don't think you can look at personality tests in isolation. Personality tests worked in the 00s because people posted results on their Facebook wall, which drew other people to the tests. The important part is not the test itself but the mechanism by which people found out about the test.
If your viral coefficient is greater than one, your userbase grows. If it's less than one, it shrinks. Sharing results of personality tests increases the viral coefficient. Taking a test and keeping the results to yourself does not.
I just don't think people share like that anymore? Maybe people are burnt out on "which character on star trek" etc they are. Maybe Facebook's algorithms have squelched most of that content. Maybe people are more concerned about privacy. It's possible that my cohort just outgrew that kind of thing and "kids these days" are still posting that stuff on snapwhatever, but I kinda doubt it.
My guess is that there's a community of people who love personality tests, and maybe you'll get some viral spread in that community via the few remaining viral mediums, but it won't be possible to achieve liftoff. So you're stuck trying to supplement the viral effects with traditional methods like advertising.
I could be wrong - I could imagine a new personality test that is so interesting and compelling that it erupts into public consciousness the way people are currently sharing Wordle scores. But deliberately crafting successful viral memes is insanely hard.
I'm an ignorant person who never used a dating site. I guess it's supposed to be obvious that "old" OkCupid was better than the current version, but I don't know why that is supposed to be the case.
Girls love personality tests. But a lot of those on OKC turned out to be just "Will I date you? Answer these 6 questions to find out..." Which suggests a real business opportunity: a site ("hoops.com" or maybe "onlyfans++.com") in which women can list a set of actions suitors can undertake to earn steps in some graduated sequence of interaction. 1. Send me a photo of yourself wearing your favorite article of women's clothing to earn a chat. 2. Here's a photo of my eyebrow, upload your woeful ballad to earn a 20 minute Zoom. Et cetera.
This has been a passion project I've had in my head for a long while, and I don't even have a use for this myself.
Trying to resist the urge to start developing it right now.
Pontifex already said this in another comment, but the biggest concern I have is that if this takes off, the usual path is to either get bought by one of the big players or become exactly like them. Which would be a shame as it would be a wasted effort then. And like Pontifex said, I would want there to be some mechanism to prevent this.
If you don't get funding for this you could go open source and ask some of the 750k horny men to contribute.
What has happened to okcupid anyhow? Wikipedia mentions that it was acquired by match.com in 2011, which was around the time I stopped using it because I found a relationship in meatspace. From the post, it is regarded mostly unusable now?
(From what I remember, okc was one of the less emotionally painful to use dating sites I used, possibly because the number of matches was small. And they showed the reply rates. I think I even got replies. Looking back in general, I think non-cynical online dating was probably the quickest way to nurse insecurity and self-doubt to a certainty of worthlessness. Short of stating a deeply unpopular opinion on twitter, perhaps. I found the micromarriage to micromort rate was really unfavorable for accumulating the former, and many activities are obviously more efficient (and less soul-sucking) for accumulating the latter.)
Call me a heartless bitch, but how is this not one of those scam email type appeals about "I am poor sick widow in Third World need urgent medical help God bless you" that I've been pestered with?
Does listing order indicate your evaluation of them? If not may want to mention how they are sorted so being no. 1 isn't taken as endorsement and vice versa
#17: Algorithms To Select The Best Systematic Review: I don't think this is the sort of thing that machine learning can currently do. Humans are probably still better at evaluating methodologies. And I think the whole system will probably rely on having humans select the best reviews, in order to train the ML algorithm. Why not just have those humans say why they selected those reviews as best?
Nor do I even think there is one best way of doing systematic reviews. It might be better to have a variety of review methodologies and rely on the biases cancelling each other out--which random-forest machine learning has proven to be surprisingly effective.
I’m definitely finding I can now abstain from certain kinds of clickbait. I see the click bait and I immediately think about how most journalists have nothing of value to say. I find via that thought, I avoid the pull.
Happy to answer any questions for those curious about #19, AnkiHub (education technology, spaced repetition, etc). Also, check out our page at https://courses.ankipalace.com/ankihub to get updates about the project.
It seems to me like this is extraordinary bad signaling for these projects -- you're essentially saying, "Here are projects that I've judged as being unworthy of receiving grant money, go ahead and pick through the slop and see if there's anything you might be interested in."
I don't understand why anyone that's savvy would want their project listed here as being explicitly judged less worthy then the least-worthy project from the set of all of the winners of the grant. Money is the unit of caring after all!
In retrospect, one way to prevent this signaling risk while still being able to show the losers of the grant process here would be to have the grant awarding process be one where submissions have to meet a certain "very happy-to-fund" standard by direct evaluation, and then randomly allocating funds among the projects that met that standard, while publishing the remaining projects explicitly as projects that definitely could have received funding but simply missed out through luck of the draw. It also would have made the grant process probably easier to administer, as you don't have to compare projects against each other to determine a ranking, just evaluate each against the "very happy-to-fund" standard.
That's not the alternative at all! They could just post their projects to the comment section under the open threads instead and still reach that audience, without associating their project with losing the grant and incurring negative signaling.
But people actually looking to fund something are more likely to look at this thread than another open thread. Especially if someone decides several months from now that they'd like to fund something, they may well come back to this post but are unlikely to go trawling through the other open threads.
I don't think people would view it that way, I see it as cheap advertisement that meets a very low bar of not being literal trolling. I think there actually may some value in a public post like this, especially since some of these proposals are very similar. There can be quite some value in connecting these people to each-other.
Presumably many of these projects would have preferred to get funding directly, but there are reasons other than "this is a bad idea" these might not have been funded.
The best example I saw was #58 (waste heat in data centers) - if you actually own a data center, this could be a wonderful opportunity. And Scott's readership does include people at the companies that run data centers (Google, Amazon, etc.). However, if Google, Amazon, etc. all don't like it, you probably shouldn't invest either.
I think we're in agreement here? Rephrasing what you've said, you agree that among other reasons, "this is a bad idea" might be one of the reasons they didn't get funded. That's precisely what I call "bad signaling".
What do you think of the idea of "threshold, then randomly fund"?
If the grant had been done that way then you would have been able to say to me that #58 was judged to be worthy of funding, but just happened to lose out. Instead we're uncertain. Maybe #58 lost out because they asked for too much money. Maybe #58 was judged inappropriate through some kind of "4d chess" where they were better suited for another kind of grant and just using ACX grants as "backup". Maybe Scott couldn't get a good advisor to evaluate it. Or maybe on reviewing the full application (and not just this summary), Scott and his advisors saw something truly damning about the project that we're missing. We just can't tell. We do know something was wrong that made this project explicitly less fundable than all of the projects that did get funding. That doubt is what I call "bad signaling".
I think that "threshold, then randomly fund" is a great idea for a lot of reasons, though I think that saving time for the reviewers, and avoiding perverse incentives in tailoring grant writing to idiosyncracies of the reviewers, have generally felt more important to me than avoiding the negative signaling.
Well, I think that there were many projects which, like mine (#26), quite clearly said they were going to do their project independent of whether they are given money, and that money would only accelerate their schedule. So, for Scott to spend his money on it anyway would be a VERY strong endorsement. And of course, there is no such thing as bad advertisement. I remain of the opinion that Scott should have made a more elaborate and more transparent selection process where this list would be posted before the money is divided, not after, but ah well. Beggars cannot be choosers.
Besides, there's not much that an ACX grant could have done for #58; they probably need seven figures in venture capital, not a five-figure grant. Sounds like an interesting idea, though with the obvious caveat that "given this is an obviously sensible idea why is nobody already doing it?"
I'm the lead on one of the projects listed above (#57). Here's why we took Scott up on his offer of sharing a précis of our project with his readers:
When you're trying to get support for your efforts to change the world for the better through the traditional nonprofit organization model, you do a whole lot of statistically longshot things like participate in the ACX grant program or send out letters of inquiry to potential donors who have a history of supporting similar efforts. You do this, of course, because you're never going to find people who want to support your change-the-world plans if you don't get out there and first make them aware of your existence. It's why "Awareness" is at the top of a lot of marketing funnel models.
This is what makes people go on "Shark Tank" or "Top Chef" or "Project: Runway" or other competition shows even though they're risking the potential for being embarrassingly savaged in public by a judge -- they know that few participants win and most leave as losers. But they might win! And even if they don't win, they're still introducing their invention or food or clothes to a huge audience that is watching because they're interested in the topic, is more likely to be discerning about what's good or not and might come seek you out on their own if what you're doing makes a particularly personal connection with them. The relatively small risk of being meaningfully tarred with the "loser" brush in such a way that it actually hurts your long-term prospects for success is far outweighed by the likelihood of "raising awareness" and then going on from that starting point to make positive connections with others who find value in what you're doing, even if you didn't end up winning the competition.
We hoped, of course, that we'd be selected as one of the projects to receive financial support. We made our best effort to put together a proposal to do just that, knowing that the odds were stacked against us, just as we do with almost every other effort we make to try to get support for our work.
Those odds suck because there's a lot of big problems in the world, and a lot of good people with smart ideas who need support to tackle those problems. (There's dozens of them above.) I talk to a lot of people and organizations who support what we're trying to accomplish, but prioritize other problems and opportunities more highly when it comes time to allocate their resources. That's just reality. The answer is to find the people and organizations who have your issue high up on their priority list -- or can be convinced that it fits within their existing priorities -- and introduce yourself to them.
This gets to the heart of our reasoning for participating in the ACX++ component of the grant program. As I said below in my post inviting questions, my board chair and I are both SSC/ACX readers. Small sample size caveats apply, but given that two thirds of our organization's founding board are engaged with ideas and community around this blog we thought it was worth seeing if others in that community shared our passion and would be interested in joining our fight.
That's why we signed up to have Scott share our project précis. Thanks to that, there are now more rationalist and rationalist-adjacent people than before who know about or organization and the work we do. Some of those newly aware people may even be moved to support our journalist education project or other facets of our work, and that would be wonderful.
Finally, even if we never see a dime of financial support from our efforts here, there's still a non-zero number of smart and engaged people who've now become slightly more informed about the problems with the way state and municipal governments do economic development policy in America. That's an incremental step toward accomplishing our mission and vision, and that's what we're here to do.
Sometimes, investors will politely decline to comment when they choose not to invest in a company. Other times, the presenter makes a pitch that compels a response. This feels to me like the latter.
And on reading #57, I assume there was no grant initially because this is a political cause. The value of the cash given would be less important than the reputational investment of giving you cash publicly. I think Scott did not publicly make any grants of that type in the first round (though I remember two where he gave money but didn't mention the political cause).
Here, Scott can promote you without any reputational investment at all; it's possible he considers you one of the pitches he anti-endorses.
As for myself - while at a glance #57 (improve US tax and subsidy policies) sounds like an excellent proposal, it's not the type of thing I have time to investigate thoroughly enough to put money behind.
I didn't mean to give the impression that we were here for the publicity and not for the money. I was trying to answer the question about "Why would you let Scott post your non-funded project and risk being tarred as a loser for all to see?" and in that specific regard, yes, the publicity was the point. That doesn't mean it was the reason for our participation in the grant program as a whole. Frankly, we're a small group that operates on a relative shoestring for a group trying to change public policy across the country -- our budget this past year was under six figures -- so we were *very* serious about the potential impact of cold, hard ACX grant cash to our project.
As for the political issue, I'd say we're more educational than political -- Americans are being intentionally misinformed about economic issues and that's creating unnecessary problems; we're hoping to change that state of affairs through public education and advocacy. The fact that the misinformation is largely being driven by the retail political benefits it generates for elected officials across the political spectrum is relevant to how we create a strategy to address the issue, but it's not the reason we care about it.
Regardless, while I don't want to get into Kremlinology about why ACX projects were or weren't funded, I don't think Scott and his advisors shied away from politically charged topics. Winning projects included the voting reform project in Seattle, the "economically literate climate change solutions" project, hiring a lobbyist to weaponize effective altruism into policy change in Australia and even a $10,000 grant to a politician "to support his political career."
(I will jokingly suggest here that if you go looking for commonalities, the more plausible reason our proposal didn't find traction isn't that it's adjacent to politics, but rather that it's intertwined with the way journalists go about their business in the modern era, which is something of a raw subject around these parts.)
I don't want to give a wrong impression -- the $10,000 grant to a politician "to support his political career." was one of the ones I was thinking of, and Scott very explicitly did not mention whom that politician was. I assume that if he was planning to promote that politician on the blog in the future, he would not have given money. The "Australian lobbyist" was similarly vague.
(I did forget that he gave money to Yoram Bauman, which does show I was partially wrong.)
"I don't understand why anyone that's savvy would want their project listed here as being explicitly judged less worthy then the least-worthy project from the set of all of the winners of the grant. Money is the unit of caring after all!"
Value is subjective. If Scott does not care, that does not mean that everyone else will not. Also even if someone who thinks exactly like Scott about all the projects, might just want to spend a lot more money than him. And he can't spend it on the things anymore, because Scott already funded them.
Asking for money like this, is a fairly uncommon skill and this is a rare practice opportunity. And failure is valuable feedback. The ACX Grant was the first time, I actually tried to put what I'm trying into words meant for others. To persuade and explain what I'm trying to do. And not sound bonkers. And from feedback, it slowly dawned on me... how I'm really bad at it. So I wrote many more drafts and I still think... it's terrible. But at least now I'm noticing it and getting better.
So next week, I can link to a "whitepaper", that's a lot more mature than my original grant. And yeah, it's likely that it won't get funded either and that's fine. Cause I might get feedback here, that tells me exactly why my idea is unpersuasive to others. And how I can improve the presentation.
Well, anyway, we can theorize all we want, but I just got 2 mails from people who appear to be interested in my project, and I can very much use them as playtesters, which they seem willing to do. That is certainly strictly better than the counter-factual where my project was not include in this list.
I'm the writer of #11 (Preserve and Categorize Web Fiction), and I'm available in this comment thread if anyone has any questions! It's a humble project in comparison to some of the literally-world-saving-if-they-work ones, but I think it's at the intersection between my skills, my volition and the greater good (for a really nerdy definition of good).
AO3 doesn't actually allow original fiction, unless it's sufficiently "fannish". The archival of rational fiction there always been in a precarious position.
That's me! I was a bit of an idiot when I sent in my original proposal, I basically said "um, money is nice as a motivator but coding time has zero marginal cost for me and I don't know what I'd concretely spend it on, so I'm not sure how much to ask for?". Some advice if you ever ask for money: actually ask for some. I totally didn't consider the ~$1k cost of attending the conference at the time.
I've just sent in my proposal to the conference selection committee last week - you can read my submitted summary (tinyurl.com/monacosummary) and abstract (tinyurl.com/monacoabstract) at those google doc links. My plan if I don't get accepted is to return any money people want to send me (or they can make funding conditional on that acceptance).
I haven't read #16 but as a note to people outside the academic systems security community, IEEE S&P is one of the four premier conference venues (this field prefers conferences over journals) in the area (the other three venues are ACM CCS, NDSS, and Usenix Security, with CCS and S&P being primary targets for applied crypto) if that helps anybody calibrate a "how seriously should I take this proposal to decide if it's worth exploring" bar
For anyone who is curious, the "thwarting assasination plots" guy is absolutely legit. UK WIred ran an article about him: https://www.wired.co.uk/article/kill-list-dark-web-hitmen. If you are somewhat surprised, so were I; I'd say he should have worked on making the advertisement sound more believable.
Having some familiarity with the darknet, this read as plausible to me.
There are low standards for people writing about the darknet in general.
The "community" is a bit underserved with good writing, since it's very niche and users tend to be rightfully paranoid and only interested in doing whatever illegal purchase they're after.
I was unaware that someone actually managed to get the murder business going, since I thought it was all law enforcement honey pots or scams.
And that it would face the same problem like darknet gun sales.
It just wouldn't work, because you can't build a reputation via user reviews, by making good on lots of orders. Like in the drug-business.
Still no good idea, how this is solved. But I guess, it could work sometimes?
Well, the thought that there is a vigilante hacker fighting this makes me feel better about the world.
Writer of #47 here, building a better social network for gathering political ideas. Happy to discuss any questions/feedback here too.
Also to be more clear about what I'm asking for, I'm really trying to gauge whether people think this is a viable idea, as there's a lot of obstacles and substantial work involved. If there's not much support, I might backburner this idea as I think it's probably too difficult to accomplish by myself, and I have various other financial responsibilities at the moment.
Also for fun, I'll gradually post my impressions of everyone else's ideas here, though I'm not in a position to really fund much. Apologies for any harshness.
#1 Attention Hijacking: I think that nonprofits have also been using the same marketing techniques for good ends (sometimes), so I'm not sure this is necessary. Also, I'd want to know what concrete actions your movement will actually take. What's the deliverable?
#2: Pain Texture: Sounds like it could be useful, though I don't know much about medical things. Good proposal, would recommend considering.
#3: Speech acoustics: The benefit is unclear to me, how will this help people?
#4: Friendmaking handbook: Sounds nice, though dependent on your writing skill. Link to outline/draft would be nice. Unclear if you have a money target or just want encouraging emails.
#5: Sounds like a good cause, but there's already YIMBY groups with money who make videos. The fear is that you release a video that is either bad or badly promoted and gets barely any views, how will you avoid that?
#6: Maybe this isn't my field, but this just sounds like a lot of fancy phrases and I can't really tell what the product will be like.
#7: This sounds nice, but $7k is substantial and I don't see much way to verify that there were any results. Maybe some video of a class would be nice, starting with just 1 school in Nepal.
#8: Big money ask, potentially revolutionary technology worth billions. But there are already well-funded solar companies so I'd need to know why they aren't considering this. Still, props for big ideas.
#9: It's unclear to me what you will be doing, and why this should be a charitable endeavor. Can't many researchers already pay you for your service? Or will you be specifically helping researchers who have no money?
#10: A tech tree sounds kinda cool, but unclear how much funding you need to produce this and what it will be used for.
>>Maybe some video of a class would be nice. Not sure about the entire videos but may be few clips and photos are possible.
>>but $7k is substantial. This is an approximate, but it can cost to travel to Nepal. And, with this project I will have to cutoff my summer funding from my research advisor.
The target of this project is to visit as many school possible in different cities. And at the end, donate the instrument bought from the grant to one of these school. This project can be tricky in many ways. When I grew up the electricity were suspended for 10-15 hrs a day. The situation is better now but it is still unreliable and this will affect many of the demos that needs power. Or may the rural school in the mountain might not have electricity.
Re “a lot of fancy phrases”: I’m bad at narratives, and it seems trying to condense a /lot/ of parts tried over the years, which I think a researchers’ toolbox should consist of - into a small post I end up with just a soup of keywords :/
E.g. #6 includes #10 as described on website (except for the payments), and that one is already working, but it’s a small piece that got condensed to just “+Kumu”. The rest is similar.
- The idea is to help research projects set up some baseline quality assurance, like a free software repository with issue tracker (for example GitLab), automated pipeline, linters, formatters, test runners, automatic dependency updates, automatic merges, commit hooks, and the like. Individually these are relatively easy skills to pick up, but a lot of groups don't know where the really easy wins are.
- I don't have the money myself, the grant would be to be able to take a sabbatical from my day job.
- As for which projects, if they are doing anything I consider at least useful, and are not already for-profit, I'd be interested.
A bit orthogonal to what you're trying to achieve, schlaugh.com is an attempt to have a social networking site that doesn't try to be poisonously addictive. Its main feature is that its users only get one post every 24 hours, and everyone's posts for that period go up at the same time. That has the nice effect of really slowing down potential escalations.
This doesn't solve the problem you're trying to solve, but if you're interested in this slightly overlapping work that has been done so far (it's been a hobby project), I can get you in touch with the site admin. (Not expecting you to jump on this, just wanted to offer! Seemed the right thing to do.)
Going out of order to counter the first-on-the-list advantage a bit.
#60 PolicyEngine: Cool idea that is somewhat similar to what I was thinking about. However, trying out the existing app gives me pause because I feel it's pretty hard to understand what's going on, and I wonder how much impact this is having in the UK. If there's not much attention, I'm not sure retargeting it to the US will help.
#61 Evo Research: The concept of evolving artificial life simulations is interesting to me, but I don't really think "find another researcher to do it" is a good funding target.
#62 Greek/Latin Lit: Sounds like it might be a nice project, but honestly the topic just doesn't interest me and the societal benefit seems small. A smaller example deliverable for less money required might be good.
#63 Human Cryopreservation: Sounds like you're successful already, but seems like a reasonable proposal for those that are interested.
#64 Ebikes: Short on details and I feel there's a huge political wall against any biking infrastructure in US. Might be cool with more details.
#65 India Intervention: I'd be interested in the results of this, though I wonder about the ethics of pretending to be an NGO.
#66 Eyesight surgery: English errors make it sound like an email scam, and general charity is not really a good fit for ACX anyway.
Hi Anon837261 - Max from PolicyEngine here. Thanks for the call-out! We definitely agree the UI needs simplification. So far, we've focused our resources on functionality, accuracy, and training policy experts to use it.
We're also working with other parties, think tanks, and advocacy groups on various research projects. The "consumer" use case is harder, and where a lot of our effort will be, especially in the US where benefit programs are more complex. We'd like to hire a contract UX designer to help us refresh it in the coming months. Could you suggest areas that you found especially confusing that we could focus on?
The home page (policyengine.org) should probably be a landing page instead of dumping people directly into the tool without any prompt. Visitors might not know what the website is about so that should be explained first. The FAQ is quite helpful but is a very small link on the bottom. There should be a way to cancel individual proposed changes on the right, and the loading time for generating impact is quite long, though I understand if that's hard to improve.
The tool is kind of fun once you figure things out. Might be cool to eventually have some game-like objectives like "Try to balance the budget while including $X UBI and whatever constraints".
* Speed when calculating the impact: Unfortunately you're right that this isn't easy. We're simulating the reform over tens of thousands of households, and parallelizing it would be a pretty involved engineering task (https://github.com/PolicyEngine/policyengine/issues/449). We do cache results of policies that have been previously entered, so predicting policies that users will enter and pre-caching those might be more tractable (https://github.com/PolicyEngine/policyengine/issues/448).
Have you tried our AutoUBI feature? It's a button at the bottom of the UBI parameter page that redirects revenue from other policy reforms into a UBI. Our previous paper on optimizing UBI levels to minimize disruption might also interest you: https://www.ubicenter.org/uk-blank-slate-ubi
One part of our plan to gamify PolicyEngine is to build a feed of policies, quickly showing users personalized and society-wide metrics for proposals from policymakers, organizations, and individuals. Users will be able to filter this feed to policies that meet certain criteria, like budget neutrality.
A highly ambitious long-term project running on niche software, that I helped testing in a limited capacity once. I'm not into climate science and I have never figured out, if anyone finds it worthwhile.
I think I vaguely approve of your ambition. But if I try to figure out what success looks like, I end up... with some kind of holographic projection of argument space. But we don't have a way to do that, since we don't even have a formal language for arguments. So we can't properly visualize it thru symbols.
So I dunno. Can you elaborate how this looks in your head?
[Not that I'd be a sponsor, I'm next week's competition, actually. I'm just curious.]
I don't think procon.org is actually that bad, but a major difference compared to my idea is that the website seems to be written by the staff, which appears to have written relatively weak arguments for both sides, and it doesn't accurately reflect the most popular viewpoints. My idea is a social network which would allow any users to post their arguments (or potentially link to relevant outside content too)
There certainly are too many arguments, which is why it's a critical aspect to be able to sort them so people can focus on the top arguments. I think it'll be complex to figure out a ranking system that works, but one simple idea is to give all users a single upvote so they can choose their favorite single argument (in contrast to Reddit, where people can upvote as many articles as they want, in addition to downvoting). One would hope that the strongest liberal, conservative, libertarian, socialist, etc arguments all bubble up as a result.
How I imagine it looking, is like searching google for climate change solutions, except instead of giving you a dozen pages that all say nearly the same thing (and many that are not giving solutions at all), you would see a dozen pages featuring the most popular arguments. The majority opinions would be highest, but you wouldn't have to go very far down to see arguments supported by small percentages of the population. The site would also hopefully be able to format the arguments in a consistent way, and allow for responses/rebuttals for each one.
Someone making this kind of argument, it would be one about good and bad regulation. Or rather technical arguments about whether a specific piece of regulation is good.
Which could lead to an argument about public choice theory and the trustworthiness/legitimacy of current regulatory institutions. Which might lead to an argument about economic systems very different from our own.
So I think that you cannot really put a charged question in a categorical box, because the discussion is never fully about the question itself, but also about all the assumptions that go into informing your answer.
And which assumptions are even relevant.
I don't know how a difference in opinion can be represented properly.
I agree that it would be really difficult to completely map out an argument space well. However, I still think it can be done better than it is now. Currently, my main options for seeing arguments about a topic are to rely on Google, Wikipedia, or Reddit, and none of those give good results. (A big reason for that is they aren't trying to!) My website would try to improve on that experience, even if the result is still far from perfect.
Have you considered building a living library, rather than a "social" network? SE is currently the biggest player in this field https://politics.stackexchange.com/ , but more recent projects like https://codidact.com/ probably have a mission that aligns better with yours.
#40 Haven Social Network: Love the idea, though I'm not sure how it differs from wordpress or other blog software yet, so I'll have to look into the features more.
#41 Biology Videos: Of course a niche thing, but great that you're making that available for people who are interested.
#42 Architecture books: I'm not really seeing that much value for the public in this, at least not for $4000. Better English might help.
#43 Dictionaries South Africa: Sounds like a good potential cause for those interested. I don't have familiarity with the field.
#44 Israel advocacy: Pretty vague stuff, I think I'm clearly not the target here.
#45 Research in AI: Lots of possibilities but light on concrete details on what you will do.
#46 Clean up space debris: Sounds ambitious expensive, can't reasonably evaluate from a description
#48 Research Audits: Sounds like it might be a good idea, all CAPS words seem unprofessional though. $50k to start is a big ask.
#49 Fund Young People: A grant program inside a grant program, seems like an unnecessary layer. Not inherently a bad idea but maybe not a good fit for this.
#50 Charter Cities: Has gotten a lot of attention here already, not much else to say.
#51 Heavy Metal Plants: Sounds like a pretty interesting project while not asking for too much, I'd love to hear the results.
#52 Slime Mold Obesity: Covered in different comment, really good goal but written with some red flags. The name is weird too.
#53 Educational Videos: Asking for quite a lot to make videos, but channel looks legit.
#54 Promote Institutions: The idea is important, but there's not a lot of details to inspire confidence that something will actually be done.
#55 Book of case studies: Sounds like it might be interesting, but could use some writing examples and specifics on how much money will be needed and how it will be used.
#56 Aella Dating: Admirable goal, but dating websites always have huge difficulty with chicken-and-egg. Implementation of so many features is also challenging. I'd think about reducing the problem, maybe focusing on a new angle for personality tests first.
#57 No subsidies: Might be a good idea in general, but I feel skepticism that putting money into writing a guide will have much effect. As usual, prefer a specific money ask.
#58 Waste heat to energy: A good potential idea, but I expect is hoping for big donors/investors here.
#59 Rapid replications: I like this idea and they already got a grant, I'll be interested in the results of the pilot.
I'm modestly surprised Scott didn't fund Slime Mold Time Mold, as the project is both interesting and the sort of cool correlative fact I'd expect Scott to be very interested in. Maybe he could say why?
Reading the proposal and not having taken the time to read the very lengthy materials on the website yet, I did have a few worries. The sentence "it will probably take several million to cure obesity" jumped out as being somewhat outrageous, and in general it felt a little non-scientific. But if the science is actually sound, then I agree that this is a pretty interesting thing to be looking into.
My read on SMTM is that it would be more useful for their cause to say "Slate Star Codex didn't believe in us" than to say "Slate Star Codex gave us 5% of the money we needed".
Also I personally don't believe the Lithium theory, but it is plausible enough I would like to see some actual scientists do an actual scientific investigation.
I will say, tying in one of these (#52) to your post yesterday, that while I absolutely love contemporary ACX and read it religiously, it certainly has a very different vibe from very early (2013-era) SSC, which feels more free-spirited and more willing to explore odd topics than the blog as it is today. (Book review posts still do often have that feel, though, and I absolutely *loved* the Book Review Contest, which had people doing their best to apply your style of writing to many works that I doubt you'd have reviewed. Of ACX-era posts, probably the most impressive one was the ivermectin post, the one that made me think the most was "WebMD, or the Tragedy of Legible Expertise", but the one that gave me the most joy to read was the review of Arabian Nights).
Anyway, my point here is that the tone/vibe/feel of contemporary SMTM strikes me as *very similar* to the early days of SSC. So if you're pining for your lost days of youth, uh...try to be more like that.
Thanks for sharing these with us! These are all very creative and/or inspiring ventures. At the very least it's gotten me asking myself whether I should be doing something at least as inspired with my life. I'm a bit overwhelmed by how much there is, but I'll be getting back to this to see if I can pitch in some help somewhere. :)
I find that I don't have anything to say about a lot of them. Also, that many separate entries is too many. I'm going to post about them in small groups.
#4: Handbook For Making Friends In The Post-College Environment
Maybe all the people who get in touch about it should be offered a chance to get in contact with each other.
#8: Alternative Solar Power Plants
Very tentatively offered, but might it work better to have smaller lenses? They might be cheaper to manufacture.
#12: Search Engine To Analyze Research Findings
I'm curious about why Scott and Tyler think it won't work.
How about a more modest goal of improving classification to make research easier, instead of looking for new insights?
#15: Book Discovery Startup
Speaking more for myself, it might make sense to be looking at what people want to read as well as looking at the books. What attracts this particular individual? Is there something outside their usual range that would fascinate them?
For #8, they're using fresnel lenses which can be much cheaper and lighter than regular ones. Have you seen those flat plastic magnifying lenses, somewhere between the size of a credit card and a piece of paper?
#17: Algorithms To Select The Best Systematic Review
I don't have a strong opinion about your project, but "best" is awfully hard to define. I wouldn't be surprising if clipping off a few of the worst is the best you can do.
Mostly my reaction is "gleep!". Doubling every nine years? And I bet some considerable fraction more or less duplicates existing work because it's too hard to find out what's already been done.
The flood of material will only get larger and less manageable.
#19: Software For Spaced Repetition And Other Education Tech
I hope this one works out.
#48: Research Transparency AUDITS Of Published PAPERS
That doesn't look like a bad idea, but I'm not sure that increasing transparency will do a lot to improve quality.
I'm intrigued by choosing the papers to replicate by a random process rather than choosing, say, the most suspicious papers or the most cited papers. Why did you decide on a random choice?
I’m not sure I get 58. Like, it feels like if you could turn (less available) waste heat into (more available) potential energy, that would violate the second law of thermodynamics, or am I misunderstanding how this works?
You're misunderstanding how it works. You can always turn heat into work, provided you have a source of cooling at a lower temperature than your heat source. The question is *what fraction* of your heat can become work, and that depends on the temperature difference between your heat source and cooling source.
In this case, they are using heat from liquid-cooling systems for CPUs as the heat source, and the atmosphere as the cold source, so you are just running a normal heat engine. They're using a low-boiling organic compound as the working fluid, because the heat source isn't very hot, not nearly enough to boil water (which is your normal working fluid for external combustion engines).
There's nothing wrong with the idea, but the low temperature difference means only a modest fraction of the heat will get turned back into useful work. Is it worth the capital and operating costs? Who knows? Presumably they have made an effort to optimize the engineering and production costs, so that at least in some cases the value of the electricity created exceeds the amortized capital cost of the engine and generator, plus whatever the installation and maintenance costs are.
It's a solid engineering venture, and one of the few on this list that I can see actually leading to real-world success. The website, for what it's worth, hits some of what would be my major concerns as an investor, emphasizing simplicity and robustness (important to hold manufacturing, distribution/installation, and maintenance costs down), as well as hinting at some IP that would keep their lunch from being eaten by GE or some other deep-pocketed big firm as soon as they show he way. But you'd have to really dig into the details to be confident the numbers work out.
Assuming the cold end is roughly room-temperature and the hot end is 20°C above that, you can in theory get an efficiency of almost 7%. Maybe liquid cooling gives you a better delta T though.
A surprising number of these seem actively harmful. Mostly between "waste of resources that might cause some direct or indirect harms", but a few to the "kill it with fire" level.
(As opposed to the ones who got the original ACX grants, which were mostly pretty good)
#64 Ebikes - Do you have any sort of plan on how to build protected bike lanes in New York? (AFAICT the only idea that has a chance of working is to just go out and set up bollards yourself, and hope the city is too dysfunctional to tear them back down - but it's a big risk and they'd probably arrest you at some point).
#40 and #47 "Build a Better Social Network" are both similar in spirit to my MeowCat project <https://github.com/cabalamat/meowcat2> so I have emailed both of them to see if we can collaborate.
Can I suggest for you and #40 a REALLY simple how-to or guided set up tutorial? Lots of pictures showing the end user interface and then some simple steps in order to get there..
Thank you, Scott, for publishing our pitch! This is Max from PolicyEngine. Happy to answer any questions or hear feedback from any readers.
Also wanted to share an exciting update since we shared our pitch: the UK Green Party is now using PolicyEngine to evaluate and design their manifesto. To our knowledge, this is the most quantitative and interactive evaluation of a party manifesto in UK history. https://blog.policyengine.org/the-green-party-manifesto-at-policyfest-ee05a2d3b06d
Regarding #52 Slime Mould, it's true that therapeutic doses of lithium case weight gain, however, it's not nearly enough to explain the obesity epidemic (only about 10-20 pounds). And that is at doses much higher than environment doses. A chemical contaminant theory of obesity would probably have to invoke multiple chemicals in combination.
If there is ever a silver bullet against obesity - I still think it's going to be treatments that target the microbiome.
While I don't buy the lithium obesity thing remember that this is something that is building up over decades for people. If a clinical dose can gain you 10+ pounds in 5 years, possibly a subclinical dose will gain you 40 pounds over 40 years.
The 55-gallon drum version is intended for homeless or refugee camps. A drum can hold 5-10 days of urine for 20 people, and can be rolled around full on wheels with bearings without tremendous effort. Yuri can be pumped empty by the user -- unlike a porta-potty. No recurring costs to bring a pump truck out to empty it.
The version being launched will hold about a week of urine for one person. Easily maneuvered around a home.
I don't mean to give the impression that you're not doing extremely useful work. The options available for the sick or lame are miserable as it is, and it would be heroic work to give them back some dignity while they take care of business. I hope you succeed. I was just...impressed with the epic scope of the vision when I read the 55-gallon drum remark. Plus alas at that poitn a whole lot of crude humor just writes itself, as in, what uses could I think of for a giant barrel of pee? Alas, too many, all of which are socially inappropriate.
Email me to zoom chat if you'd like at david.bahry@gmail.com. I'm not an expert on mental illness but have some background in cognitive science and more in life history evolution. Also look up Dr. Randolph Nesse's (2019) "Good Reasons for Bad Feelings: Insights From the Frontier of Evolutionary Psychiatry." I haven't read it but he, along with the evolutionary biologist George C. Williams, kicked off evolutionary medicine
I'm the computational biology collaborator mentioned in #37.
For what it's worth, I think this is a great project. It is worthwhile to do it by itself (figuring out whether antibiotic resistance genes really are traveling between poultry and humans), but it would also help train people locally for other projects and I have committed to helping the bioinformatics/computational biology part of that train.
The author spent several years trying to put spaced repetition into practice in their high-school classroom, and ultimately comes to the conclusion that this is mostly a bad idea, except for learning languages. Spaced repetition is very good for memorizing things, but not so much for conceptual understanding, and in most areas the second is more useful. I'm sure this is true for research and engineering, though I'm less sure about medicine.
You can't learn dick about concepts without a body of memorized facts on which to explore them. That would be like trying to learn French literature without first learning French, an exercise in futile and sterile hand waving. Not that this stops the proposal from being a common and destructive canard in modern educational theory, one supposes because it relieves both teacher and student of some genuine hard work.
#51 (Phytoremediation): How important is the role of mycorrhyzal fungi? I know that fungi can protect birch trees living on mine spoil that would ordinarily be too toxic.
Very nice. Could you set up a poll in the second half, figure out which ones are other people most excited about or which one people think deserves the most attention?
Curious why #51 didn't make it. Given how little money he asked for, I assume there is a good reason. Is it known not to work or not to have enough effect to matter?
Sounds fake, no details for how it works, how effective it is, how hard it is to grow the plants in the relevant places, no concrete evidence of domain expertise from the grant proposer, etc.
I think the key to communicate is the massive difference is code scale. If you have 100 lines of code you can get away with a lot, but it is different even trying to grasp the architecture of 100 000+ lines of code.
#11: Preserve And Categorize Web Fiction
Yes, but how to also rate or grade? And in case anyone ever missed out
#6 looks like visions of chaos (https://softology.com.au/voc.htm) but more specific and more focused on real-world simulations instead of fractals and cellular automata
Seconded. I have no clear idea what this project is even about.
Anki exists. Better Anki exists in the form of SuperMemo, which is a UIX nightmare.
Khan Academy also already exists.
"AnkiHub will empower students by democratizing accelerated learning and potentiating the ever growing wealth of quality, free, educational resources. "
I just hear buzzword buzzword buzzword?
I agree with you that I (medical student, Anki user for 10+ years, 30k+ cards) am not clear on what this proposal is suggesting, but would just like to vouch for the team behind them- it looks from the website linked that its the guys behind the AnKing deck, which is a very well established/respected deck in the medschoolanki community.
It looks like the key offering is this platform to make real-time collaborative edits to Anki collections, which is a cool idea insofar as you could get a small group of people on your course to build a useful Anki collection together in a slightly smoother way than is possible now. Other than that I'm not clear what the rest of it (Anki Mastery course etc) adds that isn't already widely available.
I heard about that deck and I don't even study medicine! Good to hear, that there's something legit behind that. Collaborative real-time editing does sound useful, too. Maybe that's what it takes for a whole class to use the same Anki deck? Or just a couple classmates building it together, which then gets the teacher's attention. And once a whole class adopts it and they outperform, it might become very obvious that using SRS is low-hanging fruit and should be a deeply integrated into any curriculum. But currently there is not a great platform to make Anki work in a class-environment.
I kinda get, why it's so vague now.
There's almost too many interesting possibilities on how that could be developed.
And there's no real limit to the ambition behind it. Once you've got a beachhead in some uni in medicine, you spread it to other unis.
Other fields. Once you have a working platform, where teachers can quantify their student's progress, why not go for high schools?
Of course, this might not work out, since the SRS revolution could have happened 20 years ago, but it's great someone taking a shot at that. Still think, they could have done a better job communicating. I would have sold this way differently to this audience, cause the way it's phrased just triggers everyone's bs detectors here. Almost tempted to try.
I'm all ears if you have any suggestions for how I can avoid sounding like a maniac when I say, "trust me, this could revolutionize education." I feel like this is something people aren't going to really grok the potential of until they can actually try it out. "Make it really easy to curate and study a comprehensive set of highly individualized digital flashcards for anything you want," just doesn't quite convey the spirit of the thing... But I thought it was worth a try!
I wrote down some pointers or something to exemplify a possible approach.
But I'm getting sick of writing publically here. Found an email address, but I'm not 100% sure, if that is you. Are you a Peggy Bia award winner?
Don't want to send the wrong Andrew Sanchez a very confusing email.
That's very kind of you! I'm ashamed to say that I'm the non-Peggy Bia award winner. inbox.asanchez@gmail.com
Although I'm on board with the idea for collaborative editing, I'm always wary of people talking about SRS having the potential to "revolutionise" education, which I feel like kind of oversimplifies a) what SRS is good and bad at and b) what education/learning/knowledge/competence really is.
This is a really interesting account from a (SRS-loving) teacher who tried to use Anki in the classroom and found it to be of very limited value https://www.lesswrong.com/posts/F6ZTtBXn2cFLmWPdM/seven-years-of-spaced-repetition-software-in-the-classroom-1. Until that's the kind of discussion to be had I find it hard to get ridiculously excited about this proposal- keen to hear what Andrew has to say via email though (joepusey1 at gmail.com).
Sure, I'll take a whack at elaborating. Collaboration is just one of many things in Anki that is super tedious. Utilizing pre-made decks from Ankiweb or elsewhere introduces another major problem for students: selecting precisely the cards they need, at precisely the moment they need them. This problem actually makes pre-made decks very difficult to use effectively. The gold standard set of Anki flashcards for medical students has over 40K cards. The current workflow med students use to take advantage of these pre-made cards is a nightmare. So one of the first features we would likely work on after collaboration is streamlining the scheduling of cards based on information about what students are currently studying. For example, students could collaborate on lecture notes in a Google Doc, an Anki add-on could take that as input and use it to search a database of all prompts, and schedule the best matches for that group of students.
Although Anki has a plugin-based architecture, add-on authors are extremely limited by the Python environment shipped with Anki. This means no NLP libraries, no statistical computing libraries, etc. So AnkiHub will also enable a web-based plugin model with far less limited computing environments that have access to a massive database of all the cards on AnkiHub.
Here is a list of other things that the platform can facilitate once we make using Anki more joyful:
- Curating a database of review stats for thousands of students
- open sourcing such a database could lead to interesting discoveries in SRS, such as optimizations for the scheduling algorithms
- Enabling smarter scheduling algorithms
- e.g., an algorithm that was aware not only of your individual reviews/ratings but of all reviews from all other students studying the same prompts
- Automated flashcard generation using ML/NLP
- A deadline-aware scheduling algorithm
- To ensure students study their prompts as much as needed before a test, for example
- Automated image occlusion
- Input an image with text labels and output prompts by occluding the text
- A suggestion engine for discovering new prompts to add to your queue
- This could be based on inputs like youtube history, articles ready in Pocket, etc.
- Integration with https://hypothes.is
- This is an easy pathway to an incremental reading workflow
- Also enables flashcards with links to primary sources so the greater context isn't lost in the minutiae
- Crowd curation of high-quality mnemonics
- Integration with ontology frameworks like https://protege.stanford.edu/
- Flashcard quality control
- automatically identify duplicate/similar prompts in a users collection and merge them into away in order to avoid redundancy/studying more than needed
- automatically suggested tweaks for a more quality card
- Generate mindmaps/zettelkastens from a collection of cards
- This would enable intuitive exploration of a collection of cards, addressing the common complaint of losing sight of the bigger picture.
- Auto discover and link to external sources like wikimedia, etc
- cards can be powered up by a bot that autolinks to relevant resources
Also, this reddit post might be of interest for a little more background and discussion: https://www.reddit.com/r/medicalschoolanki/comments/i2evp1/introduction_to_anki_palace/
Regarding the course: we have also published tons of free Anki learning resources, not to mention that we give out scholarships for the course for anybody who can't afford it. It's no secret that anybody can learn what they need for free from our YouTube channel and elsewhere. People buy the course because it offers a guided path to learn Anki quickly. We also provide individual support and a forum for people to ask for help as well as an add-on that configures Anki settings automatically.
Thanks so much for fleshing out some of the Anki palace ideas. I think some of them sound really cool, although I'm instinctively kind of wary that there's so many of massively varying complexity/difficulty.
Broadly, I think massing stats from large numbers of learners could feasibly tell us something about optimal scheduling, although the style of card, threshold for hitting different values of remembering and even differences between individual memory makes this very fuzzy data.
Some easy workflow for incremental reading->Ankifying is the holy grail that I and a lot of people have been searching for for a long time, and I've never seen anything come really close to it yet- off the top of my head, getpolarized.io looked to be doing interesting things, but I suspect its a very tiny market.
Some of your other suggestions, like generating mindmaps from cards/automated flashcard generation, I feel like underestimate the significant import of a) the process of producing, rather than just reviewing cards and b) the quality of the cards themselves- very very few people make cards of sufficiently quality to even check off all the 20 rules, never mind genuinely enable the kind of real long-term knowledge we're all looking for. Until we're better at that, I'm not sure how easy it is to automate the process- although of course the payoff could be substantial.
I don't want to sound negative, because I'm really pleased that people are doing things in the SRS space. I think used right it can be a really powerful tool, but that an understanding of Anki's limitations and correct use is both vitally important and pretty rare amongst the people I see using it day-to-day.
+1
I'm going to try commenting to these one at a time.
#1: A Movement To Fight Attention Hijacking
Oh, God, no. People's attention is being hijacked enough, How about a project to teach people to protect their attention and think about whether what they're doing makes sense. Remember that some of your benevolent uses might be getting things wrong.
One at a time! There's 66 applications! Good luck.
#2: Understand The Texture Of Pain
Sounds cool. I don't if you want to expand your subject, but one vagueness about a stabbing pain in the shoulder is that people mean different things by shoulder.
There are a lot of good ideas here. There are also a few bad ones; though I'm not going to point them out.
If I were a billionaire, I would give each of them $10k each, no questions asked. (Signaling risk is real.) Of course, since I'm not a billionaire, that's fairly easy for me to say.
I expect I will be emailing half-a-dozen of these people, a mix of "how can I help" and "would you like an investment -- and I do want my money back".
The total: 9 emails sent, one (#41) I would have emailed if they asked for anything, and about 5 where the "if this works, somebody else will give you better terms than I can, so I won't bother" principle applies.
I did get one reply to an email where I offered money, asking "I am not sure if [direct payment] is the right thing to do, or the money has to come from ACX." To be clear, I assume Scott doesn't want to be in the middle of any transactions here?
Hey Alex, I’m the person behind #41. Thanks for your interest! I’m not looking for any money, since that isn’t the limiting factor to what I’m doing right now.
But if you have any advice or criticism of my project, you can reach me at mikest@udel.edu.
Specifically, I’m trying to figure out what topics are the most useful to cover, and how to reach more people with my videos. So please let me know if you have any advice about that. Thanks!
No, it's definitely okay to do it directly. If for some reason one of you *insists* on having me as a middleman, email me.
#3: Acoustics Of Historical Speeches
Note that the wind matters for speeches made outdoors.
Does anybody else immediately think of the scene in "Life of Brian" where the people at the back of the crowd can't hear the Sermon on the Mount?
I don't know what everyone else is thinking, but I was probably influenced by an NPR piece about sound and battlefields. Until amplified sound and radio, the military was limited to what it could quite literally see and hear.
#21: Thwart Darknet Murder Plots
This one is a joke to see if I was paying attention, right?
The google sheet on the website of that person links to media coverage of murders apparently thwarted, so maybe not.
Agree; looks legit.
as i already said elsewhere in the thread, no: https://www.wired.co.uk/article/kill-list-dark-web-hitmen
Re. the one on the texture of pain: I like the idea. I just want to highlight how helpful it would be in medical practice to have a device that can measure pain. We still have no practical (eg non-fMRI) way of determining when unconscious or uncommunicative people are in pain, nor of measuring pain levels. So after they characterize the relation between textures in functional imaging, and perceived pain, it would be useful to search for correlations of fMRI-detected pain with some signal that can be found in blood samples (say, detection of endorphins), galvanic response, pulse, or skin temperature, or other simple assays.
An objective measurement of pain would help a lot with physical therapy studies, which currently rely on patient-reported subjective pain, which can change a lot from day to day because perceived pain is relative to recent events and state of mind. (BTW, the texture study might try to factor that kind of context out of its pain measurements. I mean factors like time of day, season of the year, current weather, and stress hormone levels.)
An objective measurement of pain could hinder actually getting physical therapy, though, if insurance companies decree that patients must stop receiving PT when their objectively-measured pain stops decreasing. Pro tip: When you start receiving PT, say your perceived pain is at least a 5 on the 1-10 scale. Every time you get re-evaluated to determine whether you still need PT, you need to report a lower pain level (to show the PT is working)... but this boxes you into a corner if you tell them your pain is a 3 at the start (which people often do because the level of pain that brings people in for PT is much less than intense pain from acute injuries); after 2 re-evals, you can't qualify anymore no matter what answers you give.
The fact that the only way to currently measure pain is to ask someone to subjectively provide a number (or point at a sad face!) bothers me constantly. If we could objectively measure pain it would solve so many problems. Just think of all the people who suffer chronic pain but don't have their problem taken seriously because there's no obvious cause! Or, alternatively, we could prove those people are actually in no more pain than normal and they can go see a psychologist. I guess that would depend on how it works.
Obligatory xkcd on the subjectivity of pain rating: https://xkcd.com/883/
I guess one way to go would be to define a set of calibration pain sources, but that has obvious ethical limitations as well as dual use concerns. Of course, even assuming that one could induce similar neurological signals, it could be that different brains react very differently to them.
Or it could even be that pain is not scalar at all, e.g. that there is no inter-human fixed preference ordering between nth degree burns, kidney stones, etc. And this is before we even get to emotional pain.
I think this would be relevant not just with pain, but with other sorts of subjective symptoms. Over the past two years I feel like I've grown more aware of the differences between cough caused by allergies and by generic dust or liquid stimulation (but I've also managed not to have any viral coughs, so I don't know as well whether those feel different). It seems like it could be useful for some people to be trained to know the difference between symptoms, even if most people don't pay attention.
(And of course there's an XKCD for this too: https://www.explainxkcd.com/wiki/index.php/2535:_Common_Cold_Viruses)
#15: Book Discovery Startup
This sounds almost identical to the old Book Genome Project (circa 2013) on Booklamp.com, which aimed to use semantic analysis to categorize books the way Pandora categorized music and use those to build recommendations. That project and company got quietly bought out by Apple Books as soon as their recommendation engine got good (it wasn't announced until months later), and the website was shuttered; I've never heard a word about it or the Book Genome Project again.
I'd be curious to see if Shepherd.com a) gets successful, and b) gets bought out silently the same way.
Good recommendation systems would wreck the financial planning of the publishing industry, by flattening out the distribution of sales per book. Advertising has an S-shaped ROI curve; very small and very large marketing campaigns are both low ROI. But "very small" still means very large relative to the profits on most books. One rule-of-thumb is that 50% of all books lose money, and 5% make money. This article on Forbes claims over 99% of titles published lose money, but it's counting specialty publications like agency reports: https://www.forbes.com/sites/quora/2014/05/28/what-is-the-percentage-of-books-published-on-which-publishers-actually-lose-money/?sh=1592aef36d58
The purpose of recommendation systems is to destroy the power-law distribution of sales per title (a miniscule fraction of titles account for most profits). But that power-law distribution is profitable for sellers, because it makes advertising more cost-effective. So big sellers like Amazon probably don't want recommendation systems. Notice that Amazon book recommendations are usually of best-sellers.
Remember the Netflix movie-recommendation competition? They paid out a million dollars to the winning team, but ended up not using it at all, because they realized that effective personalized recommendations weren't as profitable as strategic marketing recommendations.
> Good recommendation systems would wreck the financial planning of the publishing industry, by flattening out the distribution of sales per book.
Aren't you assuming that all books are equally interesting here? You're also assuming that being able to talk about books with other people easily is not a factor when getting a book recommandation.
Good questions. I'm not assuming those things, because this is not a Boolean decision between having the current power-law distribution, and a completely flat distribution. The goal of recommendation systems should be to flatten that power-law distribution somewhat, so that it gets closer to the normal distribution which book quality should have. Not all books are equally interesting, but the current distribution of book sales matches the distribution predicted by increasing-returns models in which quality has no effect on sales, and cannot be reconciled with any sane normal distribution of book quality.
You're right that there are increasing returns to the reader wrt how many people they can talk to about a book. But I think (but can't prove) that's only a small part of what produces the power-law distribution of sales.
Where did you hear that about Netflix?
Do we know they didn't use the techniques? Surely they heard about them while the contest was running?
The story I heard is that the big change was the move to streaming. This changed everything. The data was now in a very different shape than it had been before. There was a lot more data, but also it was revealed preference. And, the catalogue contracted.
It was in an article about the competition, written shortly after it was done. What I wrote about Netflix deciding not to use it was based on a quote from a Netflix employee, but that person didn't come right out and say "it's more profitable to do X". They said something a little vague, which I interpreted at the time as meaning the winning recommendation system wasn't profitable. But I don't remember where I read it, so I can't check that now.
Regarding being bought out, or more generally, pivoting to providing a shittier (but more profitable) service, similar comments apply to Aella's dating app (#56) and the 2 social networks (#40, #47). It would be really good if there was a social / legal / whatever mechanism such that websites were incentivized to provide a good service for their users and not chase continual growth/profits.
Oddly enough, there's this long-standing hypothesis that the *way* you get continual growth and profit is by providing a good service for your users.
In defense of the free market, the perfect is the enemy of the good. There /are/ mechanisms that incentivize good service; but they can't work perfectly without perfect competition, which we certainly don't have with Apple music, Amazon books, Netflix streaming services, or the Google Maps API.
I partly blame legislation that big companies have been busy getting passed these past few decades, which make start-ups expensive or even impossible. Especially recent Internet regulations on taxes, tax reporting, observance of laws in the state and county being delivered to (eg selling liquor on the Internet), observance of European Union laws (notably the ones to "protect" minors), and so on. Also, the interpretation of patent law, the switch to first-to-file patent priority, and excessive regulation of biotech, to name a few.
But ultimately, perfection is impossible. These companies provide services that make some compromise between the customer's interests, and their own. Most of the solutions proposed for these problems require decisions to be made by government, without considering the customer's interests at all.
Hi, I'm the author of #40/Haven. I'm trying really hard to structure this differently. I'm not raising any money. The software is open source and the protocols are dead-simple (it's just RSS). The emphasis really is on some form of self-hosting. I expect this to generally take the form of paid hosting. My hosting offering puts each Haven on it's own dedicated VPS--and there is no barrier to other entities taking the software I write and offering their own hosting, or whole different implementations that speak the same protocol. My biggest fear is if Haven catches on enough for some company to start offering a "free" version with all the tracking/ads/censorship/invasions-of-privacy that are the status quo today.
There's a whole world of non-corporate possibility that opens up if we can find some way to make self-hosting (in some form) feasible to the broader population. Haven is my attempt to make progress on that front.
Thanks for including my project here Scott. If anyone has any questions or advice about mine (making free online tutorials for Bioinformatics and Computational Biology) please let me know! I don't need any money for this, but definitely appreciate advice and constructive criticism.
(Number 41, for reference)
Hi Mike, I think your videos are great and I've actually used a few to help me learn things. I don't have any concrete advice, but I thought you'd appreciate having at least one supporter.
Thanks, yes I definitely appreciate the support! Also please let me know if there are any specific topics you'd like to see covered.
Likewise, I have no advice, but I wish your project the best.
Thanks! 🙂
Haha, losing line breaks was a bit unexpected %) I'll take the liberty to post a version of #6 above, but with line breaks as-submitted :D
--
I'm Oleksandr Nikitin, and I want to build a system dynamics simulator.
Enable independent researchers to simulate, forecast, and visualize metabolic pathways, epidemic spread, mass transit, ecology, macroeconomics, etc. Show, don't tell. Without code.
Think Airtable+Vensim+Roam+Kumu, integrated and working offline.
Why offline? Why simulate? Why a new tool?
- Complex systems must be simulated. You miss emergent phenomena if you analyze parts separately or simplify the details.
- Offline sets you free from distractions and groupthink. Free to make your own breakthroughs.
- Take your references, notes and data with you, dive deep, then return with the verified, reproducible, interactive model.
- Research can take years. Tools should outlast devices and app stores.
- And it must be fast. Isn’t it insane for a productivity tool to make you wait?
I spent years on prototypes, tested in companies since 2013, now I want to put these experiments together. Not as a startup, but as a tool accessible to everyone.
The plan:
- Create a community of curious, inquisitive makers.
- Empower them with small, fast and robust core app.
- Iterate and grow together, augment the human intelligence even more.
- Understand the world.
I seek funding to focus on this project full-time, for two years.
To launch and to guide people to the finished research.
Sounds inspiring? Worth the funding? Want more details?
Ping me at oleksandr@tvori.info. Also see https://cortex.substack.com/
My immediate observation is that IMHO a reasonable model for funding something like this is how many other tools have been made, namely, when someone gets a grant to do research where this tool would be useful (e.g. employing a few grad students/postdocs to do system dynamics modeling for a few years) then they also use part of that funding/manpower to develop such a tool or improve an existing one. However, for that to work there needs to be some basis of core functionality and structure, and that basis needs to be open source (otherwise we just get a new iteration of Vensim), so the licensing policy choice - which is not discussed in the description - is quite relevant to how the project is be positioned; developing a startup product is quite different from developing an open source community.
I'm happy to answer questions anyone might have about #57, "Advocate Against Subsidies And Tax Breaks For Local Corporations."
I'm the president of the Center for Economic Accountability, the 501(c)(3) that's putting together the "Skeptical Reporter's Guide to Covering Economic Development." Our board chair and I are both regular SSC/ACX readers, and we're grateful to Scott for the opportunity to spread the word about our work to an audience that's predisposed to actually pay attention to research findings.
I'm naturally skeptical of opportunities like these. I oppose many (though not all) tax breaks and subsidies for large corporations. However, I'm not convinced this will be effective. Most corporations' governmental benefits seem to flow in because of the company's ability to directly pay for influence, nit because of independent reporters spreading the wrong story. As an example, during early stages of the pandemic, I felt like every news article I read was outraged over commercial airlines' tax breaks — yet despite all of the bad press, they (and companies like them) continue to get big breaks. I am willing to acknowledge that:
1) I don't read the news much, so maybe most news is super in favor of big business' tax breaks. This doesn't match my personal experience, but it's definitely possible.
2) I almost never read strongly conservative news sources for opinions about taxes / subsidies. Perhaps this praise of government tax cuts is common, it's just tied to conservative news sources.
If it's the first, I'd be interested in seeing examples of major news sources praising subsidies and etc. If it's the second, I'd really doubt that telling reporters to stop praising tax breaks would overcome whatever political purpose praising tax breaks serves, though I'd be interested in at least hearing the theory.
The problem isn’t that any one newspaper story is cheerleading for any one subsidy deal. The problem is that as a matter of practice, media coverage of these deals consistently report as fact things that we know to be at least unlikely and at worst untrue about the impact that subsidies have on corporate site selection decisions. Since local media tends to be the only source of information most people will encounter about these deals, the result is a climate of public opinion in which misinformed voters reward politicians who take credit for “creating jobs” through subsidies, despite evidence that these economic development programs are at best ineffective and at worst can be extremely harmful to the communities they’re supposed to be helping. (For more on this: https://economicaccountability.org/research-resources/)
Reporters, by and large, want to get the story right. (I used to be one, decades ago.) But overwhelmingly, what we’ve found from years of reading story after story after story about the most recent deal is that they’re *not* getting the story right, because they’re not asking the right questions and including the right context.
Consider the recent headline “Boosted by state incentives, General Motors making $7B investment in Michigan, creating 4,000 jobs.” The story begins, “General Motors, building toward an all-electric future, plans to invest $7 billion and create 4,000 jobs across four manufacturing sites in Michigan. Supported by $824.1 million in state incentives, the Detroit automaker will expand an Orion Township plant, build a Lansing electric battery manufacturing facility and make upgrades to two Lansing area plants.” (https://www.mlive.com/public-interest/2022/01/boosted-by-state-incentives-general-motors-making-7b-investment-in-michigan-creating-4000-jobs.html)
The clear implication of "boosted" and "supported by" here is that the incentives changed GM’s decision about where to invest $7 billion in new and upgraded plants. But if you look at the actual dollar amounts in context, that idea’s laughable. In reality, the $824 million in subsidies is less than three days’ worth of revenues to GM, and it’s being spread out over a decade or more. It's an insignificant blip on the company's bottom line. We saw the same thing in North Carolina, where $830 million in subsidies to Apple spread over 30 years were supposed to have sealed the deal for a company that was booking more than a billion dollars a day in revenues.
The idea that a major corporation is going to base a decision about where to locate mission-critical production facilities on the equivalent of less than a couple hours’ worth of additional corporate revenues a year, rather than the best place based on actual business-critical factors such as workforce availability and cost, is just ridiculous. It's not even borne out by the site selection industry's own surveys. But that’s what gets reported as fact, over and over again, which is a big reason why the average person believes that it’s true.
Reporters also rarely include context about historical performance of these programs. If a particular tax only brought in a third as much every year as the state was budgeting for it to collect, that relevant context would be included in stories about the tax. But while it’s extremely common for official audits to find economic development agencies rife with mismanagement and only reporting half to a third of contractually agreed-upon jobs -- and that's if you give them credit for every single job being the result of a subsidy. For instance, a 2017 audit of the Wisconsin Economic Development Corporation found that the WEDC's recordkeeping and oversight was so bad that it "cannot be certain about the numbers of jobs created or retained as a result of its awards." (https://legis.wisconsin.gov/lab/media/2627/17-9full.pdf) Later that year, the WEDC was entrusted with overseeing what would become one of the greatest economic development boondoggles of the modern era, the Foxconn project. if media coverage of WEDC deals at the time had included the context that the state's own auditors were screaming from the rooftops about the agency's pervasive inability to do something as simple as get companies to report whether they'd created jobs or not, maybe there would have been less support for the disastrous Foxconn deal.
(In 2019, the next round of audits found that the WEDC had improved to the point where it could actually report a job creation number; the bad news is that the number was only 34.9%: https://legis.wisconsin.gov/lab/media/2861/19-6full.pdf)
So, we’re setting out to educate reporters on, among other things, how to find those statistics, put big dollar amounts in context and ask questions that get to the truth of why companies actually make site selection decisions and who’s really benefiting from these massive subsidies. When reporters do this, good things can happen. We’ve seen in places like Texas (https://www.houstonchronicle.com/news/investigations/article/unfair-burden-part-1-texas-tax-corporations-covid-16164744.php) and New Jersey (https://www.propublica.org/article/george-norcross-democratic-donor-tax-breaks) where informed and engaged journalists helped drive changes in state policy by pulling back the curtains on massive boondoggles.
#1. Attention hijacking...
Wasn't priming, the idea by Yale's prof Bargh, debunked? It is not replicable. Isn't the first idea based on that?
Ofcourse, it might still be true, even if not scientifically provable.
I spent about a year building and running a bespoke dating site with all kinds of interesting ways to match people. None of that matters - there's really only one important problem in online dating, and that's "how do I bootstrap a critical mass of attractive women on the site?" Personality tests (what drove OKCupid's early success) won't work anymore because Facebook has dialed the virality of these back to near-zero, and most people are burnt out on that kind of thing anyway.
Not that I think it's impossible to start a new dating site, but you have to start with the chicken-egg problem and ignore _everything else_ until you solve that. Most people who want to build dating sites spend all their time thinking about the matching stuff, because that's more fun. But it's the wrong problem.
better question is "how do I get a critical mass of women who will match with my male users". there's definitely untapped potential in the "dating app specifically for not-hot people" market for whoever can figure out how to make that sound not degrading
From what I gather, the major problem for dating sites today is trying to sort out the "only looking for casual sex and ideally a range of partners" from "want to date and maybe turn this into a long term thing".
The second problem, as you say, is "how do we get enough women on here?"
And the third problem is "how do we match up the women who want something more than 'fuck me tonight' with the guys who are the same, and avoid all the 750k horny guys just looking for some action?' because I think that drives women off.
> Sending abusive messages, money scams, fishing for validation
Or for friendship, which is more wholesome than your examples
> the major problem for dating sites today is trying to sort out the "only looking for casual sex and ideally a range of partners" from "want to date and maybe turn this into a long term thing".
The biggest problem is that individuals in your userbase don't even have very accurate guesses as to which of those two categories they're in. There are plenty of people who say they're interested in one but are actually interested in the other, in both directions. There's also people whose preferences go back and forth and they don't know which mood they're in tonight.
I still think personality tests have a good shot; personality test websites (i have a friend who runs one) gets hella traffic still; people share screenshots, and do it over platforms that aren't facebook. I also have some experience doing organic marketing over female-heavy platforms too (like tiktok).
I don't think 'critical mass' is 50%; most dating sites are functional with an imbalance of male/female users. It's hard to get around the "fewer matches will be made" with the imbalance, but there's lots of ways to get around the imbalance in in-app attention.
How many unique users are in hella traffic? What seems like a lot of users in a national or worldwide context ends up being fairly thin when you look at "who can I date in my local area?"
I also worked on Let's Date, an early Tinder competitor started by the guy behind suicidegirls. Seeded by that community and a million bucks in facebook advertising (aimed only at women), it reached maybe 100k users? But you still ran out of matches pretty quickly, even in urban areas. And the ad spend was not sustainable.
I agree that the balance ratio doesn't really matter... in as much that the men are irrelevant and you can ignore them in our your outreach. If your site has women, there will always be plenty of men. It sounds crude, but it's true.
I don't want to dishearten you - I'd love to see someone to take down IAC.
I ran out of matches pretty quickly on okcupid, but this seemed fine to me? As in, there were like maybe 10-15 eligible males in the area I wanted to date, and i would go message all of them and maybe go on dates with a few, or something. And I'd check back occasionally to see if anybody new was in the area, or answered more questions to update the match score, or of course if I was traveling for anything. This seems like the way most people used okc, and it seemed to function; I think the point isn't that you should get a lot of matches, but that you should get a few really high-potential matches. Of course I agree getting more women is better, and getting more women is hard. I do intend this to be 'niche targeted', at least at first; as in, I want to target communities where women will be organically interested in this sort of thing, and more likely to be compatible with the men who sign up.
OKCupid had a critical mass, definitely. In my mind it was the peak of online dating, though that may be an age-related judgement. I just think the number of people you need in the system to have 10-15 eligible candidates is pretty large, especially if you're filtering them by personality characteristics and not surface appearance. And getting that critical mass is the core problem because the viral mechanisms that existed in the mid-2000s don't really work anymore (or at least, don't work anywhere near as well).
Another complication is that back in the mid-2000s the free dating sites like OKC and Plenty of Fish supported themselves by advertising paid dating sites like Match.com. Now IAC owns everything and doesn't need to advertise, so that business model is probably over. Hosting costs have come down but programmers are more expensive, if anything.
I feel like okcupid's emphasis on personality tests worked great in the late 00s when online dating was more stigmatised, because you could lie to yourself and say that you were "only there for the tests".
I really don't think you can look at personality tests in isolation. Personality tests worked in the 00s because people posted results on their Facebook wall, which drew other people to the tests. The important part is not the test itself but the mechanism by which people found out about the test.
If your viral coefficient is greater than one, your userbase grows. If it's less than one, it shrinks. Sharing results of personality tests increases the viral coefficient. Taking a test and keeping the results to yourself does not.
I just don't think people share like that anymore? Maybe people are burnt out on "which character on star trek" etc they are. Maybe Facebook's algorithms have squelched most of that content. Maybe people are more concerned about privacy. It's possible that my cohort just outgrew that kind of thing and "kids these days" are still posting that stuff on snapwhatever, but I kinda doubt it.
My guess is that there's a community of people who love personality tests, and maybe you'll get some viral spread in that community via the few remaining viral mediums, but it won't be possible to achieve liftoff. So you're stuck trying to supplement the viral effects with traditional methods like advertising.
I could be wrong - I could imagine a new personality test that is so interesting and compelling that it erupts into public consciousness the way people are currently sharing Wordle scores. But deliberately crafting successful viral memes is insanely hard.
I'm an ignorant person who never used a dating site. I guess it's supposed to be obvious that "old" OkCupid was better than the current version, but I don't know why that is supposed to be the case.
Girls love personality tests. But a lot of those on OKC turned out to be just "Will I date you? Answer these 6 questions to find out..." Which suggests a real business opportunity: a site ("hoops.com" or maybe "onlyfans++.com") in which women can list a set of actions suitors can undertake to earn steps in some graduated sequence of interaction. 1. Send me a photo of yourself wearing your favorite article of women's clothing to earn a chat. 2. Here's a photo of my eyebrow, upload your woeful ballad to earn a 20 minute Zoom. Et cetera.
1. Send me a photo of yourself wearing a sark you made 'owt stich or needle-work
2. Upload a video of you ploughing the beach with a piece of leather then sow a peppercorn 'neath mean high water...
That's the spirit!
Yes! Please make this happen.
This has been a passion project I've had in my head for a long while, and I don't even have a use for this myself.
Trying to resist the urge to start developing it right now.
Pontifex already said this in another comment, but the biggest concern I have is that if this takes off, the usual path is to either get bought by one of the big players or become exactly like them. Which would be a shame as it would be a wasted effort then. And like Pontifex said, I would want there to be some mechanism to prevent this.
If you don't get funding for this you could go open source and ask some of the 750k horny men to contribute.
Or you could develop it yourself: tutorial.djangogirls.org
And finally, what's the best way to stay up to date on this?
Good luck, I hope you succeed!
What's wrong with "become exactly like them"?
What has happened to okcupid anyhow? Wikipedia mentions that it was acquired by match.com in 2011, which was around the time I stopped using it because I found a relationship in meatspace. From the post, it is regarded mostly unusable now?
(From what I remember, okc was one of the less emotionally painful to use dating sites I used, possibly because the number of matches was small. And they showed the reply rates. I think I even got replies. Looking back in general, I think non-cynical online dating was probably the quickest way to nurse insecurity and self-doubt to a certainty of worthlessness. Short of stating a deeply unpopular opinion on twitter, perhaps. I found the micromarriage to micromort rate was really unfavorable for accumulating the former, and many activities are obviously more efficient (and less soul-sucking) for accumulating the latter.)
"#8: Help Fund Eyesight Restoration Surgery"
Call me a heartless bitch, but how is this not one of those scam email type appeals about "I am poor sick widow in Third World need urgent medical help God bless you" that I've been pestered with?
I must note that there are two #8s in the list.
There are indeed - well spotted
Numbering things is hard, sorry.
Are we allowed to praise and/or deride these proposals ? Because I'm itching to do both...
Sure.
I think an index could help.
Re #10 tech trees: That's a great idea, we had successfully collaborated in this format previously :) UX is not that trivial to design though
Does listing order indicate your evaluation of them? If not may want to mention how they are sorted so being no. 1 isn't taken as endorsement and vice versa
Based on the initial qualifications, I'm pretty sure order is uncorrelated with evaluation.
#17: Algorithms To Select The Best Systematic Review: I don't think this is the sort of thing that machine learning can currently do. Humans are probably still better at evaluating methodologies. And I think the whole system will probably rely on having humans select the best reviews, in order to train the ML algorithm. Why not just have those humans say why they selected those reviews as best?
Nor do I even think there is one best way of doing systematic reviews. It might be better to have a variety of review methodologies and rely on the biases cancelling each other out--which random-forest machine learning has proven to be surprisingly effective.
Mike Saint-Antoine's channel (#41 on this list) is quite good, I've actually used a few of his videos before!
Thanks Metacelsus! 🙂
I’m definitely finding I can now abstain from certain kinds of clickbait. I see the click bait and I immediately think about how most journalists have nothing of value to say. I find via that thought, I avoid the pull.
Happy to answer any questions for those curious about #19, AnkiHub (education technology, spaced repetition, etc). Also, check out our page at https://courses.ankipalace.com/ankihub to get updates about the project.
Oh hey, didn't see you there. We kinda talked about your back downthread a little bit, sorry.
If you want to chime in on that, just ctrl+f #19.
Thanks. Responded below.
It seems to me like this is extraordinary bad signaling for these projects -- you're essentially saying, "Here are projects that I've judged as being unworthy of receiving grant money, go ahead and pick through the slop and see if there's anything you might be interested in."
I don't understand why anyone that's savvy would want their project listed here as being explicitly judged less worthy then the least-worthy project from the set of all of the winners of the grant. Money is the unit of caring after all!
In retrospect, one way to prevent this signaling risk while still being able to show the losers of the grant process here would be to have the grant awarding process be one where submissions have to meet a certain "very happy-to-fund" standard by direct evaluation, and then randomly allocating funds among the projects that met that standard, while publishing the remaining projects explicitly as projects that definitely could have received funding but simply missed out through luck of the draw. It also would have made the grant process probably easier to administer, as you don't have to compare projects against each other to determine a ranking, just evaluate each against the "very happy-to-fund" standard.
The alternative is that Scott's large audience never sees these projects at all, so I'm not surprised lots of applicants decided to roll the dice.
That's not the alternative at all! They could just post their projects to the comment section under the open threads instead and still reach that audience, without associating their project with losing the grant and incurring negative signaling.
But people actually looking to fund something are more likely to look at this thread than another open thread. Especially if someone decides several months from now that they'd like to fund something, they may well come back to this post but are unlikely to go trawling through the other open threads.
I second this
I don't think people would view it that way, I see it as cheap advertisement that meets a very low bar of not being literal trolling. I think there actually may some value in a public post like this, especially since some of these proposals are very similar. There can be quite some value in connecting these people to each-other.
Presumably many of these projects would have preferred to get funding directly, but there are reasons other than "this is a bad idea" these might not have been funded.
The best example I saw was #58 (waste heat in data centers) - if you actually own a data center, this could be a wonderful opportunity. And Scott's readership does include people at the companies that run data centers (Google, Amazon, etc.). However, if Google, Amazon, etc. all don't like it, you probably shouldn't invest either.
I think we're in agreement here? Rephrasing what you've said, you agree that among other reasons, "this is a bad idea" might be one of the reasons they didn't get funded. That's precisely what I call "bad signaling".
What do you think of the idea of "threshold, then randomly fund"?
If the grant had been done that way then you would have been able to say to me that #58 was judged to be worthy of funding, but just happened to lose out. Instead we're uncertain. Maybe #58 lost out because they asked for too much money. Maybe #58 was judged inappropriate through some kind of "4d chess" where they were better suited for another kind of grant and just using ACX grants as "backup". Maybe Scott couldn't get a good advisor to evaluate it. Or maybe on reviewing the full application (and not just this summary), Scott and his advisors saw something truly damning about the project that we're missing. We just can't tell. We do know something was wrong that made this project explicitly less fundable than all of the projects that did get funding. That doubt is what I call "bad signaling".
I think that "threshold, then randomly fund" is a great idea for a lot of reasons, though I think that saving time for the reviewers, and avoiding perverse incentives in tailoring grant writing to idiosyncracies of the reviewers, have generally felt more important to me than avoiding the negative signaling.
Well, I think that there were many projects which, like mine (#26), quite clearly said they were going to do their project independent of whether they are given money, and that money would only accelerate their schedule. So, for Scott to spend his money on it anyway would be a VERY strong endorsement. And of course, there is no such thing as bad advertisement. I remain of the opinion that Scott should have made a more elaborate and more transparent selection process where this list would be posted before the money is divided, not after, but ah well. Beggars cannot be choosers.
Besides, there's not much that an ACX grant could have done for #58; they probably need seven figures in venture capital, not a five-figure grant. Sounds like an interesting idea, though with the obvious caveat that "given this is an obviously sensible idea why is nobody already doing it?"
I'm the lead on one of the projects listed above (#57). Here's why we took Scott up on his offer of sharing a précis of our project with his readers:
When you're trying to get support for your efforts to change the world for the better through the traditional nonprofit organization model, you do a whole lot of statistically longshot things like participate in the ACX grant program or send out letters of inquiry to potential donors who have a history of supporting similar efforts. You do this, of course, because you're never going to find people who want to support your change-the-world plans if you don't get out there and first make them aware of your existence. It's why "Awareness" is at the top of a lot of marketing funnel models.
This is what makes people go on "Shark Tank" or "Top Chef" or "Project: Runway" or other competition shows even though they're risking the potential for being embarrassingly savaged in public by a judge -- they know that few participants win and most leave as losers. But they might win! And even if they don't win, they're still introducing their invention or food or clothes to a huge audience that is watching because they're interested in the topic, is more likely to be discerning about what's good or not and might come seek you out on their own if what you're doing makes a particularly personal connection with them. The relatively small risk of being meaningfully tarred with the "loser" brush in such a way that it actually hurts your long-term prospects for success is far outweighed by the likelihood of "raising awareness" and then going on from that starting point to make positive connections with others who find value in what you're doing, even if you didn't end up winning the competition.
We hoped, of course, that we'd be selected as one of the projects to receive financial support. We made our best effort to put together a proposal to do just that, knowing that the odds were stacked against us, just as we do with almost every other effort we make to try to get support for our work.
Those odds suck because there's a lot of big problems in the world, and a lot of good people with smart ideas who need support to tackle those problems. (There's dozens of them above.) I talk to a lot of people and organizations who support what we're trying to accomplish, but prioritize other problems and opportunities more highly when it comes time to allocate their resources. That's just reality. The answer is to find the people and organizations who have your issue high up on their priority list -- or can be convinced that it fits within their existing priorities -- and introduce yourself to them.
This gets to the heart of our reasoning for participating in the ACX++ component of the grant program. As I said below in my post inviting questions, my board chair and I are both SSC/ACX readers. Small sample size caveats apply, but given that two thirds of our organization's founding board are engaged with ideas and community around this blog we thought it was worth seeing if others in that community shared our passion and would be interested in joining our fight.
That's why we signed up to have Scott share our project précis. Thanks to that, there are now more rationalist and rationalist-adjacent people than before who know about or organization and the work we do. Some of those newly aware people may even be moved to support our journalist education project or other facets of our work, and that would be wonderful.
Finally, even if we never see a dime of financial support from our efforts here, there's still a non-zero number of smart and engaged people who've now become slightly more informed about the problems with the way state and municipal governments do economic development policy in America. That's an incremental step toward accomplishing our mission and vision, and that's what we're here to do.
Sometimes, investors will politely decline to comment when they choose not to invest in a company. Other times, the presenter makes a pitch that compels a response. This feels to me like the latter.
And on reading #57, I assume there was no grant initially because this is a political cause. The value of the cash given would be less important than the reputational investment of giving you cash publicly. I think Scott did not publicly make any grants of that type in the first round (though I remember two where he gave money but didn't mention the political cause).
Here, Scott can promote you without any reputational investment at all; it's possible he considers you one of the pitches he anti-endorses.
As for myself - while at a glance #57 (improve US tax and subsidy policies) sounds like an excellent proposal, it's not the type of thing I have time to investigate thoroughly enough to put money behind.
I didn't mean to give the impression that we were here for the publicity and not for the money. I was trying to answer the question about "Why would you let Scott post your non-funded project and risk being tarred as a loser for all to see?" and in that specific regard, yes, the publicity was the point. That doesn't mean it was the reason for our participation in the grant program as a whole. Frankly, we're a small group that operates on a relative shoestring for a group trying to change public policy across the country -- our budget this past year was under six figures -- so we were *very* serious about the potential impact of cold, hard ACX grant cash to our project.
As for the political issue, I'd say we're more educational than political -- Americans are being intentionally misinformed about economic issues and that's creating unnecessary problems; we're hoping to change that state of affairs through public education and advocacy. The fact that the misinformation is largely being driven by the retail political benefits it generates for elected officials across the political spectrum is relevant to how we create a strategy to address the issue, but it's not the reason we care about it.
Regardless, while I don't want to get into Kremlinology about why ACX projects were or weren't funded, I don't think Scott and his advisors shied away from politically charged topics. Winning projects included the voting reform project in Seattle, the "economically literate climate change solutions" project, hiring a lobbyist to weaponize effective altruism into policy change in Australia and even a $10,000 grant to a politician "to support his political career."
(I will jokingly suggest here that if you go looking for commonalities, the more plausible reason our proposal didn't find traction isn't that it's adjacent to politics, but rather that it's intertwined with the way journalists go about their business in the modern era, which is something of a raw subject around these parts.)
I don't want to give a wrong impression -- the $10,000 grant to a politician "to support his political career." was one of the ones I was thinking of, and Scott very explicitly did not mention whom that politician was. I assume that if he was planning to promote that politician on the blog in the future, he would not have given money. The "Australian lobbyist" was similarly vague.
(I did forget that he gave money to Yoram Bauman, which does show I was partially wrong.)
Maybe I don’t get the signaling game at all, but I def don’t see how it could be a problem.
It’s a nice presentation exercise, live-fire and safe at the same time, in a very welcoming community. What’s not to like? :)
"I don't understand why anyone that's savvy would want their project listed here as being explicitly judged less worthy then the least-worthy project from the set of all of the winners of the grant. Money is the unit of caring after all!"
Value is subjective. If Scott does not care, that does not mean that everyone else will not. Also even if someone who thinks exactly like Scott about all the projects, might just want to spend a lot more money than him. And he can't spend it on the things anymore, because Scott already funded them.
Asking for money like this, is a fairly uncommon skill and this is a rare practice opportunity. And failure is valuable feedback. The ACX Grant was the first time, I actually tried to put what I'm trying into words meant for others. To persuade and explain what I'm trying to do. And not sound bonkers. And from feedback, it slowly dawned on me... how I'm really bad at it. So I wrote many more drafts and I still think... it's terrible. But at least now I'm noticing it and getting better.
So next week, I can link to a "whitepaper", that's a lot more mature than my original grant. And yeah, it's likely that it won't get funded either and that's fine. Cause I might get feedback here, that tells me exactly why my idea is unpersuasive to others. And how I can improve the presentation.
Well, anyway, we can theorize all we want, but I just got 2 mails from people who appear to be interested in my project, and I can very much use them as playtesters, which they seem willing to do. That is certainly strictly better than the counter-factual where my project was not include in this list.
I'm the writer of #11 (Preserve and Categorize Web Fiction), and I'm available in this comment thread if anyone has any questions! It's a humble project in comparison to some of the literally-world-saving-if-they-work ones, but I think it's at the intersection between my skills, my volition and the greater good (for a really nerdy definition of good).
What does it mean to be an "anti-archivist"? Like how do you feel about museums?
What's wrong with AO3? I feel like the internet has already solved this.
AO3 doesn't actually allow original fiction, unless it's sufficiently "fannish". The archival of rational fiction there always been in a precarious position.
#29: Present An Open-Source Python Library For Monte Carlo Techniques
https://github.com/scottshambaugh/monaco
That's me! I was a bit of an idiot when I sent in my original proposal, I basically said "um, money is nice as a motivator but coding time has zero marginal cost for me and I don't know what I'd concretely spend it on, so I'm not sure how much to ask for?". Some advice if you ever ask for money: actually ask for some. I totally didn't consider the ~$1k cost of attending the conference at the time.
I've just sent in my proposal to the conference selection committee last week - you can read my submitted summary (tinyurl.com/monacosummary) and abstract (tinyurl.com/monacoabstract) at those google doc links. My plan if I don't get accepted is to return any money people want to send me (or they can make funding conditional on that acceptance).
An attempt at some organisation. (each number is placed in the first category of the list that fits if multiple categories would fit):
Science: #2, #9, #10, #12, #14, #25, #31, #45, #48, #55, #59, #65
Engineering: #6, #8 (the second), #16, #17, #24, #27, #28, #29, #30, #33, #35, #39, #46, #51, #58
Rationality: #1, #56
Humanities: #3, #11, #42, #62
Education: #5, #7, #19, #23, #26, #32, #41, #53
Society: #4, #15, #34, #36, #37, #38, #40, #44, #47, #50, #54, #57, #60, #63, #64
Altruism: #18, #20, #43
Personal Funding: #8 (the first), #13, #22, #49, #52, #61
Other: #21
I haven't read #16 but as a note to people outside the academic systems security community, IEEE S&P is one of the four premier conference venues (this field prefers conferences over journals) in the area (the other three venues are ACM CCS, NDSS, and Usenix Security, with CCS and S&P being primary targets for applied crypto) if that helps anybody calibrate a "how seriously should I take this proposal to decide if it's worth exploring" bar
For anyone who is curious, the "thwarting assasination plots" guy is absolutely legit. UK WIred ran an article about him: https://www.wired.co.uk/article/kill-list-dark-web-hitmen. If you are somewhat surprised, so were I; I'd say he should have worked on making the advertisement sound more believable.
That's a fascinating article, thank you.
Having some familiarity with the darknet, this read as plausible to me.
There are low standards for people writing about the darknet in general.
The "community" is a bit underserved with good writing, since it's very niche and users tend to be rightfully paranoid and only interested in doing whatever illegal purchase they're after.
I was unaware that someone actually managed to get the murder business going, since I thought it was all law enforcement honey pots or scams.
And that it would face the same problem like darknet gun sales.
It just wouldn't work, because you can't build a reputation via user reviews, by making good on lots of orders. Like in the drug-business.
Still no good idea, how this is solved. But I guess, it could work sometimes?
Well, the thought that there is a vigilante hacker fighting this makes me feel better about the world.
The article linked said it was a scam but after getting scammed the hirers sometimes commit the murders themselves. No actual murder-for-hire.
Evil people still failing to cooperate? I'm glad to hear.
There is a cursed Cunningham's law opportunity here
I'll take your word on there being a curse and end my foolish musings on the topic :)
Writer of #47 here, building a better social network for gathering political ideas. Happy to discuss any questions/feedback here too.
Also to be more clear about what I'm asking for, I'm really trying to gauge whether people think this is a viable idea, as there's a lot of obstacles and substantial work involved. If there's not much support, I might backburner this idea as I think it's probably too difficult to accomplish by myself, and I have various other financial responsibilities at the moment.
Also for fun, I'll gradually post my impressions of everyone else's ideas here, though I'm not in a position to really fund much. Apologies for any harshness.
#1 Attention Hijacking: I think that nonprofits have also been using the same marketing techniques for good ends (sometimes), so I'm not sure this is necessary. Also, I'd want to know what concrete actions your movement will actually take. What's the deliverable?
#2: Pain Texture: Sounds like it could be useful, though I don't know much about medical things. Good proposal, would recommend considering.
#3: Speech acoustics: The benefit is unclear to me, how will this help people?
#4: Friendmaking handbook: Sounds nice, though dependent on your writing skill. Link to outline/draft would be nice. Unclear if you have a money target or just want encouraging emails.
#5: Sounds like a good cause, but there's already YIMBY groups with money who make videos. The fear is that you release a video that is either bad or badly promoted and gets barely any views, how will you avoid that?
#6: Maybe this isn't my field, but this just sounds like a lot of fancy phrases and I can't really tell what the product will be like.
#7: This sounds nice, but $7k is substantial and I don't see much way to verify that there were any results. Maybe some video of a class would be nice, starting with just 1 school in Nepal.
#8: Big money ask, potentially revolutionary technology worth billions. But there are already well-funded solar companies so I'd need to know why they aren't considering this. Still, props for big ideas.
#9: It's unclear to me what you will be doing, and why this should be a charitable endeavor. Can't many researchers already pay you for your service? Or will you be specifically helping researchers who have no money?
#10: A tech tree sounds kinda cool, but unclear how much funding you need to produce this and what it will be used for.
#7
>>Maybe some video of a class would be nice. Not sure about the entire videos but may be few clips and photos are possible.
>>but $7k is substantial. This is an approximate, but it can cost to travel to Nepal. And, with this project I will have to cutoff my summer funding from my research advisor.
The target of this project is to visit as many school possible in different cities. And at the end, donate the instrument bought from the grant to one of these school. This project can be tricky in many ways. When I grew up the electricity were suspended for 10-15 hrs a day. The situation is better now but it is still unreliable and this will affect many of the demos that needs power. Or may the rural school in the mountain might not have electricity.
Re “a lot of fancy phrases”: I’m bad at narratives, and it seems trying to condense a /lot/ of parts tried over the years, which I think a researchers’ toolbox should consist of - into a small post I end up with just a soup of keywords :/
E.g. #6 includes #10 as described on website (except for the payments), and that one is already working, but it’s a small piece that got condensed to just “+Kumu”. The rest is similar.
Can you take a look at a bit more background at https://cortex.substack.com/p/augmenting-the-flow and comment on that pine of thinking?
Re. #9:
- The idea is to help research projects set up some baseline quality assurance, like a free software repository with issue tracker (for example GitLab), automated pipeline, linters, formatters, test runners, automatic dependency updates, automatic merges, commit hooks, and the like. Individually these are relatively easy skills to pick up, but a lot of groups don't know where the really easy wins are.
- I don't have the money myself, the grant would be to be able to take a sabbatical from my day job.
- As for which projects, if they are doing anything I consider at least useful, and are not already for-profit, I'd be interested.
You might look into the Software Sustaiability Institute (at https://www.software.ac.uk/about), which seems to have similar aims.
A bit orthogonal to what you're trying to achieve, schlaugh.com is an attempt to have a social networking site that doesn't try to be poisonously addictive. Its main feature is that its users only get one post every 24 hours, and everyone's posts for that period go up at the same time. That has the nice effect of really slowing down potential escalations.
This doesn't solve the problem you're trying to solve, but if you're interested in this slightly overlapping work that has been done so far (it's been a hobby project), I can get you in touch with the site admin. (Not expecting you to jump on this, just wanted to offer! Seemed the right thing to do.)
And good luck! :)
Going out of order to counter the first-on-the-list advantage a bit.
#60 PolicyEngine: Cool idea that is somewhat similar to what I was thinking about. However, trying out the existing app gives me pause because I feel it's pretty hard to understand what's going on, and I wonder how much impact this is having in the UK. If there's not much attention, I'm not sure retargeting it to the US will help.
#61 Evo Research: The concept of evolving artificial life simulations is interesting to me, but I don't really think "find another researcher to do it" is a good funding target.
#62 Greek/Latin Lit: Sounds like it might be a nice project, but honestly the topic just doesn't interest me and the societal benefit seems small. A smaller example deliverable for less money required might be good.
#63 Human Cryopreservation: Sounds like you're successful already, but seems like a reasonable proposal for those that are interested.
#64 Ebikes: Short on details and I feel there's a huge political wall against any biking infrastructure in US. Might be cool with more details.
#65 India Intervention: I'd be interested in the results of this, though I wonder about the ethics of pretending to be an NGO.
#66 Eyesight surgery: English errors make it sound like an email scam, and general charity is not really a good fit for ACX anyway.
Hi Anon837261 - Max from PolicyEngine here. Thanks for the call-out! We definitely agree the UI needs simplification. So far, we've focused our resources on functionality, accuracy, and training policy experts to use it.
Despite the complexity, we've had some success in the UK; for example, the Green Party is using PolicyEngine to evaluate and develop their manifesto: https://blog.policyengine.org/the-green-party-manifesto-at-policyfest-ee05a2d3b06d
We're also working with other parties, think tanks, and advocacy groups on various research projects. The "consumer" use case is harder, and where a lot of our effort will be, especially in the US where benefit programs are more complex. We'd like to hire a contract UX designer to help us refresh it in the coming months. Could you suggest areas that you found especially confusing that we could focus on?
The home page (policyengine.org) should probably be a landing page instead of dumping people directly into the tool without any prompt. Visitors might not know what the website is about so that should be explained first. The FAQ is quite helpful but is a very small link on the bottom. There should be a way to cancel individual proposed changes on the right, and the loading time for generating impact is quite long, though I understand if that's hard to improve.
The tool is kind of fun once you figure things out. Might be cool to eventually have some game-like objectives like "Try to balance the budget while including $X UBI and whatever constraints".
Thanks for this! Agree on all, here are issues on these suggestions and some comments:
* Landing page: 100% needed (https://github.com/PolicyEngine/policyengine/issues/444).
* Reset parameter change: Absolutely (https://github.com/PolicyEngine/policyengine/issues/107).
* FAQ: Glad it's helpful - we'd like to add tooltips in parameters and variables to give more info without needing people to renavigate (https://github.com/PolicyEngine/policyengine/issues/445), but we'll look into making the FAQ more prominent until then (https://github.com/PolicyEngine/policyengine/issues/446).
* Speed when calculating the impact: Unfortunately you're right that this isn't easy. We're simulating the reform over tens of thousands of households, and parallelizing it would be a pretty involved engineering task (https://github.com/PolicyEngine/policyengine/issues/449). We do cache results of policies that have been previously entered, so predicting policies that users will enter and pre-caching those might be more tractable (https://github.com/PolicyEngine/policyengine/issues/448).
Have you tried our AutoUBI feature? It's a button at the bottom of the UBI parameter page that redirects revenue from other policy reforms into a UBI. Our previous paper on optimizing UBI levels to minimize disruption might also interest you: https://www.ubicenter.org/uk-blank-slate-ubi
One part of our plan to gamify PolicyEngine is to build a feed of policies, quickly showing users personalized and society-wide metrics for proposals from policymakers, organizations, and individuals. Users will be able to filter this feed to policies that meet certain criteria, like budget neutrality.
How would your site be different from this:
https://www.procon.org/
Because, I think this is pretty terrible?
So you want to like... completely map out the argument space around climate change? I think... that's too ambitious?
There is too many arguments and most of them are stupid, for one.
But also... I don't know. Maybe you have a vision in your head how this looks like.
There's this thing:
https://www.theclimateweb.com/
There's a passionate maintainer behind this.
A highly ambitious long-term project running on niche software, that I helped testing in a limited capacity once. I'm not into climate science and I have never figured out, if anyone finds it worthwhile.
I think I vaguely approve of your ambition. But if I try to figure out what success looks like, I end up... with some kind of holographic projection of argument space. But we don't have a way to do that, since we don't even have a formal language for arguments. So we can't properly visualize it thru symbols.
So I dunno. Can you elaborate how this looks in your head?
[Not that I'd be a sponsor, I'm next week's competition, actually. I'm just curious.]
I don't think procon.org is actually that bad, but a major difference compared to my idea is that the website seems to be written by the staff, which appears to have written relatively weak arguments for both sides, and it doesn't accurately reflect the most popular viewpoints. My idea is a social network which would allow any users to post their arguments (or potentially link to relevant outside content too)
There certainly are too many arguments, which is why it's a critical aspect to be able to sort them so people can focus on the top arguments. I think it'll be complex to figure out a ranking system that works, but one simple idea is to give all users a single upvote so they can choose their favorite single argument (in contrast to Reddit, where people can upvote as many articles as they want, in addition to downvoting). One would hope that the strongest liberal, conservative, libertarian, socialist, etc arguments all bubble up as a result.
How I imagine it looking, is like searching google for climate change solutions, except instead of giving you a dozen pages that all say nearly the same thing (and many that are not giving solutions at all), you would see a dozen pages featuring the most popular arguments. The majority opinions would be highest, but you wouldn't have to go very far down to see arguments supported by small percentages of the population. The site would also hopefully be able to format the arguments in a consistent way, and allow for responses/rebuttals for each one.
Thanks!
So when we had an issue like climate change, people might argue about nuclear power. Then they might argue about price and risk.
I recently read this book review:
https://rootsofprogress.org/devanney-on-the-nuclear-flop
Someone making this kind of argument, it would be one about good and bad regulation. Or rather technical arguments about whether a specific piece of regulation is good.
Which could lead to an argument about public choice theory and the trustworthiness/legitimacy of current regulatory institutions. Which might lead to an argument about economic systems very different from our own.
So I think that you cannot really put a charged question in a categorical box, because the discussion is never fully about the question itself, but also about all the assumptions that go into informing your answer.
And which assumptions are even relevant.
I don't know how a difference in opinion can be represented properly.
I agree that it would be really difficult to completely map out an argument space well. However, I still think it can be done better than it is now. Currently, my main options for seeing arguments about a topic are to rely on Google, Wikipedia, or Reddit, and none of those give good results. (A big reason for that is they aren't trying to!) My website would try to improve on that experience, even if the result is still far from perfect.
Yeah ok, that makes sense.
Also somehow the original online comment got doubleposted. And then I deleted one, it doubledeleted them.
Substack delenda est.
Have you considered building a living library, rather than a "social" network? SE is currently the biggest player in this field https://politics.stackexchange.com/ , but more recent projects like https://codidact.com/ probably have a mission that aligns better with yours.
Continued impressions:
#40 Haven Social Network: Love the idea, though I'm not sure how it differs from wordpress or other blog software yet, so I'll have to look into the features more.
#41 Biology Videos: Of course a niche thing, but great that you're making that available for people who are interested.
#42 Architecture books: I'm not really seeing that much value for the public in this, at least not for $4000. Better English might help.
#43 Dictionaries South Africa: Sounds like a good potential cause for those interested. I don't have familiarity with the field.
#44 Israel advocacy: Pretty vague stuff, I think I'm clearly not the target here.
#45 Research in AI: Lots of possibilities but light on concrete details on what you will do.
#46 Clean up space debris: Sounds ambitious expensive, can't reasonably evaluate from a description
#48 Research Audits: Sounds like it might be a good idea, all CAPS words seem unprofessional though. $50k to start is a big ask.
#49 Fund Young People: A grant program inside a grant program, seems like an unnecessary layer. Not inherently a bad idea but maybe not a good fit for this.
#50 Charter Cities: Has gotten a lot of attention here already, not much else to say.
#51 Heavy Metal Plants: Sounds like a pretty interesting project while not asking for too much, I'd love to hear the results.
#52 Slime Mold Obesity: Covered in different comment, really good goal but written with some red flags. The name is weird too.
#53 Educational Videos: Asking for quite a lot to make videos, but channel looks legit.
#54 Promote Institutions: The idea is important, but there's not a lot of details to inspire confidence that something will actually be done.
#55 Book of case studies: Sounds like it might be interesting, but could use some writing examples and specifics on how much money will be needed and how it will be used.
#56 Aella Dating: Admirable goal, but dating websites always have huge difficulty with chicken-and-egg. Implementation of so many features is also challenging. I'd think about reducing the problem, maybe focusing on a new angle for personality tests first.
#57 No subsidies: Might be a good idea in general, but I feel skepticism that putting money into writing a guide will have much effect. As usual, prefer a specific money ask.
#58 Waste heat to energy: A good potential idea, but I expect is hoping for big donors/investors here.
#59 Rapid replications: I like this idea and they already got a grant, I'll be interested in the results of the pilot.
I'm modestly surprised Scott didn't fund Slime Mold Time Mold, as the project is both interesting and the sort of cool correlative fact I'd expect Scott to be very interested in. Maybe he could say why?
Reading the proposal and not having taken the time to read the very lengthy materials on the website yet, I did have a few worries. The sentence "it will probably take several million to cure obesity" jumped out as being somewhat outrageous, and in general it felt a little non-scientific. But if the science is actually sound, then I agree that this is a pretty interesting thing to be looking into.
My read on SMTM is that it would be more useful for their cause to say "Slate Star Codex didn't believe in us" than to say "Slate Star Codex gave us 5% of the money we needed".
Also I personally don't believe the Lithium theory, but it is plausible enough I would like to see some actual scientists do an actual scientific investigation.
Hi Scott, for #5, the link is broken: it should go to https://twitter.com/michael_wiebe/status/1455999023375011842, but your link has a ').' appended.
Sorry, fixed.
> 34: Outline A Potential Martian Legal System
"Legal Systems Very Distant From Our Own"
I will say, tying in one of these (#52) to your post yesterday, that while I absolutely love contemporary ACX and read it religiously, it certainly has a very different vibe from very early (2013-era) SSC, which feels more free-spirited and more willing to explore odd topics than the blog as it is today. (Book review posts still do often have that feel, though, and I absolutely *loved* the Book Review Contest, which had people doing their best to apply your style of writing to many works that I doubt you'd have reviewed. Of ACX-era posts, probably the most impressive one was the ivermectin post, the one that made me think the most was "WebMD, or the Tragedy of Legible Expertise", but the one that gave me the most joy to read was the review of Arabian Nights).
Anyway, my point here is that the tone/vibe/feel of contemporary SMTM strikes me as *very similar* to the early days of SSC. So if you're pining for your lost days of youth, uh...try to be more like that.
Thanks for sharing these with us! These are all very creative and/or inspiring ventures. At the very least it's gotten me asking myself whether I should be doing something at least as inspired with my life. I'm a bit overwhelmed by how much there is, but I'll be getting back to this to see if I can pitch in some help somewhere. :)
I find that I don't have anything to say about a lot of them. Also, that many separate entries is too many. I'm going to post about them in small groups.
#4: Handbook For Making Friends In The Post-College Environment
Maybe all the people who get in touch about it should be offered a chance to get in contact with each other.
#8: Alternative Solar Power Plants
Very tentatively offered, but might it work better to have smaller lenses? They might be cheaper to manufacture.
#12: Search Engine To Analyze Research Findings
I'm curious about why Scott and Tyler think it won't work.
How about a more modest goal of improving classification to make research easier, instead of looking for new insights?
#15: Book Discovery Startup
Speaking more for myself, it might make sense to be looking at what people want to read as well as looking at the books. What attracts this particular individual? Is there something outside their usual range that would fascinate them?
> Maybe all the people who get in touch about it should be offered a chance to get in contact with each other.
Shhh, don't reveal the trick!
For #8, they're using fresnel lenses which can be much cheaper and lighter than regular ones. Have you seen those flat plastic magnifying lenses, somewhere between the size of a credit card and a piece of paper?
#17: Algorithms To Select The Best Systematic Review
I don't have a strong opinion about your project, but "best" is awfully hard to define. I wouldn't be surprising if clipping off a few of the worst is the best you can do.
Mostly my reaction is "gleep!". Doubling every nine years? And I bet some considerable fraction more or less duplicates existing work because it's too hard to find out what's already been done.
The flood of material will only get larger and less manageable.
#19: Software For Spaced Repetition And Other Education Tech
I hope this one works out.
#48: Research Transparency AUDITS Of Published PAPERS
That doesn't look like a bad idea, but I'm not sure that increasing transparency will do a lot to improve quality.
#59: Rapid Replications Of Newly Published Papers
I'm intrigued by choosing the papers to replicate by a random process rather than choosing, say, the most suspicious papers or the most cited papers. Why did you decide on a random choice?
I’m not sure I get 58. Like, it feels like if you could turn (less available) waste heat into (more available) potential energy, that would violate the second law of thermodynamics, or am I misunderstanding how this works?
You're misunderstanding how it works. You can always turn heat into work, provided you have a source of cooling at a lower temperature than your heat source. The question is *what fraction* of your heat can become work, and that depends on the temperature difference between your heat source and cooling source.
In this case, they are using heat from liquid-cooling systems for CPUs as the heat source, and the atmosphere as the cold source, so you are just running a normal heat engine. They're using a low-boiling organic compound as the working fluid, because the heat source isn't very hot, not nearly enough to boil water (which is your normal working fluid for external combustion engines).
There's nothing wrong with the idea, but the low temperature difference means only a modest fraction of the heat will get turned back into useful work. Is it worth the capital and operating costs? Who knows? Presumably they have made an effort to optimize the engineering and production costs, so that at least in some cases the value of the electricity created exceeds the amortized capital cost of the engine and generator, plus whatever the installation and maintenance costs are.
It's a solid engineering venture, and one of the few on this list that I can see actually leading to real-world success. The website, for what it's worth, hits some of what would be my major concerns as an investor, emphasizing simplicity and robustness (important to hold manufacturing, distribution/installation, and maintenance costs down), as well as hinting at some IP that would keep their lunch from being eaten by GE or some other deep-pocketed big firm as soon as they show he way. But you'd have to really dig into the details to be confident the numbers work out.
Assuming the cold end is roughly room-temperature and the hot end is 20°C above that, you can in theory get an efficiency of almost 7%. Maybe liquid cooling gives you a better delta T though.
A surprising number of these seem actively harmful. Mostly between "waste of resources that might cause some direct or indirect harms", but a few to the "kill it with fire" level.
(As opposed to the ones who got the original ACX grants, which were mostly pretty good)
The one massive exception, afaict, is #52, which has a nontrivial chance of actually doing a large amount of good.
#64 Ebikes - Do you have any sort of plan on how to build protected bike lanes in New York? (AFAICT the only idea that has a chance of working is to just go out and set up bollards yourself, and hope the city is too dysfunctional to tear them back down - but it's a big risk and they'd probably arrest you at some point).
#40 and #47 "Build a Better Social Network" are both similar in spirit to my MeowCat project <https://github.com/cabalamat/meowcat2> so I have emailed both of them to see if we can collaborate.
Can I suggest for you and #40 a REALLY simple how-to or guided set up tutorial? Lots of pictures showing the end user interface and then some simple steps in order to get there..
#60: PolicyEngine
Thank you, Scott, for publishing our pitch! This is Max from PolicyEngine. Happy to answer any questions or hear feedback from any readers.
Also wanted to share an exciting update since we shared our pitch: the UK Green Party is now using PolicyEngine to evaluate and design their manifesto. To our knowledge, this is the most quantitative and interactive evaluation of a party manifesto in UK history. https://blog.policyengine.org/the-green-party-manifesto-at-policyfest-ee05a2d3b06d
Regarding #52 Slime Mould, it's true that therapeutic doses of lithium case weight gain, however, it's not nearly enough to explain the obesity epidemic (only about 10-20 pounds). And that is at doses much higher than environment doses. A chemical contaminant theory of obesity would probably have to invoke multiple chemicals in combination.
If there is ever a silver bullet against obesity - I still think it's going to be treatments that target the microbiome.
While I don't buy the lithium obesity thing remember that this is something that is building up over decades for people. If a clinical dose can gain you 10+ pounds in 5 years, possibly a subclinical dose will gain you 40 pounds over 40 years.
In re #39 I'm awed by the vision of a 55-gallon drum of piss by my bed, and a husky robot that will go dump it for me.
Good point about the weight. How about a version that just holds a night or two nights amount.
They're already on it. But it's good to know they've also got me covered if I drink a whole six pack after dinner.
Hey, author of #39 here.
The 55-gallon drum version is intended for homeless or refugee camps. A drum can hold 5-10 days of urine for 20 people, and can be rolled around full on wheels with bearings without tremendous effort. Yuri can be pumped empty by the user -- unlike a porta-potty. No recurring costs to bring a pump truck out to empty it.
The version being launched will hold about a week of urine for one person. Easily maneuvered around a home.
I don't mean to give the impression that you're not doing extremely useful work. The options available for the sick or lame are miserable as it is, and it would be heroic work to give them back some dignity while they take care of business. I hope you succeed. I was just...impressed with the epic scope of the vision when I read the 55-gallon drum remark. Plus alas at that poitn a whole lot of crude humor just writes itself, as in, what uses could I think of for a giant barrel of pee? Alas, too many, all of which are socially inappropriate.
Haha no worries! Just took the opportunity to expand on the vision.
The husky robot comment made me think of the Rick and Morty gag "What is my purpose?"
Re: #2, understand the texture of pain:
For interest look up the "Schmidt sting pain index" where he describes bug stings as fruity, spicy, etc. https://theconversation.com/amp/suffering-for-science-why-i-have-insects-sting-me-to-create-a-pain-index-50853
Maybe also hot sauce fans would be good at describing hot sauce pains
Re: #25, life history models of mental illness;
Email me to zoom chat if you'd like at david.bahry@gmail.com. I'm not an expert on mental illness but have some background in cognitive science and more in life history evolution. Also look up Dr. Randolph Nesse's (2019) "Good Reasons for Bad Feelings: Insights From the Frontier of Evolutionary Psychiatry." I haven't read it but he, along with the evolutionary biologist George C. Williams, kicked off evolutionary medicine
#3: Did Henry V really give a famous speech at the Battle of Agincourt? I thought that was Shakespeare's invention.
#39: Portable Urinal For Disabled Adults
Looks like a sensible, useful project.
I'm the computational biology collaborator mentioned in #37.
For what it's worth, I think this is a great project. It is worthwhile to do it by itself (figuring out whether antibiotic resistance genes really are traveling between poultry and humans), but it would also help train people locally for other projects and I have committed to helping the bioinformatics/computational biology part of that train.
You can reach out to Emmanuel directly, but I'm also reachable at luis@luispedro.org (and see https://luispedro.org/)
#19 seems similar to the project described in these lesswrong posts
https://www.lesswrong.com/posts/Ww2dxwWpSfkQB4NZb/a-year-of-spaced-repetition-software-in-the-classroom (2015)
https://www.lesswrong.com/posts/dtCfxYubZgRnEkGpQ/a-second-year-of-spaced-repetition-software-in-the-classroom (2016)
https://www.lesswrong.com/posts/F6ZTtBXn2cFLmWPdM/seven-years-of-spaced-repetition-software-in-the-classroom-1 (2021)
The author spent several years trying to put spaced repetition into practice in their high-school classroom, and ultimately comes to the conclusion that this is mostly a bad idea, except for learning languages. Spaced repetition is very good for memorizing things, but not so much for conceptual understanding, and in most areas the second is more useful. I'm sure this is true for research and engineering, though I'm less sure about medicine.
You can't learn dick about concepts without a body of memorized facts on which to explore them. That would be like trying to learn French literature without first learning French, an exercise in futile and sterile hand waving. Not that this stops the proposal from being a common and destructive canard in modern educational theory, one supposes because it relieves both teacher and student of some genuine hard work.
Can you please post the obvious trolls in another post? Your readership is a very creative bunch.
#51 (Phytoremediation): How important is the role of mycorrhyzal fungi? I know that fungi can protect birch trees living on mine spoil that would ordinarily be too toxic.
Very nice. Could you set up a poll in the second half, figure out which ones are other people most excited about or which one people think deserves the most attention?
Curious why #51 didn't make it. Given how little money he asked for, I assume there is a good reason. Is it known not to work or not to have enough effect to matter?
Sounds fake, no details for how it works, how effective it is, how hard it is to grow the plants in the relevant places, no concrete evidence of domain expertise from the grant proposer, etc.
#22 If anyone wants to reach me I'm at zoharatkins at gmail dot com
#8: Alternative Solar Power Plants
Nice but are you aware of the the field of https://en.wikipedia.org/wiki/Concentrator_photovoltaics ?
#9: Help Research Teams Improve Software Quality
I think the key to communicate is the massive difference is code scale. If you have 100 lines of code you can get away with a lot, but it is different even trying to grasp the architecture of 100 000+ lines of code.
#11: Preserve And Categorize Web Fiction
Yes, but how to also rate or grade? And in case anyone ever missed out
https://www.fimfiction.net/story/62074/Friendship-is-Optimal
#29: Present An Open-Source Python Library For Monte Carlo Techniques
Great!
#40: Build A Better Social Network
Yay. Good stuff
#56: Aella Wants To Start A Dating Site Like Old OKCupid
Only one comment: in my opinion the okcupid blog archive from 2009-2012 was pretty good.
#6 looks like visions of chaos (https://softology.com.au/voc.htm) but more specific and more focused on real-world simulations instead of fractals and cellular automata