"More than 1 million children are behind on their reading due to the Pandemic. 11 year olds have the reading age of 7" - the pandemic's been on for two years now.
If they're 11 now, they were 9 at the beginning of the pandemic. If they can only read at the level of age 7, then either they went back really badly, or they were stuck at a lower reading age level to begin with.
I'm always chary of these kinds of "such an age can only do that an age work" because how do you know? If kids are reading for pleasure, it's generally because they're good readers. If kids are only reading because school makes them read, and out of school they don't read textbooks and they don't read anything else, then their natural reading level is going to be bad anyway whether or not they're in school, it's a pandemic, whatever.
"They are all classic books that have stood the test of time and apart from cracking good stories they feature archetypes who share their innate human knowledge and thus empower the reader."
Anything claimed to "empower" whomever should be burned to ashes. You know what worked when I was working in a secondary school and we were lucky enough, being in a designated area of educational disadvantage, to get a pilot project for a trained school librarian working in the area of encouraging reading? Trashy books, that the kids (mainly boys) would read. Not the kind of classics that will stand the test of time, but what kids considered 'cracking good stories'. Coax 'em in with things they like, such as magazines about motor bikes, then give them access to 'unimproving' literature and break the connection in their minds between 'reading = boring because I have to read textbooks for school'.
If your programme gets kids to read for pleasure, it will succeed. If it's about "getting them to catch up to the level to pass tests", then no matter what reward structure you set up, it won't have a lasting effect because once they've gone out the other end, no longer get rewards, and use the reading improvement only for sitting tests, they won't read outside of school.
How many people did we have on here/on SSC boasting they never read fiction or never read for pleasure, if they did read it was non-fiction because they wanted to learn some facts? That's not nurturing reading as a means of empowerment or enrichment.
This past year I realized I hardly read anything at all. As a kid I used to read non-stop, I struggled to find a constant supply of books of any kind to read - math, science, scifi, fantasy. Sadly I didn't stop buying books, I just stopped reading them. But it was the stacks of unread books that motivated me to change.
>>Trashy books, that the kids (mainly boys) would read. Not the kind of
>>classics that will stand the test of time, but what kids considered
>>'cracking good stories'. Coax 'em in with things they like, such as
>>magazines about motor bikes, then give them access to 'unimproving'
>>literature and break the connection in their minds between 'reading =
>>boring because I have to read textbooks for school'.
That's what worked for me - Dresden Files of all things. I'm not back to my previous reading levels (and currently giving the evil eye to a stack of unread programming books), but now I remember why reading is so fun.
RE: #83. That's not a grant, that's an investment. I assume if you've been working with attorneys and you're asking for just 100k the returns must be miserable, but it sounds like something potentially very lucrative. Who here does litigation financing?
Our project sits in an uncomfortable position of being both socially positive and a potentially good investment. As we learn more about the medical industry and acquire data, we'll be sharing data and information with journalists who do similar work. In the long-run, we think we'll be able to tackle some of the problems with health prices and cost-disease.
Not sure why you think the returns are miserable because we're asking for 100k. Can you explain? We're actually looking for $84k in order to acquire CMS data.
I assumed the returns were bad since 1) qui tam suits can get 20 plus percent of the settlement/verdict, 2) you state there's billions of dollars in fraud, and 3) the litigation finance industry is enormous. I don't really understand why people wouldn't be throwing money at you unless qui tam litigation is so expensive as to kill the returns.
We are, actually and we've made some progress. It's a weird investment vehicle for a few reasons- It takes 2-7 years for cases to settle in most cases and cases remain under seal for over a year in most cases. So far, we've been mining our network for investment but it doesn't fit into most boxes. It's not really a non-profit. It's not a startup because it has no liquidity event.
The returns look very good. Another company in this space recently obtained support from the DoJ which means an 80% chance of a settlement which should end up being 8-9 figures. As the relator, the company would take about 10% of that after litigation costs.
Would appreciate any advice on what kind of investor would be interested in this kind of work.
I would suggest having the law firm you are working with reach out to LexShares or one of its many competitors. They are going to want to see you have a viable claim. In this context, you'll likely have to show them examples of the evidence of fraud you've been able to sniff out from the data you already have and convince them that the evidence is strong enough to lead somewhere.
If you don't have any way to show a viable claim, and this is just something you personally think has a chance, I doubt traditional litigation financing would be interested in something so speculative. But if you shop around a little, talk to a litigation finance broker, you may find someone.
There's a whole section on entheogens over at Erowid. It's been years since I've been into them, but at that time ibogaine was reputed to be the most effective cure for opiate addiction around, and given the mortality of each its likely that would be more of a focus than alcohol
This was a fucking lie created by scam artists who ran ibogaine treatment centers that charged rich yuppies way too much to do nothing for their kids.
It fit the pattern that has been seen several times now, for a psychedelic that supposedly instantly cures you of opioid addiction, then when any research at all is done it turns out this just isn't true
Not necessarily, but in many cases yes. If you think a rejection would be amusing, it's at least 50-50 I think it's fake. If you think a rejection would be boring, I probably skipped it for that reason. A third possibility to consider is that I might have declined to reject a pitch because I actually liked it -- no comment on whether that actually happened.
Hanania's pitch isn't fake in the sense "this isn't Richard Hanania". However, it is something of a fake pitch in that the reasons given only loosely correlate with the reason he wants money.
The reason he wants money is "I'm semi-public figure Richard Hanania and I have a Substack". I'm familiar with him, and the "I will not mute him when I see a Matt Yglesias type re-tweet him" level is all the support I am interested in offering.
Also, I don't see why I would pay him to hire think-tank people when I could just start my own think tank.
>They have an online textbook. Unfortunately, it is unreadable. Anything that is about "quantum" is likely fake, and if they explain "spaced repetition" before they get to the uncertainty principle it is surely crap.
[Michael Nielsen](https://scholar.google.com/citations?user=ZP0eZ94AAAAJ&hl=en) is a great researcher and a co-author of one of the best Quantum Information texts. Dismissing something because of its association with the word "quantum" is uh... unwarranted.
Yeah I read it as a hasty dismissal, though kind of par for that list.
Thinking of the core idea, SRS definitely improves retention, and upgrading end of chapter review questions to incorporate SRS seems like an easy patch. The basics sound good.
The opportunity cost to think about here though is: just pay people to work through the VSI series from Oxford, or the leading 101 textbooks across topics, or the leading online courses, and have them make very high quality Anki decks to accompany each one and distribute those for free (much as the anki medschool community has basically done for those courses). Or hire devs to improve Anki's usability.
It is easy to get enamored with SRS. I love it as a tool. But have you actually taught bored kids with it? There are other pressing challenges in education aside from "meatware is bad at recall," and maybe that one is not actually the problem we think it is.
Could you elaborate on your reply to 104? To me it looks like having such a software package (or a platform) available open source would be a good thing. I am less optimistic than the applicant about the resources required to actually create something like this, but still it seems like it could benefit many people.
There's both a "software library" pitch and a "consumer product" pitch mixed together.
For a pure software library, the "do the work first and then ask for grant money" approach is the only way. If you can actually write such a library, you're probably either a student or a professional computer programmer -- and getting money won't actually help you write the library.
For a consumer product, this absolutely must be a business for liability issues. And unless I know your home address I cannot help you form a business.
Also as a consumer product, it has the problem that it doesn't do anything interesting. It might be able to say "you have indigestion the day after you eat 5x a normal amount of black pepper", which I can figure out on my own. But if it says "you are depressed two days after you bake a casserole" ... well, the software v1 is too stupid to realize both events are correlated with it being miserable weather outside. And if it says "you tend to get good news in the mail two days after you lose three chess matches in a row", it is lying with statistics.
There is no detectable signal at least 90% of the time, and if there isn't someone very visible to sue, it's immoral to encourage you to even try to do this for people you don't know.
#89 WikiCiv.org - A reader emailed noting that I didn't make a specific ask. I would mostly like editors to contribute content, the site is cheap enough to run that I don't especially need funds. Also putting me in touch with anyone who might like to collaborate would be very helpful too. Thanks!
How do you plan to make sure it is accurate? Wikipedia works well enough, so clearly that isn't impossible--but are there particular ideas from other wikis you are consciously taking on board, or are there unique challenges to a wiki on this topic?
I would like to take on as many of the lessons as possible from other wiki's: neutral point of view, references for claims etc. The only area I'm trying to innovate in is the scope. To ensure accuracy, any claim that can't be straightforwardly cited will need replicable demonstrations. If on the "steel" article I want to claim you can decarbonize pig iron with FeO, there needs to be a credible independent demonstration of this.
Have you looked at worldbuilding.stackexchange.com ? There are quite some people with the required expertise, some may be willing to help. (If you want to advertise, don't do it on the main site, but on chat or meta. If some of your articles are directly relevant for some question, you can refer to it, of course)
>#85: Study The Neuroscience Of How The Self Matures
>"In earlier stages the focus is on exterior stimuli, while in later stages the exploration is of one’s interior experience, and how the exterior world is interpreted through our interior experience."
What is "the focus" and how is this falsifiable?
>"Qualities such as compassion, dis-identification from the concept of self, and an understanding of the constructed nature of experience, become stronger and more nuanced at each stage."
Where are these "Qualities" and how can I measure them? How can I measure "an understanding of the constructed nature of experience?"
>"Our team, led by renowned Harvard neuroscientist Sara Lazar, is seeking funding to conduct a series of experiments to characterize the maturity process in scientific language and situate it within the fields of psychological and cognitive neuroscience. Charitable gifts of $50,000 to $500,000 will allow us to conduct essential preliminary studies to establish proof of concept and enable us to seek federal funding. Please contact slazar@mgh.harvard.edu with questions."
The American chestnut tree was once the dominant tree in much of the Eastern US. In 1904, a blight from Asian chestnuts arrived in New York. It obliterated the American chestnut, killing 3-4 billion trees. Today, less than 100 are left in their native range. Chestnuts are significantly more nutritious and delicious to people and animals than acorns and (unlike acorns) are produced in large quantities every year. Their nuts were an important source of food for people, livestock, and wildlife, especially in rural Appalachia, and their lumber was world renown. Now, they're just a line in a Christmas song. Several groups are trying to restore the American chestnut tree. I am not associated with any of them.* The American Chestnut Cooperators Foundation (https://www.accf-online.org/) breeds the few surviving blight resistant trees. The American Chestnut Foundation (https://acf.org/) backcrosses American chestnuts with Chinese chestnuts. SUNY Environmental Science and Forestry (https://www.esf.edu/chestnut/) has made a resistant tree using genetic engineering and is trying to get legal permission to release it into the wild from the USDA-APHIS, EPA, and FDA.
*Which is why I didn't submit this as a grant proposal. I did do a service project with ACF years ago, and met two of the surviving trees.
My name is Dan Sparkman and I want your help to plant trees for carbon capture. There are plenty of people planting trees and they are mostly worthy of support. However most reforestation projects are eventully going to be cut down. I aim to plant forest gardens which should stand for hundreds of years. "
Why? In hundreds of years, technology improvements will mean emissions will be vastly lower and/or carbon capture will be vastly more effective.
Furthermore, there's nothing unique about this idea. Other people with more knowledge and resources are already working towards reforestation efforts so I cannot possibly imagine it could be cost effective to donate to this cause instead of many others.
To your first point as much as I like technology even after 500 years of scientific advancement trees are going to be 370 million years ahead. Any technological solution is going to be energy hungry, energy in the current economy is carbon. That's a 2 steps forward 1 step back problem. Trees are also massively reproducible. They look after themselves and all of their externalities are positive.
The more I look into the issue the more I think the roll of plowing in carbon emissions is under appreciated. Plowed fields are dead, the decay, they oxidize They emit carbon. By planting a food crop as trees you avoid plowing. Sequestering more carbon in the soil than in the wood.
To your second point. Most reforestation projects are done by pulp and paper companies. They can plant more trees than me cheeper but they will just be cut down, releasing all that carbon and tearing up the soil when the machines move it. They also plant mono crop forests, they have all the biodiversity of a corn field.
Other carbon capture forests are straight up scams. Which for all you know mine is as well.
I'm going to plant trees with or with out your help. I see a place to apply some leverage. The climate is changing the boreal forest is leaving. Nut trees will be more viable every year (where I am). There are plenty of abandoned orchards and fields. There is land to plant trees on, I can get the trees and with a bit of help I can put them in the ground. If I can prove a viable crop which doesn't involve plowing, other people will steal my method and do it fast better and more efficient than me.
With a bit of help I can plan 1 to 2 hundred acres. That is enough to get attention if it works out.
1) Sequestering carbon by trees is quite limited. It exists but is happening on scale smaller than expected by most people
2) Tree planting is a good idea in general, I wish you a good luck!
3) "Any technological solution is going to be energy hungry, energy in the current economy is carbon" +1
4) "trees are going to be 370 million years ahead" - that is silly, in many cases dedicated engineering surpassed any biological solutions within years.
If I were a prospective investor, Dan, I would be interested in some details:
How much carbon will the nut trees sequester? As the trees get older, do they continue to absorb CO2 at the same rate as smaller trees? Are there other uses of the land that might be better CO2 sinks?
How much maintenance is required? Will competing species need to be suppressed? Will the trees need annual pruning or any other maintenance to remain productive? fertilizer? Irrigation?
Of course, food production must be considered as well. You might want to examine different tree species and show which provide the best combination of food production and CO2 absorption.
It seems to me that this project is close to your heart, and I think that's a very good thing. If you could do the math to back up your ideas, it would make a big difference. I'll bet with a few days, maybe a few weeks, of research you could show the science is sound.
To #91 re: Aquaporins, I have to assume you are already familiar, but if you're not there is a company doing Aquaporin based filtration and desalination.
Wouldn't a more likely MOA for weight loss while wearing weighted belts be the increased caloric expenditure required for every step? I wear 30+lbs of gear all days at work. Just climbing a few flights of stairs is a workout, and I'm famished by the end of the day. Without counting calories as well as tracking weights, this project seems of a limited scope.
On Jason Fung's youtube channel he just had a 9m video "Does exercise make us hungry". He references 2 meta analysis and another study (titles on the screen) at 3m, 4m, and 6m in. Summary: exercise immediately kills your appetite, but overall you just eat more to make up for it.
I would seem like the increase caloric expenditure would be most of the reason for weight loss, but in "Body Weight Homeostat that Regulates Fat Mass Independently of Leptin in Rats and Mice" (2018) Jansson et al. show that mice depleted of osteocytes don't get the weight loss effects seen in normal mice.
The weight loss is from appetite suppression dependent on your bones! (at least in mice)
And to your point "without counting calories as well as tracking weights, this project seems of a limited scope", the project is less about the biology behind the weight loss and more about compliance and effectiveness of different weighted clothing. Figuring out more about the biology is important, but turning it into a successful weight loss technique is more than just the biology.
"#68: Educational Software To Hit Developmental Windows In Babies
One way to counteract the growing burden of knowledge and increase innovation is to teach people more in less time. Early childhood education could be useful to this end. ... I plan to create a simple computer interface with accompanying educational software specifically targeted at babies. The interface will be simple enough that a baby can intuitively use it, and the software will make it easy to develop learning modules for babies."
This proposal makes me grind my teeth. Realistically, when you're dealing with a three month old baby, your childcare is most concerned with feeding them, changing their nappy, giving them a warm, safe place for naps, and then handing them back to the parents at the end of the day.
People *are* already working on "explicitly teaching skills and knowledge", I'll leave some links to Irish standards below. I'd be more sympathetic if this was aimed at helping at-risk children to catch up to and achieve age-appropriate developmental milestones, not "treat babies like rats on a treadmill". You want to teach babies perfect pitch, woo-hoo. Meanwhile, the kind of babies most at risk are not going to be very much helped with "the burden of knowledge" by computer software they can use to learn music, and the babies whose parents will load up on this are the kids who are already in the rat-race to pile up perfect grades, the right mixture of extra-curriculars and hobbies, and craft the best application to the good university for the good degree for the good job. Now they can start as soon as they're out of the cradle, instead of waiting until they're two years old. How nice. Because the world will stop turning if we don't churn out 15 year olds ready to start their adult working lives at that age.
Is perfect pitch even teachable, or is it a genetic (or epigenetic) trait that must be inherited ? Can babies even use computer UIs ? Forget all the moral/ethical/social implications -- this proposal sounds like an outright scam to me. In which case, I guess you could say we should encourage it, because separating rich people from their money would redistribute the wealth, or something...
Yes, perfect pitch is teachable if children are young enough. There may be genetics involved determining to some extent how easily a child picks up on it or what probability they may learn it without being explicitly taught.
Also, maybe my use of the word baby is confusing. I think it's clear that it'd be really hard and probably not useful to make this for a newborn. But babies are able to use simple apps on phones at a pretty young age and could probably use something more obvious (the board with large buttons I plan to make) at a younger age and with greater proficiency. Of course, in order to determine the details of what works and what doesn't, I'll be iterating on different designs.
Ok, assuming you and @Seta Sojiro are correct, and perfect pitch is teachable, then a). How do we know this ? Are there studies where babies who were taught developed perfect pitch at a higher rate than the control group, and b). Assuming (a) is true, why not use whatever method they used (I'm guessing, human-guided instruction), instead of this computerized UI ?
There aren't really great controlled studies, but there's good evidence. One of the best papers I've read on it shows convincingly that older students enrolled in a perfect pitch program don't learn it while the younger ones do. Maybe there's some crazy selection bias going on, but it'd have to be crazy.
Human-guided instruction is expensive, and the point of this project is to make something that can teach a lot of different things to a baby with little expense and little need for parent/teacher oversight. A lot of parents sadly don't have the time or means otherwise.
I understand the motivation, but still, how is this proposal substantially better from a myriad others that promise something to the extent of, "play Mozart to your baby 24/7 to make it into a super-genius uber-baby" ? Ok, so obviously I'm exaggerating for effect, but still, AFAIK the track record for these autonomous baby-teaching methods is quite poor. Also, how old is the original study ? I ask because, if the perfect-pitch training method actually works, and the study is sufficiently old, then (capitalistically speaking) most of the rich people today should have perfect pitch -- but AFAIK they do not.
It's not randomized but that's really not necessary when we're talking about effect sizes this large. Perfect pitch is incredibly rare (1 in 10,000), and we're talking about a greater than 90% success rate even when some of the kids enrolled were probably a bit too old.
You wouldn't demand a randomized trial proving that hiring a Spanish tutor can teach a child Spanish. You just observe that people do this all the time and it works. That's how perfect pitch programs are in Japan.
Interesting -- so, as I said above, is there a reason why rich people in the US don't all have perfect pitch ? Come to think of it, do all rich people in Japan have perfect pitch ? I mean, if I could give my baby superpowers, I'd consider it money well spent, so maybe they do and I just don't know ?
Perfect pitch isn't a superpower. For a lot of people it's little more than a party trick. I think I should make it clear that my project isn't about perfect pitch and teaching babies perfect pitch is just for the proof of concept.
Because few people know it’s an option, it’s time consuming, fewer people care, and it’s not a superpower. Perfect pitch is perhaps a misnomer. I have a friend who has it and she is a terrible singer and beginner level pianist. It’s just the ability to hear sounds as discrete categories like color, rather than as varying heights like greyscale. Seeing in color doesn’t make you da Vinci and perfect pitch doesn’t make you Mozart.
Speaking three languages is more of a superpower, everyone knows it’s possible to teach a child (children all around the world grow up speaking 3+ languages), and yet the majority of kids of rich Americans only speak English.
I'm pretty confident perfect pitch is teachable, but only to toddlers. The Eguchi method claims a near 100% success rate as long as it is used between the ages of 2 and 4 (the critical period closes after age 4). Thousands of people in Japan have used the method and developed perfect pitch.
I think this is a good idea, but my very small two cents is that perfect pitch might not be the best thing to teach. From my experience being/hanging around musicians:
1. For a child who will go on to become a musician, a strong sense of relative pitch will serve them better than perfect pitch, and having perfect pitch can be a crutch that weakens their sense of relative pitch. In addition, having perfect pitch comes with a "curse" where, as you age, your sense of perfect pitch may drift, resulting in a mismatch between the pitch you perceive and the pitch you're actually hearing. From what I've heard, this sucks, a lot.
2. If you're not going to become a musician, perfect pitch won't be much more than a party trick anyway
Is it just me, or do most of these sound like outright scams -- or, to be more charitable, well-intentioned proposals that are obviously doomed to failure ? Is there any evidence that babies can use computer interfaces, or that *this time* the video courses can finally replace in-person classroom instruction (unlike all those other ones that crashed and burned), or that spirits do exist after all, etc. ? I know there's lots of evidence against all of these propositions (and more), but maybe I'm missing the one slam-dunk argument for them ? If so, why wasn't it in the proposal ?
I'm very surprised Scott didn't fund #122: In Vitro Gametogenesis. I'm guessing the problem was the quality of the applicants themselves, rather than the quality of the idea.
The actual problem was a combination of "they needed too much money" and "I was sure some investors would fund it so I threw it at them instead, and then they didn't fund it". I actually think it's very high-quality.
Thanks for the support and the opportunity. To be honest, it is very speculative at this point. Would be considered pre-seed for most investors. That of course hasn’t stopped them from throwing money into the alternative technology I mentioned. I actually think our time horizons for getting it to work in humans are quite similar. And then there is the issue of human embryo research which some people find controversial. Anyway, I thought there might be someone in the ACX community with the means are interest in helping move the science forward.
Hi, I'm a scientist and I'm just... not sure how that's going to work?
Consider these sentences that I just made up but that could be taken from the scientific literature:
"We induced tumors in the Drosophila midgut by overexpressing the yki transcription factor in intestinal stem cells using Gal4/UAS."
"We observed a 5-fold increase in GFP fluorescence intensity after 24-h incubation of the reporter cells with interferon-gamma."
"TGF-beta signaling triggers SMAD phosphorylation followed by changes in gene expression."
How do you make these sentences any easier to understand for the lay reader? I promise, I'm not deliberately trying to make these sentences difficult. Just, they're dealing with complex concepts, and pretty much every scientific article is composed of sentences like these. Sure, the Introduction section is supposed to ease you into the article, but the author assumes that you already are at least somewhat conversant with the field. If you're reading a paper on Drosophila, they're not going to explain to you what Gal4/UAS is.
I really hope I'm not coming across as super arrogant and stuck-up here. If there is a simple way to make scientific papers more accessible without sacrificing scientific rigor and information density, I would really like to know!
Yeah, that is basically the problem with all science reporting. "We reduced the prevalence of type-ABC tumors in Drosophila by overexpressing transcription factor XYZ" becomes "Scientists cure cancer !". I don't think anything can be done about this. You either understand biology/computer science/etc., or you don't, and there's no amount of NLP datamining that can do the work for you. As the famous quote goes, "there's no royal road to geometry".
As a PhD student, I've become very aware that there are many opportunities to improve your writing. I feel that every year, the papers I've written the year before are filled with embarrassing writing mistakes I don't make anymore. What was most useful to me is discussing a specific paper with a scientific writing expert, who works for the university in this capacity. While she couldn't claim to understand my paper, she did find many ways to improve the writing, usually after a discussion of the form "I don't know what these words mean, but judging from the sentence structure... , so perhaps do..." "These thing basically mean this and that" "alright, so what if you write this instead..." And this was a paper filled with mathematics. While she of course didn't look at the proofs, it is still interesting that general scientific writing expertise is still very relevant in a field with rather idiosyncratic writing.
I recently moved to computer science from biology, and in fields that are built more on theory (like CS) and less on mountains of previous empirical work (like biology), I think there would be more potential for this kind of work making papers easier to read.
> How do you make these sentences any easier to understand for the lay reader?
Not the original proposer, but I've thought about something like this for a while, though for software documentation. The idea would be to build a "definition checker". Every word you use has to be defined. The easiest way to do this would be software, something like wikipedia or Gwern's blog on steroids. As an example: You see "drosophila", don't know what it is, press shift and hover it and see a short definition of drosophila. Drosophila itself is defined (thus the Gwern blog reference https://www.gwern.net/index, you can seee when you hover a link it opens something, and the hover thing still works inside, so you can open things very deep).
There are two parts to this:
- the easy part, the software. Find a way to extract the "roots" of words and build dependency graphs (DAG ideally). This is basically how we manage dependencies in software, so we can use that. Someone already built something that looks like this here https://github.com/mjambon/vanity. To apply that to all things written, we can combine that to spellchecker tech. Fortunately someone reimplemented the most popular spellchecker in Python while explaining how it works, so this part wouldn't be too hard too https://zverok.github.io/spellchecker.html
- And now there's the hard part: how do you choose what you start with? What are the axioms? My initial and naive idea is using basic english and leveraging basic wikipedia, but of course it won't be that easy.
The fun part is that we could work with #89, as it's basically for me the same idea: how to bootstrap human knowledge?
If I'm understanding you correctly, it sounds like this idea would already be super helpful without making it fully automatic. There could be a manually curated standard ontology of biological acronyms and experiments somewhere, and authors could manually link to it from the right words in their papers (with some easy spell checking help, probably)
Like other commenters, I also think there are generally many potential avenues of improvement for papers.
I enjoy many posts on ACX that summarize (review) research or books. Maybe this project could be something like that but if you kept reading, you'd actually get all the info from the paper.
If you ever had to review papers for a conference, you know that this is not about making things easier for a layman to understand.
A badly written paper is an obstacle course. You might see missing steps in reasoning, flaws in logic, graphs that are supposed to show something but show something else entirely, sentences that take 10 lines, narrative that does not start at the beginning and/or randomly jumps around, confusing notations, errors in arithmetics... I could continue.
Luckily, a reviewer has the option of saying that a paper is too terrible to try to salvage. Sometimes, however, the reviewers are not paying attention. And sometimes a paper gets publicized without any review.
Making a badly written paper readable is not an easy task, and it's one that cannot be automated. You have to go back and forth with the authors, trying to figure out exactly what they meant. If the authors are not available, and you have to guess, your rewrite might get the meaning wrong.
Getting authors to follow certain guidelines would be nice, but it's hard for me to be optimistic. Strunk and White is 105 pages, and most people who write scientific papers haven't even gone through that.
I wish the authors of #96 the best of luck, but it's very hard for me to believe that they can make a difference.
If you believe that #82 is in any way for real, then I hereby propose #82.5: give me 3/4 of the money given to #82, and I will develop a sure-fire solution that will stop it in its tracks before it can eat the Earth. 100% guarantee, or your money back.
(I'm talking about AGI here; obviously boring old self-driving cars already exist)
#79: A) Get people to state their intentions explicitly and honestly. B) Put everything in place that’s necessary to ensure that people actually do A, instead of just pretending to do it. The core insight is A, but it’s irrelevant without B’s infrastructure.
...if you have solved that one, I believe you have a decent shot at the Nobel memorial price in Economics in Stockholm. Since this is THE question to solve in the human sciences. You can write the story of human evolution as our attempt to be able to differentiate the genuines from the fakes.
Re: Darwin's theory (embryo version) in The Expression of the Emotions in Man and the Animals, where he suggests that involuntary blushing evolved as a way for humans to signal to others that they are self-bound to behaviour like honesty and chastity, thereby once-and-for-all providing signals impossible to fake, therefore gaining more trust from others, therefore being better able to spread their genes. A theory of why altruism toward non-kin exists builds on the same idea: Giving money to total strangers is a way to appear pseudo-self-bound to honesty in your everyday dealings, making more people willing to take the risk to deal with you, and gaining an evolutionary advantage that way. (Wrote a paper about that myself some decades ago.)
So again, if you have solved this one, humanity will thank you in the for-ever future. And you will probably influence the future path of human evolution. But when something is too grand to be true, there is usually some snag in the theory.
Best of luck though!
PS...and we have just got a demonstration of how difficult it is to do this differentiation, in Scott's blog post of the trials and tribulations he went through in order to decide who to fund and who not to fund among the hopeful.;-)
I’m a little bit mortified about that pitch. I noticed that my original whitepaper was doing a bad job of translating my idiosyncratic understanding into good, well-reasoned explanation. And I started wondering if the bitcoin addresses that Coinbase generates for deposit are persistently linked to my account :)
Will get my own wallet when I get around to it.
Of course, I still should have written the pitch by February 5, as I promised.
But in the meantime I was already looking more into memory palaces and long-term memory retention, areas that I previously neglected. And what I learnt from research made me want to reconsider how those aspect fit into the larger problem.
So, I asked Scott to rescind it, so that this didn’t have the “vague promises, secret knowledges and I can totally turn lead into gold, trust me bro”-vibe.
And I figured, that he might just sort me out anyway, because this must have appeared borderline scummy. Well, he said, “no takebacks” and stood by that. And apparently, I’m not that obviously untrustworthy. Fair enough.
I’m glad he posted it anyway, because I got five emails already expressing interest and/or telling me about similar projects and what appear to be similar approaches.
So this worked out great!
Although I’m quite excited to discuss that, this is a rare weekend within the next couple months, where I don’t have time to respond to questions or work on/think about this project.
But if you want to leave questions in the comments or by email, feel free to do so.
I will state for the record that this is one I did not review under the "it seems fake" clause.
However, I would entertain a revised request (probably on a classified thread) if there was a Substack blog on your efforts (with at least 3 real posts).
"I will state for the record that this is one I did not review under the "it seems fake" clause."
Totally fair.
"However, I would entertain a revised request (probably on a classified thread) if there was a Substack blog on your efforts (with at least 3 real posts)."
Not opposed to getting a public presence once I have the time for it.
If somebody wants to be paid for their writing these days, the best place to do so is Substack.
You can do it with a blog elsewhere plus a Patreon, but ... if you want to make it easy for people reading an ACX thread to pay you money, I don't know why you would.
Re: #96: Improve The Readability Of Scientific Writing
Look up Steven Pinker's "The Sense of Style" and his talks about it; he's a famous guy to get endorsement from if you can, once you've got some actual work published, and also the book has some good tips.
Don't try to "create a new journal to publish rewritten versions of existing papers"; that'll get you into weird copyright issues at worst, or just throw people off as weird and radical at best. A thing you could do instead is try to sell your services as an "editing plus" or even "academic ghost-writing" company, working closely with the paper authors / scientists to optimize their language for readibility. (I'd say "editing plus" so they can still call themselves authors while acknowledging your company, explicitly, so you can get advertised in the journals they publish in; maybe even get a contract with an existing journal to be their go-to place to send things to edit.)
Your e.g. website might be able to rewrite a *paragraph* of an existing paper and call it fair use—don't quote me on that, I'm not a lawyer—to illustrate the value and style, but yeah, not whole papers.
Is there someplace one can buy weighted belts to give it a try yourself (I know that it won't be useful for the study but it's an interesting thing to try...and I figure it makes every movement a bit more of an exertion).
Yes, you can buy a scuba belt and accompanying weights on Amazon, for instance. There are also a lot of different styles of weighted vests for sale that are commonly used to add resistance to exercise.
I think it's important to point out that it's theorized that it isn't so much the extra exertion that added weighted provides but actually *the loading of your bones leading to appetite suppression* which leads to weight loss.
What I do not get about this model is that normal, plain obesity will exert extra force on your bones as well, and it does not seem to lead to weight loss.
Regarding #77 (Computer programs which write themselves) how is this different than the Racket project or, more generally, the problem of creating new programming languages? I'm a big fan of Racket (a programming language optimized for writing programming languages) but there are reasons that most programs aren't written in Racket.
Namely, that programming languages are all about tradeoffs. What things are easy to do and easy to think about and what is less easy. We build a bunch of different languages because different domains are easier to program when different things are easier to write/think about.
The more a language/development environment focuses on being able to manipulate programs themselves the less optimized your language will be to deal with other kinds of problems. More generally, as you increase the number of layers involved in creating a program (e.g. you go through some kind of advanced program transformation) you often make it harder to reason/conceptualize what the program is doing. That's a cost.
From what I remember when I looked at Racket, it doesn't allow custom syntax or precise control over computations executed (only the output of that computation). The language layers of this project are meant to be much more freeform. Racket mostly lets you write Scheme like languages. Yes, you can hack it a bit with macros but I find that makes it less first class. Here, no layer is off limits for editing!
I don't understand why what you're describing isn't a cost for other (i.e., popular, existing) programming languages and environments. They have just as many layers, but instead you can't easily edit them. Since the layers are made smaller and since you wrote them yourself, I'd think it'd have a lower cost. Or you'd at least have the power to reduce that (by making less leaky abstractions, editing them out, etc).
The part of the language/environment meant to deal with other kinds of problems is something you'd yourself. The same way you can create a missing library in something more traditional.
The project of preserving dead peoples' legacies is really cool. I thought about this idea a long time ago, it's nice that someone is actually trying to do it.
I just started a two year unpaid sabbatical to pursue my personal interests and to complete my MBA. I am 36 years old. About five months ago I discovered Data Science. I've been teaching myself SQL and R for about two hours per day. I am seeking 149$ to purchase a subscription to Datacamp.com . This website is a better teaching platform compared to my University. Please DM @streamust on twitter if you can sponsor me. 📊
For Michael Arntzenius, I understand the motivation is to create tools to program in existing languages, that's directly applicable to work now.
Is there much interest out there in creating new languages from scratch with voice programming in mind? There are a lot of weird conventions in languages trying to make them compact that don't necessarily make sense outside of typing.
#88: as somebody with a physics education I always thought "emergent phenomena" meant stuff like temperature and ferromagnetism, that don't exist in an underlying physical theory but emerge from collective behaviour of many particles. Then when I read Less Wrong I always got a bit confused about how EY liked to rant about "emergent phenomena", which I thought were so uncontroversial. But it turns out it's a euphemism for woo? You learn something new every day.
"More than 1 million children are behind on their reading due to the Pandemic. 11 year olds have the reading age of 7" - the pandemic's been on for two years now.
I don't see a logical impossibility there; reading skills can get worse, not just freeze.
+1
If they're 11 now, they were 9 at the beginning of the pandemic. If they can only read at the level of age 7, then either they went back really badly, or they were stuck at a lower reading age level to begin with.
I'm always chary of these kinds of "such an age can only do that an age work" because how do you know? If kids are reading for pleasure, it's generally because they're good readers. If kids are only reading because school makes them read, and out of school they don't read textbooks and they don't read anything else, then their natural reading level is going to be bad anyway whether or not they're in school, it's a pandemic, whatever.
"They are all classic books that have stood the test of time and apart from cracking good stories they feature archetypes who share their innate human knowledge and thus empower the reader."
Anything claimed to "empower" whomever should be burned to ashes. You know what worked when I was working in a secondary school and we were lucky enough, being in a designated area of educational disadvantage, to get a pilot project for a trained school librarian working in the area of encouraging reading? Trashy books, that the kids (mainly boys) would read. Not the kind of classics that will stand the test of time, but what kids considered 'cracking good stories'. Coax 'em in with things they like, such as magazines about motor bikes, then give them access to 'unimproving' literature and break the connection in their minds between 'reading = boring because I have to read textbooks for school'.
If your programme gets kids to read for pleasure, it will succeed. If it's about "getting them to catch up to the level to pass tests", then no matter what reward structure you set up, it won't have a lasting effect because once they've gone out the other end, no longer get rewards, and use the reading improvement only for sitting tests, they won't read outside of school.
How many people did we have on here/on SSC boasting they never read fiction or never read for pleasure, if they did read it was non-fiction because they wanted to learn some facts? That's not nurturing reading as a means of empowerment or enrichment.
This past year I realized I hardly read anything at all. As a kid I used to read non-stop, I struggled to find a constant supply of books of any kind to read - math, science, scifi, fantasy. Sadly I didn't stop buying books, I just stopped reading them. But it was the stacks of unread books that motivated me to change.
>>Trashy books, that the kids (mainly boys) would read. Not the kind of
>>classics that will stand the test of time, but what kids considered
>>'cracking good stories'. Coax 'em in with things they like, such as
>>magazines about motor bikes, then give them access to 'unimproving'
>>literature and break the connection in their minds between 'reading =
>>boring because I have to read textbooks for school'.
That's what worked for me - Dresden Files of all things. I'm not back to my previous reading levels (and currently giving the evil eye to a stack of unread programming books), but now I remember why reading is so fun.
for #102--how can waterfighting be made competitive? How do you win?
I can think of one way. Secure sheets of paper to your back, front and both sides, and if a paper gets wet you're dead.
RE: #83. That's not a grant, that's an investment. I assume if you've been working with attorneys and you're asking for just 100k the returns must be miserable, but it sounds like something potentially very lucrative. Who here does litigation financing?
Our project sits in an uncomfortable position of being both socially positive and a potentially good investment. As we learn more about the medical industry and acquire data, we'll be sharing data and information with journalists who do similar work. In the long-run, we think we'll be able to tackle some of the problems with health prices and cost-disease.
Not sure why you think the returns are miserable because we're asking for 100k. Can you explain? We're actually looking for $84k in order to acquire CMS data.
Hey, nothing wrong with being both those things.
I assumed the returns were bad since 1) qui tam suits can get 20 plus percent of the settlement/verdict, 2) you state there's billions of dollars in fraud, and 3) the litigation finance industry is enormous. I don't really understand why people wouldn't be throwing money at you unless qui tam litigation is so expensive as to kill the returns.
Are you specifically not looking for investors?
We are, actually and we've made some progress. It's a weird investment vehicle for a few reasons- It takes 2-7 years for cases to settle in most cases and cases remain under seal for over a year in most cases. So far, we've been mining our network for investment but it doesn't fit into most boxes. It's not really a non-profit. It's not a startup because it has no liquidity event.
The returns look very good. Another company in this space recently obtained support from the DoJ which means an 80% chance of a settlement which should end up being 8-9 figures. As the relator, the company would take about 10% of that after litigation costs.
Would appreciate any advice on what kind of investor would be interested in this kind of work.
I would suggest having the law firm you are working with reach out to LexShares or one of its many competitors. They are going to want to see you have a viable claim. In this context, you'll likely have to show them examples of the evidence of fraud you've been able to sniff out from the data you already have and convince them that the evidence is strong enough to lead somewhere.
If you don't have any way to show a viable claim, and this is just something you personally think has a chance, I doubt traditional litigation financing would be interested in something so speculative. But if you shop around a little, talk to a litigation finance broker, you may find someone.
So last time I tried to be as nice as possible to every pitch. I think this time, to be fair, I will do the opposite.
https://newslettr.yevaud.com/p/contra-acx-grants-part-ii/comments
There's a whole section on entheogens over at Erowid. It's been years since I've been into them, but at that time ibogaine was reputed to be the most effective cure for opiate addiction around, and given the mortality of each its likely that would be more of a focus than alcohol
This was a fucking lie created by scam artists who ran ibogaine treatment centers that charged rich yuppies way too much to do nothing for their kids.
It fit the pattern that has been seen several times now, for a psychedelic that supposedly instantly cures you of opioid addiction, then when any research at all is done it turns out this just isn't true
Done. 23 rejection statements. I've skipped over quite a few I believe are fake proposals.
The one complaint I could make about your analysis is that some of the rejections are, arguably, not harsh enough.
His take on #109 was hardly even negative. It was just a wish list.
Should that be read as "if it's not there, Alex thinks it's fake"? I'm interested particularly in why you would think Hanania's to be fake.
Not necessarily, but in many cases yes. If you think a rejection would be amusing, it's at least 50-50 I think it's fake. If you think a rejection would be boring, I probably skipped it for that reason. A third possibility to consider is that I might have declined to reject a pitch because I actually liked it -- no comment on whether that actually happened.
Hanania's pitch isn't fake in the sense "this isn't Richard Hanania". However, it is something of a fake pitch in that the reasons given only loosely correlate with the reason he wants money.
The reason he wants money is "I'm semi-public figure Richard Hanania and I have a Substack". I'm familiar with him, and the "I will not mute him when I see a Matt Yglesias type re-tweet him" level is all the support I am interested in offering.
Also, I don't see why I would pay him to hire think-tank people when I could just start my own think tank.
>#80: A “Mnemonic Medium” To Replace Textbooks
>They have an online textbook. Unfortunately, it is unreadable. Anything that is about "quantum" is likely fake, and if they explain "spaced repetition" before they get to the uncertainty principle it is surely crap.
[Michael Nielsen](https://scholar.google.com/citations?user=ZP0eZ94AAAAJ&hl=en) is a great researcher and a co-author of one of the best Quantum Information texts. Dismissing something because of its association with the word "quantum" is uh... unwarranted.
Yeah I read it as a hasty dismissal, though kind of par for that list.
Thinking of the core idea, SRS definitely improves retention, and upgrading end of chapter review questions to incorporate SRS seems like an easy patch. The basics sound good.
The opportunity cost to think about here though is: just pay people to work through the VSI series from Oxford, or the leading 101 textbooks across topics, or the leading online courses, and have them make very high quality Anki decks to accompany each one and distribute those for free (much as the anki medschool community has basically done for those courses). Or hire devs to improve Anki's usability.
More though, read about tanagrabeast's journey:
https://www.lesswrong.com/posts/F6ZTtBXn2cFLmWPdM/seven-years-of-spaced-repetition-software-in-the-classroom-1
It is easy to get enamored with SRS. I love it as a tool. But have you actually taught bored kids with it? There are other pressing challenges in education aside from "meatware is bad at recall," and maybe that one is not actually the problem we think it is.
Could you elaborate on your reply to 104? To me it looks like having such a software package (or a platform) available open source would be a good thing. I am less optimistic than the applicant about the resources required to actually create something like this, but still it seems like it could benefit many people.
There's both a "software library" pitch and a "consumer product" pitch mixed together.
For a pure software library, the "do the work first and then ask for grant money" approach is the only way. If you can actually write such a library, you're probably either a student or a professional computer programmer -- and getting money won't actually help you write the library.
For a consumer product, this absolutely must be a business for liability issues. And unless I know your home address I cannot help you form a business.
Also as a consumer product, it has the problem that it doesn't do anything interesting. It might be able to say "you have indigestion the day after you eat 5x a normal amount of black pepper", which I can figure out on my own. But if it says "you are depressed two days after you bake a casserole" ... well, the software v1 is too stupid to realize both events are correlated with it being miserable weather outside. And if it says "you tend to get good news in the mail two days after you lose three chess matches in a row", it is lying with statistics.
There is no detectable signal at least 90% of the time, and if there isn't someone very visible to sue, it's immoral to encourage you to even try to do this for people you don't know.
#67: Make the weighted vests in the form of mail hauberks. Encourage compliance by periodically attacking test subjects with a sword.
Little known fact: the beginnings of obesity epidemic line up with the time period that mail hauberks went out of fashion!
#89 WikiCiv.org - A reader emailed noting that I didn't make a specific ask. I would mostly like editors to contribute content, the site is cheap enough to run that I don't especially need funds. Also putting me in touch with anyone who might like to collaborate would be very helpful too. Thanks!
How do you plan to make sure it is accurate? Wikipedia works well enough, so clearly that isn't impossible--but are there particular ideas from other wikis you are consciously taking on board, or are there unique challenges to a wiki on this topic?
I would like to take on as many of the lessons as possible from other wiki's: neutral point of view, references for claims etc. The only area I'm trying to innovate in is the scope. To ensure accuracy, any claim that can't be straightforwardly cited will need replicable demonstrations. If on the "steel" article I want to claim you can decarbonize pig iron with FeO, there needs to be a credible independent demonstration of this.
"Please just edit Wikipedia instead." -- Jimbo Wales
"Wikipedia is an encyclopedia. Encyclopedias summarize knowledge, rather than try to contain all of it." - Wikipedia Scope page: https://en.wikipedia.org/wiki/Wikipedia:Scope
Here's an challenge: make https://en.wikipedia.org/wiki/Lithium_iron_phosphate_battery a "Good Article".
You mean be a good article on Wikiciv.org? Challenge accepted! (Though it may be a little while as we have to work our way up the "tech tree").
If you want a tech tree, don't write an encyclopedia, play Civilization 4.
This project is Civ inspired: What if Gandhi nukes you?
Have you looked at worldbuilding.stackexchange.com ? There are quite some people with the required expertise, some may be willing to help. (If you want to advertise, don't do it on the main site, but on chat or meta. If some of your articles are directly relevant for some question, you can refer to it, of course)
I had not seen this. This is very helpful. Thank you!
I just wanted to say I think it's a great project, and I hope you succeed!
I really appreciate you saying that. I'm going to give it my best shot!
>#85: Study The Neuroscience Of How The Self Matures
>"In earlier stages the focus is on exterior stimuli, while in later stages the exploration is of one’s interior experience, and how the exterior world is interpreted through our interior experience."
What is "the focus" and how is this falsifiable?
>"Qualities such as compassion, dis-identification from the concept of self, and an understanding of the constructed nature of experience, become stronger and more nuanced at each stage."
Where are these "Qualities" and how can I measure them? How can I measure "an understanding of the constructed nature of experience?"
>"Our team, led by renowned Harvard neuroscientist Sara Lazar, is seeking funding to conduct a series of experiments to characterize the maturity process in scientific language and situate it within the fields of psychological and cognitive neuroscience. Charitable gifts of $50,000 to $500,000 will allow us to conduct essential preliminary studies to establish proof of concept and enable us to seek federal funding. Please contact slazar@mgh.harvard.edu with questions."
Half a million? Really? I can't believe Harvard took her, good thing they're not all like that! https://www.epsilontheory.com/gell-mann-amnesia/
>and a few that violated my sensibilities for one or another reason
And yet somehow, a woke loony like Alice Evans didn't offend your sensibilities enough not to hand over 60 thousands dollars
# 126: Restore the American Chestnut Tree
The American chestnut tree was once the dominant tree in much of the Eastern US. In 1904, a blight from Asian chestnuts arrived in New York. It obliterated the American chestnut, killing 3-4 billion trees. Today, less than 100 are left in their native range. Chestnuts are significantly more nutritious and delicious to people and animals than acorns and (unlike acorns) are produced in large quantities every year. Their nuts were an important source of food for people, livestock, and wildlife, especially in rural Appalachia, and their lumber was world renown. Now, they're just a line in a Christmas song. Several groups are trying to restore the American chestnut tree. I am not associated with any of them.* The American Chestnut Cooperators Foundation (https://www.accf-online.org/) breeds the few surviving blight resistant trees. The American Chestnut Foundation (https://acf.org/) backcrosses American chestnuts with Chinese chestnuts. SUNY Environmental Science and Forestry (https://www.esf.edu/chestnut/) has made a resistant tree using genetic engineering and is trying to get legal permission to release it into the wild from the USDA-APHIS, EPA, and FDA.
*Which is why I didn't submit this as a grant proposal. I did do a service project with ACF years ago, and met two of the surviving trees.
"#125: Plant Trees For Carbon Capture
My name is Dan Sparkman and I want your help to plant trees for carbon capture. There are plenty of people planting trees and they are mostly worthy of support. However most reforestation projects are eventully going to be cut down. I aim to plant forest gardens which should stand for hundreds of years. "
Why? In hundreds of years, technology improvements will mean emissions will be vastly lower and/or carbon capture will be vastly more effective.
Furthermore, there's nothing unique about this idea. Other people with more knowledge and resources are already working towards reforestation efforts so I cannot possibly imagine it could be cost effective to donate to this cause instead of many others.
To your first point as much as I like technology even after 500 years of scientific advancement trees are going to be 370 million years ahead. Any technological solution is going to be energy hungry, energy in the current economy is carbon. That's a 2 steps forward 1 step back problem. Trees are also massively reproducible. They look after themselves and all of their externalities are positive.
The more I look into the issue the more I think the roll of plowing in carbon emissions is under appreciated. Plowed fields are dead, the decay, they oxidize They emit carbon. By planting a food crop as trees you avoid plowing. Sequestering more carbon in the soil than in the wood.
To your second point. Most reforestation projects are done by pulp and paper companies. They can plant more trees than me cheeper but they will just be cut down, releasing all that carbon and tearing up the soil when the machines move it. They also plant mono crop forests, they have all the biodiversity of a corn field.
Other carbon capture forests are straight up scams. Which for all you know mine is as well.
I'm going to plant trees with or with out your help. I see a place to apply some leverage. The climate is changing the boreal forest is leaving. Nut trees will be more viable every year (where I am). There are plenty of abandoned orchards and fields. There is land to plant trees on, I can get the trees and with a bit of help I can put them in the ground. If I can prove a viable crop which doesn't involve plowing, other people will steal my method and do it fast better and more efficient than me.
With a bit of help I can plan 1 to 2 hundred acres. That is enough to get attention if it works out.
1) Sequestering carbon by trees is quite limited. It exists but is happening on scale smaller than expected by most people
2) Tree planting is a good idea in general, I wish you a good luck!
3) "Any technological solution is going to be energy hungry, energy in the current economy is carbon" +1
4) "trees are going to be 370 million years ahead" - that is silly, in many cases dedicated engineering surpassed any biological solutions within years.
If I were a prospective investor, Dan, I would be interested in some details:
How much carbon will the nut trees sequester? As the trees get older, do they continue to absorb CO2 at the same rate as smaller trees? Are there other uses of the land that might be better CO2 sinks?
How much maintenance is required? Will competing species need to be suppressed? Will the trees need annual pruning or any other maintenance to remain productive? fertilizer? Irrigation?
Of course, food production must be considered as well. You might want to examine different tree species and show which provide the best combination of food production and CO2 absorption.
It seems to me that this project is close to your heart, and I think that's a very good thing. If you could do the math to back up your ideas, it would make a big difference. I'll bet with a few days, maybe a few weeks, of research you could show the science is sound.
#70 & #71 are the same application.
And, unless I missed it, there are no contact details for either of them.
support at(@) vrlanguages.co.uk
To #91 re: Aquaporins, I have to assume you are already familiar, but if you're not there is a company doing Aquaporin based filtration and desalination.
https://aquaporin.com/
I guess your angle is more getting them into vesicles and seeing if they will passively uptake water from a salt solution?
Something like that yes, except all in one step.
Wouldn't a more likely MOA for weight loss while wearing weighted belts be the increased caloric expenditure required for every step? I wear 30+lbs of gear all days at work. Just climbing a few flights of stairs is a workout, and I'm famished by the end of the day. Without counting calories as well as tracking weights, this project seems of a limited scope.
On Jason Fung's youtube channel he just had a 9m video "Does exercise make us hungry". He references 2 meta analysis and another study (titles on the screen) at 3m, 4m, and 6m in. Summary: exercise immediately kills your appetite, but overall you just eat more to make up for it.
I would seem like the increase caloric expenditure would be most of the reason for weight loss, but in "Body Weight Homeostat that Regulates Fat Mass Independently of Leptin in Rats and Mice" (2018) Jansson et al. show that mice depleted of osteocytes don't get the weight loss effects seen in normal mice.
The weight loss is from appetite suppression dependent on your bones! (at least in mice)
And to your point "without counting calories as well as tracking weights, this project seems of a limited scope", the project is less about the biology behind the weight loss and more about compliance and effectiveness of different weighted clothing. Figuring out more about the biology is important, but turning it into a successful weight loss technique is more than just the biology.
I used to train with a weighted vest and sometimes even wore it in a normal day. Its max weight was 30 kgs (65 lbs.)
My experience:
a) wearing a weighted vest will train your breathing muscles in a way that normal exercise won't,
b) all the important leg joints hate, hate, hate it,
c) my weight loss was minimal, especially when compared to intermittent fasting. But possibly my body composition changed (less fat, more muscle).
"#68: Educational Software To Hit Developmental Windows In Babies
One way to counteract the growing burden of knowledge and increase innovation is to teach people more in less time. Early childhood education could be useful to this end. ... I plan to create a simple computer interface with accompanying educational software specifically targeted at babies. The interface will be simple enough that a baby can intuitively use it, and the software will make it easy to develop learning modules for babies."
This proposal makes me grind my teeth. Realistically, when you're dealing with a three month old baby, your childcare is most concerned with feeding them, changing their nappy, giving them a warm, safe place for naps, and then handing them back to the parents at the end of the day.
People *are* already working on "explicitly teaching skills and knowledge", I'll leave some links to Irish standards below. I'd be more sympathetic if this was aimed at helping at-risk children to catch up to and achieve age-appropriate developmental milestones, not "treat babies like rats on a treadmill". You want to teach babies perfect pitch, woo-hoo. Meanwhile, the kind of babies most at risk are not going to be very much helped with "the burden of knowledge" by computer software they can use to learn music, and the babies whose parents will load up on this are the kids who are already in the rat-race to pile up perfect grades, the right mixture of extra-curriculars and hobbies, and craft the best application to the good university for the good degree for the good job. Now they can start as soon as they're out of the cradle, instead of waiting until they're two years old. How nice. Because the world will stop turning if we don't churn out 15 year olds ready to start their adult working lives at that age.
Early Childhood Curriculum Framework:
https://ncca.ie/media/4151/aistear_theearlychildhoodcurriculumframework.pdf
National Quality Framework for Early Childhood Education
https://siolta.ie/media/pdfs/siolta-manual-2017.pdf
Is perfect pitch even teachable, or is it a genetic (or epigenetic) trait that must be inherited ? Can babies even use computer UIs ? Forget all the moral/ethical/social implications -- this proposal sounds like an outright scam to me. In which case, I guess you could say we should encourage it, because separating rich people from their money would redistribute the wealth, or something...
Yes, perfect pitch is teachable if children are young enough. There may be genetics involved determining to some extent how easily a child picks up on it or what probability they may learn it without being explicitly taught.
Also, maybe my use of the word baby is confusing. I think it's clear that it'd be really hard and probably not useful to make this for a newborn. But babies are able to use simple apps on phones at a pretty young age and could probably use something more obvious (the board with large buttons I plan to make) at a younger age and with greater proficiency. Of course, in order to determine the details of what works and what doesn't, I'll be iterating on different designs.
Ok, assuming you and @Seta Sojiro are correct, and perfect pitch is teachable, then a). How do we know this ? Are there studies where babies who were taught developed perfect pitch at a higher rate than the control group, and b). Assuming (a) is true, why not use whatever method they used (I'm guessing, human-guided instruction), instead of this computerized UI ?
There aren't really great controlled studies, but there's good evidence. One of the best papers I've read on it shows convincingly that older students enrolled in a perfect pitch program don't learn it while the younger ones do. Maybe there's some crazy selection bias going on, but it'd have to be crazy.
Human-guided instruction is expensive, and the point of this project is to make something that can teach a lot of different things to a baby with little expense and little need for parent/teacher oversight. A lot of parents sadly don't have the time or means otherwise.
I understand the motivation, but still, how is this proposal substantially better from a myriad others that promise something to the extent of, "play Mozart to your baby 24/7 to make it into a super-genius uber-baby" ? Ok, so obviously I'm exaggerating for effect, but still, AFAIK the track record for these autonomous baby-teaching methods is quite poor. Also, how old is the original study ? I ask because, if the perfect-pitch training method actually works, and the study is sufficiently old, then (capitalistically speaking) most of the rich people today should have perfect pitch -- but AFAIK they do not.
There have been studies, unfortunately the vast majority are in Japanese. But I found one in english:
https://journals.sagepub.com/doi/full/10.1177/0305735612463948?casa_token=rKW29zt5svkAAAAA%3AJqLB6ycMKbmSsbgOTfqutSeiUWFfxR0OKLWlB_3-8iVj2Zpebt6TF9T3Cf5hNvZlNPaHn2OU2asNwA
It's not randomized but that's really not necessary when we're talking about effect sizes this large. Perfect pitch is incredibly rare (1 in 10,000), and we're talking about a greater than 90% success rate even when some of the kids enrolled were probably a bit too old.
You wouldn't demand a randomized trial proving that hiring a Spanish tutor can teach a child Spanish. You just observe that people do this all the time and it works. That's how perfect pitch programs are in Japan.
Interesting -- so, as I said above, is there a reason why rich people in the US don't all have perfect pitch ? Come to think of it, do all rich people in Japan have perfect pitch ? I mean, if I could give my baby superpowers, I'd consider it money well spent, so maybe they do and I just don't know ?
Perfect pitch isn't a superpower. For a lot of people it's little more than a party trick. I think I should make it clear that my project isn't about perfect pitch and teaching babies perfect pitch is just for the proof of concept.
Because few people know it’s an option, it’s time consuming, fewer people care, and it’s not a superpower. Perfect pitch is perhaps a misnomer. I have a friend who has it and she is a terrible singer and beginner level pianist. It’s just the ability to hear sounds as discrete categories like color, rather than as varying heights like greyscale. Seeing in color doesn’t make you da Vinci and perfect pitch doesn’t make you Mozart.
Speaking three languages is more of a superpower, everyone knows it’s possible to teach a child (children all around the world grow up speaking 3+ languages), and yet the majority of kids of rich Americans only speak English.
K Anders Andersson's book on expertise cites research that East Asians are more likely to have perfect pitch by a lot.
The book is Peak. Pages xiv-xv. The research is by Ayako Sakakibara 2014.
I'm pretty confident perfect pitch is teachable, but only to toddlers. The Eguchi method claims a near 100% success rate as long as it is used between the ages of 2 and 4 (the critical period closes after age 4). Thousands of people in Japan have used the method and developed perfect pitch.
I think you underestimate how much this (maybe not specifically the lessons on perfect pitch) has potential to help the children who are most at need.
I think this is a good idea, but my very small two cents is that perfect pitch might not be the best thing to teach. From my experience being/hanging around musicians:
1. For a child who will go on to become a musician, a strong sense of relative pitch will serve them better than perfect pitch, and having perfect pitch can be a crutch that weakens their sense of relative pitch. In addition, having perfect pitch comes with a "curse" where, as you age, your sense of perfect pitch may drift, resulting in a mismatch between the pitch you perceive and the pitch you're actually hearing. From what I've heard, this sucks, a lot.
2. If you're not going to become a musician, perfect pitch won't be much more than a party trick anyway
Is it just me, or do most of these sound like outright scams -- or, to be more charitable, well-intentioned proposals that are obviously doomed to failure ? Is there any evidence that babies can use computer interfaces, or that *this time* the video courses can finally replace in-person classroom instruction (unlike all those other ones that crashed and burned), or that spirits do exist after all, etc. ? I know there's lots of evidence against all of these propositions (and more), but maybe I'm missing the one slam-dunk argument for them ? If so, why wasn't it in the proposal ?
There's a reason they haven't been funded yet!
118 is mis-titled, it has the same title as 117 even though the content is different
I'm very surprised Scott didn't fund #122: In Vitro Gametogenesis. I'm guessing the problem was the quality of the applicants themselves, rather than the quality of the idea.
The actual problem was a combination of "they needed too much money" and "I was sure some investors would fund it so I threw it at them instead, and then they didn't fund it". I actually think it's very high-quality.
Oh no, they didn't fund it?
Thanks for the support and the opportunity. To be honest, it is very speculative at this point. Would be considered pre-seed for most investors. That of course hasn’t stopped them from throwing money into the alternative technology I mentioned. I actually think our time horizons for getting it to work in humans are quite similar. And then there is the issue of human embryo research which some people find controversial. Anyway, I thought there might be someone in the ACX community with the means are interest in helping move the science forward.
Sorry you didn't get funded, but "our work on gametogenesis would be considered pre-seed" is very funny!
Ha ha! No pun intended. Didn't think of that, but I see what you mean :)
#96 Improve the Readability of Scientific Writing
Hi, I'm a scientist and I'm just... not sure how that's going to work?
Consider these sentences that I just made up but that could be taken from the scientific literature:
"We induced tumors in the Drosophila midgut by overexpressing the yki transcription factor in intestinal stem cells using Gal4/UAS."
"We observed a 5-fold increase in GFP fluorescence intensity after 24-h incubation of the reporter cells with interferon-gamma."
"TGF-beta signaling triggers SMAD phosphorylation followed by changes in gene expression."
How do you make these sentences any easier to understand for the lay reader? I promise, I'm not deliberately trying to make these sentences difficult. Just, they're dealing with complex concepts, and pretty much every scientific article is composed of sentences like these. Sure, the Introduction section is supposed to ease you into the article, but the author assumes that you already are at least somewhat conversant with the field. If you're reading a paper on Drosophila, they're not going to explain to you what Gal4/UAS is.
I really hope I'm not coming across as super arrogant and stuck-up here. If there is a simple way to make scientific papers more accessible without sacrificing scientific rigor and information density, I would really like to know!
Yeah, that is basically the problem with all science reporting. "We reduced the prevalence of type-ABC tumors in Drosophila by overexpressing transcription factor XYZ" becomes "Scientists cure cancer !". I don't think anything can be done about this. You either understand biology/computer science/etc., or you don't, and there's no amount of NLP datamining that can do the work for you. As the famous quote goes, "there's no royal road to geometry".
As a PhD student, I've become very aware that there are many opportunities to improve your writing. I feel that every year, the papers I've written the year before are filled with embarrassing writing mistakes I don't make anymore. What was most useful to me is discussing a specific paper with a scientific writing expert, who works for the university in this capacity. While she couldn't claim to understand my paper, she did find many ways to improve the writing, usually after a discussion of the form "I don't know what these words mean, but judging from the sentence structure... , so perhaps do..." "These thing basically mean this and that" "alright, so what if you write this instead..." And this was a paper filled with mathematics. While she of course didn't look at the proofs, it is still interesting that general scientific writing expertise is still very relevant in a field with rather idiosyncratic writing.
I recently moved to computer science from biology, and in fields that are built more on theory (like CS) and less on mountains of previous empirical work (like biology), I think there would be more potential for this kind of work making papers easier to read.
> How do you make these sentences any easier to understand for the lay reader?
Not the original proposer, but I've thought about something like this for a while, though for software documentation. The idea would be to build a "definition checker". Every word you use has to be defined. The easiest way to do this would be software, something like wikipedia or Gwern's blog on steroids. As an example: You see "drosophila", don't know what it is, press shift and hover it and see a short definition of drosophila. Drosophila itself is defined (thus the Gwern blog reference https://www.gwern.net/index, you can seee when you hover a link it opens something, and the hover thing still works inside, so you can open things very deep).
There are two parts to this:
- the easy part, the software. Find a way to extract the "roots" of words and build dependency graphs (DAG ideally). This is basically how we manage dependencies in software, so we can use that. Someone already built something that looks like this here https://github.com/mjambon/vanity. To apply that to all things written, we can combine that to spellchecker tech. Fortunately someone reimplemented the most popular spellchecker in Python while explaining how it works, so this part wouldn't be too hard too https://zverok.github.io/spellchecker.html
- And now there's the hard part: how do you choose what you start with? What are the axioms? My initial and naive idea is using basic english and leveraging basic wikipedia, but of course it won't be that easy.
The fun part is that we could work with #89, as it's basically for me the same idea: how to bootstrap human knowledge?
If I'm understanding you correctly, it sounds like this idea would already be super helpful without making it fully automatic. There could be a manually curated standard ontology of biological acronyms and experiments somewhere, and authors could manually link to it from the right words in their papers (with some easy spell checking help, probably)
They've posted a sample paper rewrite here. https://etiennefd.com/dgm/platypus-paper-rewritten/ and earlier gave their thoughts in comments here https://jawws.org/blog/2021/10/07/platypus-annotated.html
Like other commenters, I also think there are generally many potential avenues of improvement for papers.
I enjoy many posts on ACX that summarize (review) research or books. Maybe this project could be something like that but if you kept reading, you'd actually get all the info from the paper.
If you ever had to review papers for a conference, you know that this is not about making things easier for a layman to understand.
A badly written paper is an obstacle course. You might see missing steps in reasoning, flaws in logic, graphs that are supposed to show something but show something else entirely, sentences that take 10 lines, narrative that does not start at the beginning and/or randomly jumps around, confusing notations, errors in arithmetics... I could continue.
Luckily, a reviewer has the option of saying that a paper is too terrible to try to salvage. Sometimes, however, the reviewers are not paying attention. And sometimes a paper gets publicized without any review.
Making a badly written paper readable is not an easy task, and it's one that cannot be automated. You have to go back and forth with the authors, trying to figure out exactly what they meant. If the authors are not available, and you have to guess, your rewrite might get the meaning wrong.
Getting authors to follow certain guidelines would be nice, but it's hard for me to be optimistic. Strunk and White is 105 pages, and most people who write scientific papers haven't even gone through that.
I wish the authors of #96 the best of luck, but it's very hard for me to believe that they can make a difference.
Hi! Jess here from #119. Happy to answer questions; also, the short link in my post seems to be broken; this is the url it's pointing to: https://docs.google.com/document/d/1LljPlu4X2V3p4WqQ51dD99LFYWdp6f3QSl9k4OUA7qg/edit#heading=h.sja1agog8fc8
Cut to 10 years later and we're all dead due to #82. The ghost of Scott says to the ghost of Eliezer: "didn't expect that, did you?"
If you believe that #82 is in any way for real, then I hereby propose #82.5: give me 3/4 of the money given to #82, and I will develop a sure-fire solution that will stop it in its tracks before it can eat the Earth. 100% guarantee, or your money back.
(I'm talking about AGI here; obviously boring old self-driving cars already exist)
(yes, that's the joke 😆
Tesla: Hold my beer
the fact that I don't know if you're referring to the scientist or the car company makes this even better
I really like your chutzpah Bendini:
#79: A) Get people to state their intentions explicitly and honestly. B) Put everything in place that’s necessary to ensure that people actually do A, instead of just pretending to do it. The core insight is A, but it’s irrelevant without B’s infrastructure.
...if you have solved that one, I believe you have a decent shot at the Nobel memorial price in Economics in Stockholm. Since this is THE question to solve in the human sciences. You can write the story of human evolution as our attempt to be able to differentiate the genuines from the fakes.
Re: Darwin's theory (embryo version) in The Expression of the Emotions in Man and the Animals, where he suggests that involuntary blushing evolved as a way for humans to signal to others that they are self-bound to behaviour like honesty and chastity, thereby once-and-for-all providing signals impossible to fake, therefore gaining more trust from others, therefore being better able to spread their genes. A theory of why altruism toward non-kin exists builds on the same idea: Giving money to total strangers is a way to appear pseudo-self-bound to honesty in your everyday dealings, making more people willing to take the risk to deal with you, and gaining an evolutionary advantage that way. (Wrote a paper about that myself some decades ago.)
So again, if you have solved this one, humanity will thank you in the for-ever future. And you will probably influence the future path of human evolution. But when something is too grand to be true, there is usually some snag in the theory.
Best of luck though!
PS...and we have just got a demonstration of how difficult it is to do this differentiation, in Scott's blog post of the trials and tribulations he went through in order to decide who to fund and who not to fund among the hopeful.;-)
As a stem cell biologist and oogenesis researcher, I endorse #121 (cell culture media optimization) and #122 (meiosis induction by SCNT).
I hope that #107: RADVAC: Open Source Vaccines doesn't get lost in the shuffle because of pandemic burnout.
#113, David Gretzschel here.
I’m a little bit mortified about that pitch. I noticed that my original whitepaper was doing a bad job of translating my idiosyncratic understanding into good, well-reasoned explanation. And I started wondering if the bitcoin addresses that Coinbase generates for deposit are persistently linked to my account :)
Will get my own wallet when I get around to it.
Of course, I still should have written the pitch by February 5, as I promised.
But in the meantime I was already looking more into memory palaces and long-term memory retention, areas that I previously neglected. And what I learnt from research made me want to reconsider how those aspect fit into the larger problem.
So, I asked Scott to rescind it, so that this didn’t have the “vague promises, secret knowledges and I can totally turn lead into gold, trust me bro”-vibe.
And I figured, that he might just sort me out anyway, because this must have appeared borderline scummy. Well, he said, “no takebacks” and stood by that. And apparently, I’m not that obviously untrustworthy. Fair enough.
I’m glad he posted it anyway, because I got five emails already expressing interest and/or telling me about similar projects and what appear to be similar approaches.
So this worked out great!
Although I’m quite excited to discuss that, this is a rare weekend within the next couple months, where I don’t have time to respond to questions or work on/think about this project.
But if you want to leave questions in the comments or by email, feel free to do so.
Next week, I’ll have time to get to them.
I will state for the record that this is one I did not review under the "it seems fake" clause.
However, I would entertain a revised request (probably on a classified thread) if there was a Substack blog on your efforts (with at least 3 real posts).
"I will state for the record that this is one I did not review under the "it seems fake" clause."
Totally fair.
"However, I would entertain a revised request (probably on a classified thread) if there was a Substack blog on your efforts (with at least 3 real posts)."
Not opposed to getting a public presence once I have the time for it.
Substack specifically? Dunno. I'll let you know.
If somebody wants to be paid for their writing these days, the best place to do so is Substack.
You can do it with a blog elsewhere plus a Patreon, but ... if you want to make it easy for people reading an ACX thread to pay you money, I don't know why you would.
Re: #96: Improve The Readability Of Scientific Writing
Look up Steven Pinker's "The Sense of Style" and his talks about it; he's a famous guy to get endorsement from if you can, once you've got some actual work published, and also the book has some good tips.
Don't try to "create a new journal to publish rewritten versions of existing papers"; that'll get you into weird copyright issues at worst, or just throw people off as weird and radical at best. A thing you could do instead is try to sell your services as an "editing plus" or even "academic ghost-writing" company, working closely with the paper authors / scientists to optimize their language for readibility. (I'd say "editing plus" so they can still call themselves authors while acknowledging your company, explicitly, so you can get advertised in the journals they publish in; maybe even get a contract with an existing journal to be their go-to place to send things to edit.)
Your e.g. website might be able to rewrite a *paragraph* of an existing paper and call it fair use—don't quote me on that, I'm not a lawyer—to illustrate the value and style, but yeah, not whole papers.
Is there someplace one can buy weighted belts to give it a try yourself (I know that it won't be useful for the study but it's an interesting thing to try...and I figure it makes every movement a bit more of an exertion).
Yes, you can buy a scuba belt and accompanying weights on Amazon, for instance. There are also a lot of different styles of weighted vests for sale that are commonly used to add resistance to exercise.
I think it's important to point out that it's theorized that it isn't so much the extra exertion that added weighted provides but actually *the loading of your bones leading to appetite suppression* which leads to weight loss.
What I do not get about this model is that normal, plain obesity will exert extra force on your bones as well, and it does not seem to lead to weight loss.
Regarding #77 (Computer programs which write themselves) how is this different than the Racket project or, more generally, the problem of creating new programming languages? I'm a big fan of Racket (a programming language optimized for writing programming languages) but there are reasons that most programs aren't written in Racket.
Namely, that programming languages are all about tradeoffs. What things are easy to do and easy to think about and what is less easy. We build a bunch of different languages because different domains are easier to program when different things are easier to write/think about.
The more a language/development environment focuses on being able to manipulate programs themselves the less optimized your language will be to deal with other kinds of problems. More generally, as you increase the number of layers involved in creating a program (e.g. you go through some kind of advanced program transformation) you often make it harder to reason/conceptualize what the program is doing. That's a cost.
From what I remember when I looked at Racket, it doesn't allow custom syntax or precise control over computations executed (only the output of that computation). The language layers of this project are meant to be much more freeform. Racket mostly lets you write Scheme like languages. Yes, you can hack it a bit with macros but I find that makes it less first class. Here, no layer is off limits for editing!
I don't understand why what you're describing isn't a cost for other (i.e., popular, existing) programming languages and environments. They have just as many layers, but instead you can't easily edit them. Since the layers are made smaller and since you wrote them yourself, I'd think it'd have a lower cost. Or you'd at least have the power to reduce that (by making less leaky abstractions, editing them out, etc).
The part of the language/environment meant to deal with other kinds of problems is something you'd yourself. The same way you can create a missing library in something more traditional.
Regarding #124 Are you familiar with Steve Hsu's work using machine learning to improve predictions of phenotypes from genetic information?
The project of preserving dead peoples' legacies is really cool. I thought about this idea a long time ago, it's nice that someone is actually trying to do it.
I just started a two year unpaid sabbatical to pursue my personal interests and to complete my MBA. I am 36 years old. About five months ago I discovered Data Science. I've been teaching myself SQL and R for about two hours per day. I am seeking 149$ to purchase a subscription to Datacamp.com . This website is a better teaching platform compared to my University. Please DM @streamust on twitter if you can sponsor me. 📊
For Michael Arntzenius, I understand the motivation is to create tools to program in existing languages, that's directly applicable to work now.
Is there much interest out there in creating new languages from scratch with voice programming in mind? There are a lot of weird conventions in languages trying to make them compact that don't necessarily make sense outside of typing.
#88: as somebody with a physics education I always thought "emergent phenomena" meant stuff like temperature and ferromagnetism, that don't exist in an underlying physical theory but emerge from collective behaviour of many particles. Then when I read Less Wrong I always got a bit confused about how EY liked to rant about "emergent phenomena", which I thought were so uncontroversial. But it turns out it's a euphemism for woo? You learn something new every day.