Isn’t the answer to the last question just that most journalists don’t know what forecasting engines are or think they are valuable? And even for the few who do, they know their audience doesn’t know or care?
I daresay that is a big part of it. I also get the impression that typically journalists write news articles (as opposed to analysis or commentary) on the basis of interviews done with people or reports written at institutions. So if prediction markets are important it would be for the people being interviewed to bring them up.
I have seen prediction markets and betting odds discussed in articles about elections and the like, already quite a long time ago, so it does happen.
Yes, the media like Salon used to cite the Iowa Electronic Markets and then Intrade all the time. This was during the big wave of interest in election polling & forecasting that more or less culminated in Nate Silver. I am not sure why it all disappeared, although Intrade shutting down due to some sort of embezzlement issue certainly didn't help. I suspect that the culture-war turn didn't help matters.
I think you nailed it. Journalists write articles to get reads, and if consumers are more likely to read an article because it relies on them because it's part of the current zeitgeist they'll include them. But once it leaves the zeitgeist journalists don't really have any reason to care about a prediction market.
Yes, exactly, from my experience as a journo (not a generalist one, but still), the second bit would be one of the main reason: in the context of a word-limited piece (as most are), you don't want to have to explain (i) what it is; and (ii) why it's worth taking into account. This might change when prediction markets become common knowledge.
Add to this that citing sources or external info is not something you like to do in practice. Sure, if you are being asked, you should be able to back all statements. But when drafting, you get the urge to make it as though you had the knowledge all along, so as to appear smart, but also ward off competitors.
In this context, citing people/experts on a given matter is much easier, partly because it's also a way to build up a relationship with that person/expert to obtain more info down the line. You don't get that added value with public-facing websites or sources.
As a (admittedly noncentral) reader, I hate the norm of not citing sources. I think ~all numerical claims should come with citations. I think a writer comes off as smarter and having better epistemics when they cite sources.
My strong suspicion is it's due to the Silicon Valley-New York (or tech-media more broadly) interelite conflict, and the journalists are obviously on the side of media. But I'm not sure how you would falsify that hypothesis.
But if that were the complete answer, sports journalists would include odds from bookmakers and betting sites in their stories far more often than they do. After all, those are fairly well known and understood by ordinary people, and are easily accessible.
I think another reason is that there are two sides to journalism: Reporting and commentary.
Reporting is to check facts and report on what has happened. True reporters love numbers and hard data, but focus on the past and undeniable, and avoid anything that looks like speculating about future outcomes. (Pure reporters are very old-fashioned, and rare in our age.)
But commentary has become far more common, and most journalists now see their job more as something like "telling important stories" rather than reporting. Of course, good journalists can do both – that's what they try to teach you at J-school (at least when I went, a good while ago) – but most journalists aren't good.
Mediocre journalists don't understand or like numbers, but perform their job work by gathering gossip and anecdote, and compiling BS that aligns with their biases. They prefer salient quotes and soundbites to impersonal data and bright lines for lots of reasons: Numbers don't evoke enough emotion (show, don't tell, etc.). Cold facts rarely line up neatly enough, but complicate the narrative. Also, they are hedging their bets by being "nuanced" (read: vague).
Most journalism is a mix of these two, but whereas pure commentary has become very common, pure reporting has become rare and old-fashioned.
One factor, IMO, is that a forecasting engine that worked would work against the actual business of journalism -- producing an endless string of horse-race stories about something. The last thing a journalist wants to write is that it's clear who is going to win an upcoming election because that means they have to find something else to write about tomorrow.
A particularly annoying case was the Fetterman vs. Oz election, which IIRC was the last undecided seat in the US Senate in 2022 and thus attracted a lot of attention. The web site Five Thirty Eight originally made its name by predicting elections accurately, but in this case they reported daily on the election as it was coming up detailing which factors favored each candidate, as if there was a reasonable chance either could win. In fact, Ladbrokes was offering 10-to-1 odds that Fetterman would win. So a maximally factual article would say, It would take a miracle for Oz to win and not bother reporting on it until election day.
The rationalist fascination with Aella is bizarre to me. I really don't have a better explanation than 'nerds enamored with only cute girl to pay them any attention'. She gets treated like she's some sort of intellectual, but watching the Lex Fridman podcast episode that featured her, it's very clear that she's...not. Does she have any kind of impressive prediction record that I'm ignorant of or something?
I find her blog and survey pretty interesting - see https://aella.substack.com/ . As far as I know she doesn't have any amazing formal prediction record, but neither do I or lots of other rationalists. She just writes well, thinks clearly, and has good opinions.
She's a replacement-level controversial blogger (who already often get some following) who's also if roughly the 90-95th percentile of both "attractive woman" and "willing to get naked online". This makes her more famous both for the direct effect (some guys will follow her for that) and the schelling point effect - if you've got a dozen similar mid-decent Hanson types but want to coordinate on everyone following the same one, might as well focus on the one who also stands out for an unrelated reason.
This assumes there are many good bloggers and we need to find some reason to prefer one over another. This whole conversation is predicated on "she's just a person who writes well and consistently, why do we care about her so much?"
At the risk of saying self-interested arrogant things: basically nobody writes well and consistently. I write moderately well and consistently, and make more money as a blogger than I ever did practicing medicine. Writing well and consistently is so rare that it it fully explains all of her fame and then some.
I don't think we have "a few dozen mid Hanson types". We have Hanson himself - who has 80,000 followers on Twitter - Aella - who has 176,000 - and a handful of others who are also popular. If there were other people as good as those people, they would also have tens of thousands of Twitter followers and be famous.
I actually sometimes occasionally read Aella for fun when I'm bored. This is a really high bar; anyone who reaches it will be famous no matter how pretty they are or aren't.
To put my original claim a different way*: I think aella is about, say, a +2sd blogger in terms of writing skills/edginess. It's a nontrivial and fairly impressive level of skill, but not a massive outlier, and is thus insufficient to explain her fame (which is +3-4sd). However, being +2sd on two other things that are mostly uncorrelated (but help with fame) combines to make her a significant enough outlier to get this level of attention.
*I'm tempted to get into a long aside about how to estimate the exact number of similar-skill bloggers and the ability distribution, but while I think we'd still have different opinions on the exact distribution and her placement in it I think my main claim should survive them.
It often seems the other way around: People assuming Aella is "fascinating" (as you say) to others only because of her attractiveness seem to be the ones who can't decouple her appearance from her other work. Or at least overemphasize it. She's prolific in very different ways that can equally explain the fascination.
I agree. I have paid virtually no attention to her naked activities and find her totally fixating. She writes in that brilliant rationalist way that emphasizes plain language to avoid your thought process from getting lost in complex language and tripped up by assumptions built into it
This goes with her attractiveness to set up a road block for people who aren't engaged by her work to assume it's all that. I'm currently avoiding her because my inner Chesterton's Fence is chronically annoyed but I can't think of anyone else particularly comparable to the plasticity of her curiosity
Well, a survey with a sample size of thousands has a value in and of itself, I'd say. You really can't discount that.
There's also the fact that she's willing to be un-PC, which obviously a lot of rationalists like.
The fact that she's the sort of girl nerds dream about and rarely get obviously has a lot to do with why she gets those huge sample sizes. But if it lets her do research academia wouldn't dare do, well...I might not think Musk deserves his billions, but I like his work with electric cars, and I want humanity to go to Mars.
> She gets treated like she's some sort of intellectual, but watching the Lex Fridman podcast episode that featured her, it's very clear that she's...not.
I came away with the exact opposite impression. She was having none of Lex's tendency towards imprecise and poetic descriptions/questions, and rightly drilled him on exactly what he was trying to convey.
I'm tickled that Destiny is the kind of guy who gets invited to speak at conferences these days. The last time I remember hearing about him, he was "that Starcraft II streamer who keeps getting in hilarious fights with Deezer and CombatEX over Skype."
Just gotta try hard and believe in yourself, I guess.
Aella's done more genuine data science work than the supermajority of this comment section. (And probably contributed more to the corpus of human knowledge, tbh)
I don’t think her invite is important enough to argue over. She’s a rationalist internet celebrity; that’s all you need to know.
But I do think the idea that she has done “genuine data science” work should be pushed back against. She can’t have done more genuine data science work because she hasn’t done any genuine data science work. Her work is both hopelessly compromised by methodological issues (no controls and poor/non-random sampling) and does not utilize the scientific process (opinion essays that use rationalist rhetoric to present personal experiences). This is not to say that she is incapable of doing some but to this point she has not done any so she cannot possibly have done more than anybody else in this comment section.
She posts informal polls that are entertaining and occasionally suggestive about the opinions of people who pay attention to her, which should be enough to justify further interest anyway if you must justify it.
I would not be surprised if many of her critics would have equally skeptical stance to studies of college students who walked through the wrong door in a library or internet surveys.
In many ways, humans are quite similar to each other, so it is easy to make case for it when studying mechanisms that are shared by all humans. As a point on the critics' favor, the social and psychological phenomena, everything that can concerns the mind and behavior in fact is the dimension where the humankind shows most variety. (To educated layman, like me, I should add). And the internet has the weirdest niches of it all: easily available, easy to self-sort into bubble networks. It makes so much intuitive sense why any finding with such sample would fail to replicate in some other population due to some bizarre selection effects. But on the other hand, sometimes intuitive common sense is wrong. Though I have no expertise in psych-related, from what I understood, the replication crisis didn't mean all such studies failed to replicate.
Tangential thoughts:
+ Any sample, including the college student / random internet study sample, becomes is more interesting if you can perform randomized controlled experiments. Causal effects are still causal effects, especially if they replicate, even weird causal effects that replicate in weird niche populations.
+ There are many things to dislike about scientific peer review as part of publication "stamp of approval" process, but all comments I ever received were useful. (If the reviewer just didn't understand something or had ultimately inapplicable advice, still thinking it through was useful experience.) I don't know if she has tried to publish, but that would be a way to get feedback (and occasionally, the stamp of approval).
+ Suppose one has a study and set of results. A question of some kind of bias arises. Various, increasingly clever sensitivity analyses are presented to address it. A fact that I find weird is that it is surprisingly rare to see a numerical exercise showcasing exactly how much bias there would be needed to observe any particular result.
She does way more than twitter polls. What part of her methodology do you take issue with?
As for non-random samples, there's a dozen reasons why that's not a valid critique (not the least of which is that her samples are already much better than a huge portion of published research). But I'll just link her post on it here:
Anyways, are YOU a data scientist? Have you been paid to do data science? Because I see multiple actual data scientists telling her that what she does qualifies.
I will fully admit to having a higher standard for rigor than many so-called scientists of different stripes. I don’t think I’ll be lowering that standard in the name of inclusivity.
I am not a data scientist, though my job involves certain elements of the prototypical skill set. It would be bizarre if the world of ‘data science’ was so subjective that it could only be evaluated from within, as if it was the qualia of an acid trip.
>It would be bizarre if the world of ‘data science’ was so subjective that it could only be evaluated from within
No, but presumably people with the title should have more say in the boundaries of the category then someone who is not a data scientist. The criteria you've implied to fail her for would also exclude many, if not most, data scientists.
But this is all moot, since the criteria you initially gave makes it seem like you either aren't actually familiar with her actual survey/analysis work (since you claim it's all personal experiences, which is false, and lacks controls, which is either false or not applicable) and are refusing to update on the information you've been provided with, or are arguing in bad faith by using an esoteric definition of "data science" and refusing to provide that definition.
"Make your claims clear, explicit, and falsifiable, or explicitly acknowledge that you aren't doing so (or can't)."
He was the first (and so far only) streamer to be interested in prediction markets, his fans got into it, they created a lot of Manifold markets on who he would get in beefs with or whatever, and now I guess he is part of history.
Other big political streamers are commies allergic to anything market-related. Outright rightist can't get big in the current climate, and Destiny is a weird contrarian always balancing on the verge of being completely cancelled, but until that happens he's the best we're going to get.
I'm curious what it is about this format that a rightist can't get big in it. I'm not familiar with political streamers, really, and hadn't heard of Destiny. Maybe I'm too old for this stuff? (Older Millennial). Or maybe I'm just allergic to spending my time listening to unscripted ranting, which is what this sounds like to me.
Following the pattern of other formats, you would normally think that even if the market is dominated by leftists, there would still be room for one rightist to grab some market share by sole virtue of being the ONLY rightist voice.
Well, the way that people get big in streaming nowadays is through networking and collaborations, with all the attendant echo chamber effects. If some lone rightist threatens to get big in spite of that, it's easy to find an excuse to ban him, because "trust and safety" structures are also staffed with leftist activists.
Speaking as someone who had a real job for decades, that describes a lot of rightists. I remember when Ayn Rand was the "in" thing. Generally among people with no experience.
Kids tend to get swayed in various different ways. One of the characteristics is that the way they are swayed is ... unpleasant ... to their respected elders.
Yes, I was going to say, Martin Blank's comment doesn't ring entirely true to me. As you point out, Ayn Rand is someone who appeals disproportionately to the young. As does, to a lesser degree, libertarianism more broadly. I thought it was the young that catapulted figures like Jordan Peterson and, dare I say, Andrew Tate.
I'd say one of the most salient and enduring political tendencies of intellectually-inclined young people is a preference for unnuanced idealism over a more pragmatic or nuanced politics. Marxism is still a leading brand for unnuanced idealism, but there are others.
Key quote: "Last summer forecasters on Metaculus, an online prediction platform that is a favourite of many techies, thought it would take until the early 2040s to produce an ai capable of tricking humans into thinking that it was human after a two-hour chat, had good enough robotic capabilities to assemble a model car and could pass various other challenging cognitive tests. After a year of astonishing ai breakthroughs, Metaculus forecasters now think that this will happen by the early 2030s."
Thank you! Though I wouldn't say that's the Economist citing it in exactly that way, even though it's on the way to it. First, it has to introduce what Metaculus is (of course, at this point a necessity), which it others with the "that is a favourite of many techies", biasing those against tech against it more than they already would be. And, the "AI Astrology" in the graph really rubs me the wrong way.
I wish metaculus had a more general LK-99 replication question. It seems silly and short-sighted to restrict to only the first replication attempt, knowing how slow and difficult the process can be.
re: "why would any agency give a no-action letter now". Because it forces the targets to engage in expensive, time-consuming legal and regulatory activities to fight back, and a lot of times targets won't or can't. Then the precedent gets established for free. Basically it's an end-around the regulatory process. The PredictIt case isn't exceptional. Agencies often do this kind of thing (they all have their own mechanisms) before initiating regulatory proceedings, just to see if it works. It costs them very little, especially compared to the normal regulatory process. Worst case, a judge tells them to do it better, so they do. Worst-worst case, they step back and follow the process they should have followed to begin with.
Obligatory not a lawyer, but I have spent a lot of time working with regulatory data and processes. Things like no-action letters are already considered legally binding commitments to a degree (I don't think this is a novel ruling), and agencies are held to something resembling normal standards of justification. Biggest difference is that they don't need to go through the public comment process this way, which is tedious and annoying and which they mostly hate. Because, as you note, a bunch of wingnuts spam them and they have to (are supposed to, at least) prove they've responded to every comment.
"Because it forces the targets to engage in expensive, time-consuming legal and regulatory activities to fight back, and a lot of times targets won't or can't"
I'm confused - I thought a no-action letter was when an agency said something was okay. Who is fighting back? Can't they just never tell anyone that what they're doing is okay, and just leave things up in the air?
It's kind of weird that they ever decided on an official policy of selective enforcement, right? Like a no action letter is basically saying "we have a rule against this, but you guys seem pretty nice and harmless, so I guess we won't shut you down right now. Maybe later," and then if you decide to shut them down later they can sue you because... what were you thinking. The whole idea seems arbitrary and capricious.
Wiki says a no-action letter is given when an behaviour is technically, legally NOT ok but the agency won't do anything right now because common sense.
The wiki article is rather bare bones and has a [citation needed] for the above claim, so there's that.
"In addition, the SEC staff reserves the right to change the positions reflected in prior no-action letters."
So I suppose "fighting back" against a no-action letter doesn't make much sense except if you want a clear-cut, legally binding approval of your proposed actions. Then again, why would you request a no-action letter in the first place.
I don't know, with no money on the line I could definitely see people making wacky answers just for fun (which is what 'lizardman' originally refers to, 7% of Americans answering 'yes' to 'lizardmen controlling the government' because they think it's funny).
The caveat is that silly answers will probably reduce the amount of play-money they have, reducing their impact, unless they spend real money to increase it again.
This works if the number of new accounts increases slower than the rate at which they lose play money by betting on silly things. It would be nice to model that
I would just like to formally register that there is rather little steel in either the 757 (American Airlines Flight 77 and United Airlines Flight 93) or 767 (AA Flight 11 and UA Flight 175).
Oh, so now you want a mystical prophet who is also an aeronautical engineer? You will be insisting that they make clear and unambiguous predictions next, I suppose?
I kind of thought that, had missiles struck the US somewhere, those would be considered the steel birds. Or maybe if a crowd was attacked by a bunch of drones. Or a big plane crashed accidentally where a lot of people were.
Looking at it, I considered the prophecy too vague to be dis-proven.
Amazing that LK-99 has 25~30% probability. When I first read it in Korean news I obviously thought it would be some ridiculous scam. Then I read that superconductivity related stocks are soaring. And now the prediction markets. If this resolves true, I would be seriously impressed, in both ways.
I still think it is a scam/mistake, but the soaring stocks and prediction markets are exactly what I would expect, and why I don't think prediction markets would/will be a great way to make decisions on Issues Of Note. 'If this is true we'll be rolling in the dough' is enough to make people rush out and buy stocks in the hopes of money money money, and then prediction markets see the stocks rising and go likewise because surely if all this money is being poured out, there must be a basis for the optimism! As if bubbles never happen.
There's more than one group. One of the groups is almost certainly either a fake or a useless product. (Needs extremely high pressure, etc. Specs to reproduce it aren't clear, etc.)
The other group is claiming essentially STP for the final product. (Actually boiling water is supposed to still be stable.) And they also provided enough info to allow others to reasonably try to replicate it. That's the one that's 25-30% probable.
It's just happenstance that both groups made announcements at about the same time.
B, but if as seems likely it turns out to be yet another "oops, sorry" dud, then A because Trump running or not for the next presidential election will affect the USA,. the lives of those in the USA, and by the ripple effect of influence the rest of us.
I would think that if the precedent set is that you can hound high profile opposition by endless lawsuits then you would have seen a lot against Joe Biden. Instead the precedent seems to be that you can hound high profile criminals by endless lawsuits, which is why you see a lot against Hunter Biden.
Right - but he does have some charges against him, unlike Joe Biden. This all seems to me to suggest that, while politically prominent people might be slightly more or less likely to be prosecuted for some things than other random people, there doesn't seem to be open license to persecute any politician one wants, and there also doesn't seem to be total impunity.
Trump got his hush money payment sliced and packed into 26(?) different charges.
Hunter got his three explicit Federal gun felonies consolidated into one misdemeanor. And I have no idea how many state/local gun crimes weren't charged at all.
> Trump got his hush money payment sliced and packed into 26(?) different charges.
If the charges are not valid then Trump can challenge them in court. He has appointed 234 Federal Judges and we all know he is a great judge of character and we can be sure none of those judges are part of deep state.
Besides Trump is not a stranger to courts, Trump & Friends filed around 63 lawsuits challenging various aspects of 2020 election but unfortunately lost in all of them. This must be due to the deep state. Hopefully if this goes upto supreme court they will support him and make him the president. All Hail president Trump.
Yeah. The document precedent was already set by Hillary Clinton, she didn't return the top secret documents when ordered by Court and later tried to delete video footage.
She made electronic copies, and kept them on an insecure server where they could have been (and were unless you're pretending the Chinese are incompetent) accessed by anybody. And yes, she NEVER turned over that server.
But yeah, paper copies are magically worse. Go with that.
Hey! Let me ask you a honest question, are you mentally challenged? Cause that's the impression I'm getting from your comments.
> Return what? She made electronic copies, and kept them on an insecure server where they could have been (and were unless you're pretending the Chinese are incompetent) accessed by anybody.
Stop spreading bullshit dude. Before you make absurd claims like this at-lease read the Wikipedia https://en.wikipedia.org/wiki/Hillary_Clinton_email_controversy As part of the FBI's Midyear investigation (code name "Midyear Exam"),[141] at the request of the IC inspector general, Clinton agreed to turn over her email server to the U.S. Department of Justice, as well as thumb drives containing copies of her work-related emails. If you are mentally challenged and can't comprehend Wikipedia articles maybe you learn how to read and then you can think about commenting in astral codex.
> And yes, she NEVER turned over that server.
Trump didn't return the documents when ordered by court and tried to delete the video recordings. And being incompetent he failed to delete the video recordings.
Re journalists, I think it's pretty simple. English/history kids who are afraid of math and numbers become journalists who are afraid of AI taking our jobs, and numbers. The intersection of people who write articles for mainstream publications and people who read publications like this/know much about prediction markets is vanishingly small. Not to mention, there's a bandwagon effect: no one wants to be the first one to start using this weird probability thing when no one else is doing it -- especially considering the hate the weatherman, Nate Silver, etc. get for "being wrong". There's little incentive to do it, and plenty not to.
Hmm, I disagree. As a sort-of-journalist-adjacent-person, I'm constantly being asked to have opinions on things outside my area of expertise, and prediction markets are a great way to do that ethically.
A big part of journalism seems to be about trying to come as close to the brink of expressing an opinion as you can, while still seeming objective. So if two countries are getting close to war, journalists know their readers want some kind of clarity about "will they go to war, or won't they?", but if they interject their personal opinion, they'll be breaking norms plus lose trust if their opinion turns out wrong. So they try to launder their opinion by eg asking experts, especially experts on their side of the aisle who they know will agree with them. Then they can have an article like "HARVARD INTERNATIONAL RELATIONS PROFESSOR: WAR INEVITABLE". Prediction markets would be one more way to do this.
Journalism is dominated by people-people, like most establishment types. They don't like or intuitively understand numbers, data, the scientific method, or abstract distributed systems of knowledge processing. They keep these things as far away as possible. They feel sure that knowledge, moral purity, wisdom and so on are rare and highly concentrated in a small number of excellent individuals. That's why so many of them hold markets in disdain, talk about algorithms as if they're magic spells, and many are deeply unsure about democracy (being as it is a system for collecting distributed wisdom) - frequently preferring terms like "populism" and "fascism" which are understood to be left-coded ways to signal disagreement with the foundational assumptions of markets and democracies.
Instead they work in terms of individual people and their brands. A friendly local academic is right up their street: you can phone them up at any time and they will talk for hours if you let them, so you can just ask them a question, they'll give you an opinion, and nobody will blame you if you repeat whatever they say and it turns out to be wrong because how could you possibly have known that a Smart Person was wrong? Like, that whole expectation would be completely unreasonable! They have a title! They work at a famous institution! The brands are impeccable!
Quoting some obscure prediction market runs counter to all these instincts. Firstly, it's a market i.e. assumes that wisdom is distributed instead of concentrated. Secondly, none of the markets have any well known brands, and none of the traders are even identified at all. Thirdly, they yield outcomes in the form of numbers. Finally, if it turns out the market was wrong, you can't deflect blame onto the market itself because none of your peers respect this form of knowledge either.
Democracy may occasionally function as a means of collecting distributed wisdom, but that's not it's purpose. It's purpose is to keep people moderately satisfied with their rulers, so there are fewer insurrections and civil wars.
Note that there are other associated functions whose purpose is to ensure that nobody with sufficient power to start a civil ware feels sufficient motivation to do so. That's not directly part of democracy, but it's been associated with it since the Magna Charta.
Democracy is a terrible way to determine truth but, since preferences are inherently subjective, it's better confined to aggregating those than wisdom.
Yeah and it feeds on itself, since there are people who got quoted once as an "expert". They pretty much advertise themselves as having a reliable opinion.
When a journo wants an expert to back up their slant they call the relevant person and ask for another quote. Voila: expert-backed, so it's not opinion it's news.
I think that news stories are often written to suggest that some kind of interesting outcome is possible, or that an event which occurred has possible interesting implications. On a lot of these topics, the prediction market will often be pretty sure the interesting thing won't happen, and including something like "Metaculus forecasts a 5% probability of China invading Taiwan by the end of 2024" would undercut the rationale of printing an article on a potential invasion.
I also think newspapers would rather print "July 2023 is on track to be the hottest month ever" than a percentage chance of 2023 being the hottest year ever, for example.
Journalists definitely look dumber than they are a lot of the time because they're reporting on something they know little about, but on the other hand, this seems like a good guide to how seriously you should take an article on, say, fusion power or gain of function experiments or AI, written by someone who knows approximately nothing about these fields except that he briefly interviewed a couple experts whose statements he mostly didn't understand.
And at the same time, as best I can tell, most journalists are innumerate and scientifically illiterate, at least compared to anyone I'd want to hear talk about science and technology. (They're probably mostly less so than their readership, though.)
Sure, but as you mention, they already have a way to do it -- by asking "experts". In fact, the current way, from their perspective, is probably better, since "war inevitable" sells much better than "64% chance of war". It also allows them to package the same opinion multiple ways "Harvard IR prof" vs "Stanford historian" vs "geopolitical expert" vs ..., might all say slightly different things, and there you get plenty of articles out of it, whereas "the markets say X" is about all you get out of that style of reporting. It's boring, and more importantly, it's illegible. No one knows what 23% means, really. They just round down to 0%, and then get mad at you if it ends up happening if you talked about the 23% too much (cf. 2016 election). It's opinion laundering, but in a way that makes the end product really lame to read and might get people mad at you, and definitely will confuse some people, if not the journalist themselves (which I think is a larger part than you addressed in your reply).
I think it’s much simpler. Regardless of who a *journalist* is, if they want to cite a source to back things up to their *audience* then they need to appeal to a source whose expertise is legible to the audience. I’d the audience doesn’t know what a prediction market is, then it’s better to just cite a random government official or university professor.
A10(TO4)6X2±x, where A = alkaline or rare earth metal;
M = Ge, Si, or P; and X = halide, O, or OH. The
name ‘apatite’ derives from the Greek apat¯e meaning
‘deceit’ as a result of the diverse range of forms it can
take [17].
Here I consider the lead-phosphate apatite
Pb10(PO4)6(OH)2."
a) "TO4", then "M = Ge, Si, or P"
1) I assume TO4 should have been MO4
2) Ge and Si are in the same group, but not the same as P. Maybe As was intended?
b) "X = halide, O, or OH" Yes, halide and OH can substitute (F/OH substitution
is why fluoride helps tooth enamel) but O has an extra charge...
c) "where A = alkaline or rare earth metal" And lead is neither - though it _does_ have a +2 charge in many compounds, as do the alkaline earth metals, so that works chemically.
These are all nits, since the compound of interest is explicitly specified, but I wonder about the proofreading of the paper...
I'm surprised that people are taking a purely computational paper so seriously. I don't know any DFT myself but I talked to some colleagues who do and they seemed rather indifferent to it.
Lol thanks for featuring my extremely uninteresting comment but unfortunately my name is not Muhammed Wang, it is indeed just a pseudonym that I use sometimes
Which is funny because both the Polymarket and Manifold ones are both spiking sharply up in the past few minutes (I assume some new positive news must have come out)
Not surprising though, for the reasons Scott pointed out.
Gaia says: "Resolves NO, since neither of two replication attempts published on the ArXiv today by teams in India and China, respectively, were successful:
Synthesis of possible room temperature superconductor LK-99: Pb9Cu(PO4)6O
Semiconducting transport in Pb10-xCux(PO4)6O sintered from Pb2SO5 and Cu3P"
It's mildly amusing that, due to the manifold markets being a live, updating link, at the time I'm reading this, rather than being within a few points of each other, the manifold market is up to 50% while the static polymarket view is still showing 24% (although it has also increased to a little over 30% at the time of writing...must be some news for them both to increase!)
Baba Vanga is famous in post-Soviet countries; so much that in Russian her name turned into a verb meaning "to predict baselessly/ironically".
Although I'm of the opinion that she was a KGB asset (https://www.vagabond.bg/russia-brings-vanga-3460) to influence leaders who base their decisions based on mystics, astrology and such; and nowadays a common psyop method.
Re: election betting, maybe there’s a legitimate fear that already-partisan people may exaggerate their positions more if their own money is directly tied to the election. If a media editor has $250,000 riding on an election, they might be less motivated to resist their outlet’s biases in pursuit of truth. That seems like a really high limit to me.
Naive question: Which *fundamentally new capabilities* has current-generation AI got which were not already present in expert systems, Markov chains, etc? I see the ChatGPT responses, and all I can think of is "cute, that's a bit like the dozens of articles back in the day about how we can generate fake Italian-looking names, except with whole articles."
This is a weird question to me. If you abstract it away enough then the LLMs aren't doing anything "fundamentally different" than a simple python script that outputs "hello world!" The dramatic increase in quality over what was previously available is what is novel and surprising. Yes, earlier AI was able to do some of the same things that these LLMs can do (although nothing was able, as far as I'm aware, to do everything that these LLMs can do in a single AI), but GPT-4, Claude 2, and others are _dramatically_ better at almost everything than previous AIs were.
First, there is a question of scale: Eniac was not "fundamentally different" from my laptop: both are in principle Turing complete (at least given infinite external memory), so they are capable of any computation (in principle), but in practice we all see rather massive difference. Also, current LLM's in particular, contain a fundamental innovation in the form of transformer architecture, compared to earlier recursive networks which had very short attention span and poor parallelism. Thirdly, deep learning in general, compared to earlier statistical methods, including Markov chains, but also shallow networks represents a kind of "fundamental change" because of its ability to form and use abstractions. And, finally, the fact that even simple statistical methods like Markov chains can generate text which can sometimes plausibly look like a text in a chosen human language with some traces of emergent grammar and sense tells us a lot about ourselves. Does this mean we are close to AGI (whatever it means) or at least human level intelligence? I am deeply skeptical, I think there are still many pieces of the puzzle missing (frankly, we do not even understand well the dynamics of learning in deep networks), and that we may soon see again blog posts discussing whether progress in AI has stalled (you might find amusing to read some discussions in the recent past about stalling of AI research, I found one from 2018!) but it is not all hype by Open AI and other big AI enterprises. The current progress is real, relevant, and quite fundamental, also for understanding of our own nature
I mean, the absolute biggest one is true parsing of natural language input.
It's weird to even include expert systems in this question. Expert systems are essentially just a static set of (usually) yes-or-no questions, hooked up to a bunch of "if" statements (every outcome of which is hand-picked). It's fancy version of a uQuiz. Like, not even similar.
Markov chains *are* sort of similar, but they're just... not good? They don't display even a facsimile of reasoning or recognition of the content of the input.
Also I think the year we've had with them undersells just how much of a leap chatGPT was towards passing the most rote kinds of humans. Just yesterday I had an interaction with a tech support agent who, based on our conversation, acted *less* capable of cognition and problem solving than my experience with CGPT.
Tech support agents have a script to follow and for a lot of them English is a second language (at best). So what you're seeing there is the speed at which the human can follow the script plus the number of off-script answers you make.
Rather like the previous answer saying that ENIAC was not fundamentally different than your laptop. That's true, but ENIAC was not able to compute fast enough to e.g. produce video output in real time. I don't know if you would consider this capability a difference in kind though.
Writing grammatically correct sentences in meaningful paragraphs about a broad range of topics was way beyond anything that expert systems or Markov chain models could do.
Code synthesis. Before GPT-3, code synthesis was essentially a pipe dream - there's an "old" (2015) article¹ by Gwern where he attempts to train a pre-Transformer architecture RNN to output CSS (which isn't even a Turing complete language), and it's... bad. "Success" in that case meant "it produces random output that sort of looks like CSS". GPT-3+ can synthesize syntactically and semantically correct code in every popular programming language just from an english description. This is fundamentally new.
Taking about the use of Kalshi as a way for businesses to hedge against world events (like elections), does this distort the market as a pure prediction market? I have no economic knowledge about how hedging impacts say stock or commodity futures markets, but obviously hedging is a big part of the trading that goes on there and those tend to be accurate. But there the “bet” made when trading is directly tractable to the outcome being predicted, i.e. the future value of a company or commodity adjusted for risk, and hedges represent a form of risk adjustment that is directly related. It seems to me like substantial hedging on events that resolve to a yes or no would change the dollar value of a position in the prediction market from purely representing the market’s predicted likelihood of that outcome occurring to “the likelihood of this event occurring weighted by financial risk to the participants in this market if the event does occur” so there’s information being represented about things beyond the event itself. Maybe that balances out when market participants have risk on either side but for your example if Bernie Sanders we’re elected I would think all of those Wall Street betters are assuming the same direction and a very similar magnitude of the financial risk it represents for them. Of course if the market gets too skewed someone comes in to take the other side, but is the balance of the seesaw still weighted toward market-wide directional risk like this? Or am I not understanding how hedging impacts a prediction market (would be easy for me to do)?
It comes down to volume. If the volume is high enough, these hedges won't move it, because to whatever extent a hedge _does_ move the market, someone else sees that movement as an opportunity to make money (which it is, to whatever extent it moves the market away from a correct prediction), and so more money comes in and it corrects.
If the volume isn't high enough though, then yes, these kinds of hedges would indeed reduce it's quality of prediction. But it's hard to imagine that state of affairs lasting very long. the EMH is by no means absolute, but it doesn't _consistently_ leave the _same_ $100 lying on the ground.
Citation needed. I'm skeptical you have an example that is similar in the ways that matter to the above hypothetical. This is the kind of example that is one of the very few ways where the EMH would actually operate mostly as expected in the real world. There are lots of cases where the EMH doesn't work in the real world for all kinds of reasons. I just don't think that "hedgers move the prediction market away from the consensus prediction" is one of those.
Yes, if there's a systemic factor causing one side of the hedge to be more desirable (e.g. there's a presidential candidate everyone thinks would be bad for the economy), this can happen. It already does happen in the stock market - put options on the S&P500 are worth more than you'd expect purely based on the distribution of actual S&P returns because they're used as hedges.
(This does in principle mean you can make money by selling SPY puts, but it's picking up pennies in front of a steamroller, so not enough people are willing to do it to overcome this imbalance).
Right, which is exactly the point: There is a limit to how far hedgers can move a market. If they move it too far, those pennies become dollars and someone dashes out in front of the steamroller until all that's left is pennies again. So in an extremely pedantic sense, yes, hedgers can and will move the market. But they will only move it so far that no one thinks it's wrong enough to be worth correcting.
lol at "a couple of times a day I check the prediction market on [superconductors]" as if it were the stock market or expert opinion or anything except the agglomeration of what some random retards online think
"It’s always “We Are Various Prestigious Economists, And Have You Considered That Economics Tells Us That Prediction Markets Are Good?”
Feck it, now I am *doubly* against prediction markets if the economists are out there telling us we should have them.
Look, the Smart People want their fancy betting markets because they're too high class to step inside William Hills and put 50p each way on the outsider in the Gold Cup? Sure, let them, but let's not kid ourselves this is anything other than "I want to use my Giant Brain to make money" and not a way of "we will find out the best way to decide 'is arsenic a health supplement?' by using the forces of the market".
(2) Robert Miles will be at this bunfight, eh? Maybe he can give you all a selection of his hits:
(Yes, I'm sure he's never heard *that* joke before).
(3) I do absolutely want to get some informed opinions on the superconductor. My priors are that this is going to turn out to be a hoax, fraud, or over-enthused reporting of minor results like the previous efforts. If anyone can give me a semi-solid "nah, this looks sorta legit" rundown I'd appreciate it.
(4) "I ran into some finance people at the NYC meetup this week and asked why they weren’t using more advanced forecasting technology - prediction markets, superforecaster tournaments, calibration training, that kind of thing. A common reply was “who says we aren’t using it?”
That's the part that makes me repose even less confidence in "Allowing election bets makes Wall Street less interested in elections, not more!" As you say, knowing the outcome of elections is worth hundreds of millions to them. If they have their own internal markets, which live or die by being correct, and that market is telling them "X is likely to win the next election", it may well be in their interest to try and make sure Y is selected instead of X during the primaries, or to help out Z to get elected instead. Maybe they throw money at Z's campaign, or they run advertising, or they put out subtle influences to influence *other* prediction markets, or they get polling companies to do polls crafted to return "Z is the people's choice!" and then release those results.
My take on it is that if there's a lot of money riding on the result, then hell yeah somebody will try and influence the outcome:
I am very confused about the US government's overregulation of prediction markets. Is there any rhyme or reason for them to come down so hard on this compared to other countries, when the general rule (or general perception anyway) is that the US regulates most things much less than most other western countries?
Also, is this extreme hostility to gambling (political betting is downright illegal, which I was very surprised to learn; it's a completely normal part of political campaigns in Australia) primarily seen as a left-wing socialistic thing, or a right-wing moralistic thing? I'm guessing the former, based on the tone of all the rhetoric, but if it were the latter that would at least (sort of) fit better with common preceptions of US governance. I'd still find it very weird, though.
I think the perception of the US not regulating stuff much is mostly wrong (especially financial stuff). Specifically on gambling and securities there's some pretty strict regulations.
Well, the USA is the country that did the Prohibition. It was founded by Puritans, so whilst it may be more capitalist than most countries in general, when it comes to the regulation of moral vices ....
Left vs right wing morality is just a confusing concept here, because "right wing" is usually meant to mean conservative in the US context and the US is historically very religious, but left wing morality and conventional Christian morality are just two sides of the same coin. The actual opposite to left wing socialism would be something like libertarianism or anarchism, neither of which are big on regulation of gambling.
Left or right is hard to define, particularly if you're defining "left" as "socialistic" and right as "moralistic".
The historical core of opposition to gambling in the US is religious, rooted in the teachings of the churches most affected by the early 20th Century "Social Gospel" movement of American Protestantism (the United Methodist Church still officially denounces all gambling to this day).
And the "Social Gospel" movement was the religious wing of the early 20th Century Progressive movement in the US. (Other causes of the "Social Gospel" movement included alcohol prohibition, the elimination of child labor, the living wage, and political corruption.)
Assuming that the US is just “less regulation” is just as bad as an assumption as assuming that Western Europe is just “more left”. These might be useful shorthands for certain central policy debates, but not across the board.
Historically there may have been hostility towards gambling, but I dont think thats been true for quite some time. It also depends on what type of gambling. Horse and Dog racing have long been legal in many many states. Jai alai is a game popular in miami that has long had gambling associated with it.
For long periods only Nevada and New Jersey (and some river boat casinos which i am not very familiar with) allowed gambling such as cards or dice or slots, but that has changed a lot. Native American reservations have long been able to open their own casinos. And some states started allowing horse and dog tracks to open casino related operations.
The big changes recently have been to allow sports betting in more places as well as remove requirements for animal racing to be attached to casino operations. Now many, if not most, states allow sports betting.
Changes to gambling regulations over the past ~15 years have been similar to the changes to marijuana regulations over the same period. States saw how much money they could make by allowing it but heavily taxing it. So more and more states are removing regulations. There has been some opposition from religious groups or other moral objection; and at first New Jersey and Nevada based casino operators and the states of New Jersey and Nevada tried to prevent other states from allowing gamblings for fear their monopolies would be lost. The casino companies just adapted while Las Vegas became more about entertainment than just gambling (new jersey had been on a long slow decline for a while already).
I suspect the difference in attitudes to gambling between Australia and the US might mostly come down to the roots of Australia's political culture being way more Catholic compared to the mostly Protestant US political class of the past.
Thanks for linking to https://fatebook.io! Interested to hear people's impressions
Also - the "who says we aren’t using it" anecdote about using Metaculus in finance is surprising, I'm curious whether this extends beyond the intersection of ACX readers and finance people.
Does anyone know of a similar site or tool for implementing group prediction markets (multiple people can make predictions, aggregate score, etc). I’d like to implement it at my work, but wouldn’t be able to use a public market like manifold.
On Fatebook you can share questions with a list of email addresses who can all make predictions. I'll probably add forecast aggregation soon, maybe today!
Fatebook is for registering predictions (e.g. 70%), it's not a prediction market (e.g. "Buy YES for $0.58"). I think this works better for very low numbers of participants and is more intuitive, but you miss out on nice properties of prediction markets.
On Manifold you can now create private groups, which might fit your usecase?
"the 'who says we aren’t using it' anecdote about using Metaculus in finance is surprising"
Why is that surprising? Finance is full of number-heads (aside from those who are farther from the numbers and make money primarily due to being born rich and/or schmoozing). They're exactly the types I would expect to read this blog and/or be interested in prediction markets.
Also, the incentives for them to use every source of information possible to determine the right financial bet to make are...huge. That's the whole deal with the stock market. So it doesn't surprise me at all, tbh. The only reason I can think of for this being surprising to someone is if they didn't think predictions markets were a source of meaningful information distinguishable from noise. Which, if that's your opinion, fair, though I don't agree.
Surprising only because I'd spoken to someone who'd spent some time pitching superforecasting-style forecasting consulting work to finance (and other) companies and had concluded that they largely weren't interested. It might be that I'd over-updated on this and a couple of other anecdotes.
Depends on how they pitched it too, I imagine. "This is the best way to figure out the best way" - yeah, nice, see you later. "This will make you gazillions and put you ahead of the competition!" - tell me more!
Yeah, in house proprietary software for...prediction markets? What super secret, hard to explain design elements would that have? My best guess is that some people misinterpreted the intent of Scott's question and are referring to tools that scrape Twitter/other sources of big data for consumer sentiment to predict earnings.
Oh, sweet lord. I don't know this Sinclair Chen, but good grief: we have no god but Mammon?
"Capital markets today serve a function not just of allocating investments, but also of aggregating information. They incentivize smart ambitious professionals to work hard, grinding 60 hour weeks in their glass towers, striving be the best predictors of the value of assets - and yet, any American from any background, whether black or white, straight or gay, christian or atheist, female or male, can beat the market so long as they see or infer something that the NY suits don't see. And by beating the market they improve it. It is the purest meritocracy and the purest truth-finding machine and there's nothing else quite like it in the world."
All hail the blessed and divine market! Let none gainsay the market, for in Mammon our pelf there is neither gay nor straight, black nor white, or male nor female! There is nothing like it in this world for it is not of this world but comes from the one true holy source of all motivation - getting stinking rich*!
No wonder they go for "you have blood on your hands" over the top J'Accuse! How many ordinary Americans read the stock market or base their behaviour on "Ah, I see Disney stock is down"? Yet we are supposed to give full credence to "a bunch of math nerds playing with a toy market would be followed, understood, and believed by the man in the street, who would then have changed his behaviour about masks and Covid"?
Lady, I wore a mask. I got vaccinated. I isolated during the lock-down. I still got Covid, and no bunch of smart maths predictions would have prevented that.
*If those smart ambitious professionals got paid in "here's an apple and a pat on the head", how many grinding 60 hour work weeks for the pure truth-seeking sake of it would they put in?
I'm reminded of this excerpt from "A Cynic's Guide To Fintech", particularly the first few sentences of the second paragraph:
"Fintech business model #5. Assuming that the regulators will be more inclined to listen to your whining than to the incumbents'
"Usually a bad idea in financial services. Regulators basically don't like small financial services companies. There are severe diseconomies of small scale in supervising them, they are more prone to blowing up and they don't do very much for your career. And financial services is an intrinsically regulated industry where consumer protection is often very rigorous for a good reason. So the whole Uber idea of just blatantly breaking the law and then sending out a press release about how uncool and obstructive everyone is being is not going to go down well. Several fintech startups have already found out that there is no exemption for tech companies from the money-laundering or consumer finance laws, and that regulators usually don't care if they've driven someone they regard as a rule-breaker out of business. The existence of financial regulations also tends to mean that fintech companies need to have a lot more capital lying around than they would if they didn't need a financial services licence."
To be honest the superconductor thing is a big embarrassment for the prediction markets. I am a physicist, and among people who are experts in superconductivity (myself included) I can't find anyone who is remotely optimistic. The data looks way too fishy and does not really resemble the usual data for confirmed superconductors, it has logical inconsistencies, the authors are clearly ignorant of relatively basic concepts in superconductivity, and this report is part of a long history of dozens or perhaps 100 "unidentified superconducting objects". If this forecast had been made only by experts in superconductivity there's no way it would have gone above 1-2%. To see the prediction markets at 50% makes me inclined to dismiss the whole approach as anything that can remotely be described as "predictive".
The bar of “a successful replication” is fairly low though. If this thing is really just a powerful diamagnet, then some group will replicate that and get the same misinterpretation, and it will be “a successful replication”.
Disagree: the original data seem inconsistent with just a powerful diamagnet, so if somebody showed that it was a powerful diamagnet, it would not be a successful replication.
(Not intended as a gotcha, just sharing useful additional information from elsewhere in the thread)
FWIW, given the openness of the original researchers to offer their original material for others to test independently, I think intentional hoax/fraud is unlikely, and most likely it’s just researchers making mistakes.
I really would like an "Idiot's Guide" to what is going on right now, because my initial reaction to this was "oh yeah, yet another hype that is going to explode when poked", but I'm seeing claims that the results have been replicated elsewhere, so I don't know what to think.
I'm still very dubious, but it looks a bit like there *might* be something in the theory there? Help?
It has been replicated. The theory behind why it works has been simulated. It's basically an engineering problem in the developed world at this point. Maybe the US will crank out a Nature article by the end of the year but really who cares about them. The "Rationalists" as per yoozh are behind the curve and have a Western-centric conservatively-slow-to-update bias. They, as a whole, don't even accept UAPs despite the vast evidence in favor of their existence. The less time you spend in Flat Earth level communities like this one the better.
(1) Allegedly replicated on somebody's kitchen table in Russia and via a dodgy video from China. I want something a teensy bit more rigorous than that
(2) Dunno about "Rationalists", being slow to accept extraordinary claims without good supporting evidence just seems sensible. We've been at this rodeo before, remember?
(3) I don't belive in UFOs, even if they have been rebranded to UAPs. Though according to the guy telling all and sundry that they're real and here, the Vatican knows about them too. So until Pope Francis issues some kind of pronouncement, yeah I remain to be convinced.
> Although I like the result of this decision, I’m worried about the ruling that no-action letters constitute binding commitments whose amendment or cancellation requires careful agency action with every t crossed and every i dotted.
What is it that you like about the decision? It seems to me that the bad thing you describe in the second clause here is the primary result of this decision, so I don't know what there is to like about it.
The LK-99 market on Manifold is going absolutely crazy right now. It's a weird dynamic, where hundreds of newly registered users are throwing in 500 mana on yes (that's the initial play money allocation), sometimes even buying mana for this purpose, while established users already feel overleveraged and don't have the liquidity to correct the market.
"My my, whoever would have thought that the prospect of EASY MONEY even if it's only play money would attract hordes of treasure seekers and the SENSIBLE RATIONAL WISE ONES would be overwhelmed so that the market was not, in fact, self-correcting", she said drily.
Replying only to say that this -- "although 180 of those were by a person named Chris Greenwood who somehow submitted the same message 180 times. This is probably a metaphor for something" -- made me LOL.
> A no-action letter is a letter written by the staff members of a government agency, requested by an entity subject to regulation by that agency, indicating that the staff will not recommend that the agency take legal action against the entity, should the entity engage in a course of action proposed by the entity through its request for a no-action letter.
This makes a no-action letter sound like something that should exist (i.e. you ask an agency if something is legal, rather than guessing and fighting a long legal battle later) but if I remember, this case makes it seem like they don't actually do that? You wrote last year (https://astralcodexten.substack.com/p/mantic-monday-81522?):
> On August 4, the CFTC reversed itself, saying the PredictIt had “not operated its market in compliance with the terms of the letter” and that it had to shut down by February
And then wrote that it wasn't clear why this reversal had happened since they stuck to the stated rules, but that it seemed like some sort of regulatory capture happened, or the agency just changed its mind. If these things can happen, then what's the point of such a letter? It doesn't actually reduce uncertainty. Rather than engaging in legal bone-reading around what laws say and what (if any) precedent exists, now companies have to engage in psychological bone-reading around how binding the commitment is. Making them binding is the only way for such a letter to be of any use.
This seems to confirm that contra what Scott says about "Allowing election bets makes Wall Street less interested in elections, not more", prediction markets can indeed be rigged.
Maybe you *want* insider trading for prediction markets, because that's based on insider info because the Platonic Ideal of a prediction market is laid out in that piece:
"Prediction markets allow people who lack expertise about certain topics to form precise models of what the future could look like thanks to the live-updating probabilities that are generated by traders."
But uh, in that case you could just go directly to the experts and people in the know and ask them "are the aliens in contact with the Vatican little green men or Greys?" instead of going the extra step with a betting market.
Insider trading is about making a profit, not adjusting the market, and the story shows this is what people were doing, and that despite all the lofty rhetoric, if A puts a ton of money on "YES" then people will be reluctant to buy "NO" even if they know better, or think they know better. So it is demonstrably possible to influence public sentiment and change opinion. The big Wall Street firms might be very happy to dump a million or so into prediction markets, even if they lose it, because it shifts public sentiment towards Bob instead of Billy, and they want Bob because he is not going to Sarbanes-Oxley them if he wins the election.
"Users started commenting on the market, with weak speculation that there could be insider trading. But because this was Quinesweeper’s first market, others feared he was planning to load up on YES shares and fraudulently misresolve the market to yes regardless of the outcome. The Manifold team assured users that we would fix the resolution if it was resolved incorrectly so that users could trade on their true beliefs, and not have to factor into their probability the chance of fraud."
But they don't seem to have been able to control the insider trading; they didn't want to delete the market, and the efforts to bury it only made it *more* conspicuous and got *more* people interested and jumping aboard.
"A lesson I would recommend learning is that many AI markets on Manifold are heavily insider traded"
I would imagine if prediction markets do get off the ground/are allowed use real money, this will spread to other markets. When there is the chance of money being made, there is always the incentive to manipulate or rig the market in your favour, even if it is 'only' using insider info.
"Firstuserhere has also informed me that a few YES insiders agreed to not buy the market above 55% so they could conceal their continued accumulation of cheap shares. They really were working together to maximise how much profit they could make off of the ignorant NO bettors."
"From what I can tell, turning the market into one big meme worked surprisingly well, and some users genuinely believed there was no statement, and that the only reason the probability was so high was “for the memes”.
And these are prediction market users who presumably have a good idea of what way it all works, and yet they could be fooled into thinking the entire market was a fake, or that some outcome other than the reality was true.
I don't think this shows as hopeful a way forward for prediction markets as the author thinks; forget predicting AI risk, work on "if we do manage to get prediction markets mainstreamed and if we do get permission to use real money, how do we avoid manipulation, scams, and profiteering if what we really want is to get the best information, not just make money?"
I think the biggest problem with Scott and the other pro-prediction market and pro-crypto people is that they think that markets just magically work in a state of nature. The reason that our real-money markets work so well is *because* of the regulation from SEC and friends, after a century of lessons won the hard way. But then when the regulators say "hold on a moment, this seems prone to manipulation and insider trading", they scream that the government is holding back innovation.
Obviously, not *everything* the SEC does is beneficial or necessary, but on a scale from "no SEC" to "status quo", I'd guess that it is at least 95% necessary. We've seen what happens without regulation from all the financial panics and scams of 100 years ago, and then the crypto world helpfully speed-ran the entire history of scams and fraud in a decade for the benefit of anyone who forgot. And now we have demonstrations like this.
>We've seen what happens without regulation from all the financial panics and scams of 100 years ago
I can't remember the exact details but I read a novel from the late 19th century which involved, amongst other things, this kind of financial scams - not quite crimes because, as you say, regulations hadn't been imposed then. It involved the representative of a large American financial combine who worked in London and was making trades on the stock market; using such advanced technology as THE TELEGRAPH he was able to get early information from America about shares and buy cheap to sell high. Fair enough, sez you, insider trading maybe but not illegal at the time.
But what he *also* did, on instructions from the combine, was manipulate share prices so they could pluck pigeons. He might buy up a lot of shares in a particular company, and either he'd 'let slip' to selected investors that this was a Good Thing and he was letting them in on the ground floor early before it became public knowledge, or others would see this and think "Hmm, if the XYZ Combine is buying this, it must be about to boom" and the pigeons would put their shirts on buying shares, thus driviing the price up.
At which point the combine would sell up, the price would crash, and the pigeons would be ruined, while the combine either knew the company was going to collapse and used this way to maximise their returns, or they could then buy back the shares cheaply in the company which was going to grow naturally. No regulations, no regulatory bodies around to stop this kind of sharp practice.
The main plot was about a guy who invested this way, was ruined, and his scheme to get his revenge (and his money back). The difference between Then and Now is also shown in the Sherlock Holmes story "The Stockbroker's Clerk" (where between you and me, I often wonder exactly how stupid was the client? But then again, the con artist tricking him knew exactly what to say to a young, desperate, and not too bright guy to push all his buttons):
“Well, the whole thing hinges upon two points. The first is the making of Pycroft write a declaration by which he entered the service of this preposterous company. Do you not see how very suggestive that is?”
“I am afraid I miss the point.”
“Well, why did they want him to do it? Not as a business matter, for these arrangements are usually verbal, and there was no earthly business reason why this should be an exception. Don’t you see, my young friend, that they were very anxious to obtain a specimen of your handwriting, and had no other way of doing it?”
Nowadays we don't let "business matters" be "usually verbal" for the reason that we no longer do things on a handshake and 'a gentleman's word is his bond'.
The BBC broadcast a radio adaptation of the 60 canon stories in the 90s and here is a link to the shows, including The Stockbroker's Clerk:
I really like what Bert Coules did with The Lion's Mane; he took the bare bones of the plot and did a completely different story around it, with bonus "The Man from U.N.C.L.E." and Star Trek references if you can catch them!
The problem isn't that regulation can't fix an issue. The problem is when you constantly complain about said regulation, *including in this very post*.
I think there should be low enough regulation to allow things to exist, but high enough to allow them to be good. I don't know exactly what that level is, but I think a good first step would be allowing it to exist, then seeing whether it is good or not and what steps would be necessary to make it so.
I do hope at least some regulation can be at the market level - ie some market decides insider trading is bad and it will ban anyone who commits it. Compare to eg casinos regulating cheating. That way we can have more flexibility and diversity in level of different regulations.
Well, just right there - the market exists, and we've seen someone (a couple of someones, indeed) hopping in to Win The Game. I think we have a good general idea of how human nature is going to work out, even with the nice civil rationalists (and that last is not a sneer, you lot generally do try hard to be good people). But even good people succumb to temptation.
Yes, prediction markets would be unnecessary if every human could identify the people who know the true information, go to them, and trust them to give a true answer. But we can't.
It looks like this prediction market ended somewhere between the accuracy level it would have had if there had been no insider trading, and the accuracy level it would have had if every insider had disclosed all the knowledge they had publicly. I think this is good for prediction markets (though not for AI orgs who want to keep their information secret)
yeah, but you had people using inside information to make money (even if it was only play money) so I don't think that prediction markets will be free of the problems that plague ordinary markets unless and until the prediction market bods lay down the law with some heavy regulation.
And the piece says that Metaculus was reluctant to do anything much in the name of transparency and infosec, so the market-riggers were enabled to go right ahead.
Imagine this was a wider world, public, market with real money and some immediately vital topic (like an election). I have no confidence the same wouldn't happen.
The question: "Will Israel pass the judicial reform bill?" is not very well formed, since "Israel" is a diffuse entity. The parliament (Knesset) might pass the bills, but the Supreme Court (Bagatz) might rule them unconstitutional, which is weird, since Israel doesn't have a constitution. Then... well, forecast that.
Most hedge funds, index fund managers, and mutual fund providers have theirs filthy green hands in the pockets, arseholes and mouths of both major presidential candidates by Election Day.
Since these organizations are soon going to own all the property in America and make us rent our lives away, I think it’s fair to say no entity can exceed their investment and capacity for over handed and underhanded force should the situation call for it.
Rigging the election might be possible with either large Russian or Chinese intelligence investment, but really only with the explicit compliance of a major town square owner like Musk or Zuck. But why on earth would they stick their neck out for a state that just experienced a minor coup, and a state with its own private internet where they don’t stand to profit much. Especially not Musk who would lose SpaceX overnight if he was ever discovered to being letting the enemies of the US tamper with ... twit... twi.... X ‘gag’.
Suffice it to say, it would need a trial run with oversight, but it’s prob fine to bet on the election. Maybe a little funny too since both parties seem to be competing to see who can jam their own foot in their mouth harder.
Would a rationalist content search engine be something you, reader, would pay a small amount for? Say $2-6/month. Imagine it has all the features you want.
>"Why would any government agency ever give a no-action letter now?"
Cynically, perhaps because it'll be binding(ish) on your successors who might otherwise be inclined to more energetic enforcement than you would prefer?
Just came across this interesting article profiling several highly successful traders on PredictIt. It's amazing how much trouble people will go to to win money on PredictIt.
User banned for this comment.
Isn’t the answer to the last question just that most journalists don’t know what forecasting engines are or think they are valuable? And even for the few who do, they know their audience doesn’t know or care?
I daresay that is a big part of it. I also get the impression that typically journalists write news articles (as opposed to analysis or commentary) on the basis of interviews done with people or reports written at institutions. So if prediction markets are important it would be for the people being interviewed to bring them up.
I have seen prediction markets and betting odds discussed in articles about elections and the like, already quite a long time ago, so it does happen.
Yes, the media like Salon used to cite the Iowa Electronic Markets and then Intrade all the time. This was during the big wave of interest in election polling & forecasting that more or less culminated in Nate Silver. I am not sure why it all disappeared, although Intrade shutting down due to some sort of embezzlement issue certainly didn't help. I suspect that the culture-war turn didn't help matters.
I think you nailed it. Journalists write articles to get reads, and if consumers are more likely to read an article because it relies on them because it's part of the current zeitgeist they'll include them. But once it leaves the zeitgeist journalists don't really have any reason to care about a prediction market.
Yes, exactly, from my experience as a journo (not a generalist one, but still), the second bit would be one of the main reason: in the context of a word-limited piece (as most are), you don't want to have to explain (i) what it is; and (ii) why it's worth taking into account. This might change when prediction markets become common knowledge.
Add to this that citing sources or external info is not something you like to do in practice. Sure, if you are being asked, you should be able to back all statements. But when drafting, you get the urge to make it as though you had the knowledge all along, so as to appear smart, but also ward off competitors.
In this context, citing people/experts on a given matter is much easier, partly because it's also a way to build up a relationship with that person/expert to obtain more info down the line. You don't get that added value with public-facing websites or sources.
As a (admittedly noncentral) reader, I hate the norm of not citing sources. I think ~all numerical claims should come with citations. I think a writer comes off as smarter and having better epistemics when they cite sources.
My strong suspicion is it's due to the Silicon Valley-New York (or tech-media more broadly) interelite conflict, and the journalists are obviously on the side of media. But I'm not sure how you would falsify that hypothesis.
I think this is a huge part of it.
But if that were the complete answer, sports journalists would include odds from bookmakers and betting sites in their stories far more often than they do. After all, those are fairly well known and understood by ordinary people, and are easily accessible.
I think another reason is that there are two sides to journalism: Reporting and commentary.
Reporting is to check facts and report on what has happened. True reporters love numbers and hard data, but focus on the past and undeniable, and avoid anything that looks like speculating about future outcomes. (Pure reporters are very old-fashioned, and rare in our age.)
But commentary has become far more common, and most journalists now see their job more as something like "telling important stories" rather than reporting. Of course, good journalists can do both – that's what they try to teach you at J-school (at least when I went, a good while ago) – but most journalists aren't good.
Mediocre journalists don't understand or like numbers, but perform their job work by gathering gossip and anecdote, and compiling BS that aligns with their biases. They prefer salient quotes and soundbites to impersonal data and bright lines for lots of reasons: Numbers don't evoke enough emotion (show, don't tell, etc.). Cold facts rarely line up neatly enough, but complicate the narrative. Also, they are hedging their bets by being "nuanced" (read: vague).
Most journalism is a mix of these two, but whereas pure commentary has become very common, pure reporting has become rare and old-fashioned.
One factor, IMO, is that a forecasting engine that worked would work against the actual business of journalism -- producing an endless string of horse-race stories about something. The last thing a journalist wants to write is that it's clear who is going to win an upcoming election because that means they have to find something else to write about tomorrow.
A particularly annoying case was the Fetterman vs. Oz election, which IIRC was the last undecided seat in the US Senate in 2022 and thus attracted a lot of attention. The web site Five Thirty Eight originally made its name by predicting elections accurately, but in this case they reported daily on the election as it was coming up detailing which factors favored each candidate, as if there was a reasonable chance either could win. In fact, Ladbrokes was offering 10-to-1 odds that Fetterman would win. So a maximally factual article would say, It would take a miracle for Oz to win and not bother reporting on it until election day.
The rationalist fascination with Aella is bizarre to me. I really don't have a better explanation than 'nerds enamored with only cute girl to pay them any attention'. She gets treated like she's some sort of intellectual, but watching the Lex Fridman podcast episode that featured her, it's very clear that she's...not. Does she have any kind of impressive prediction record that I'm ignorant of or something?
I find her blog and survey pretty interesting - see https://aella.substack.com/ . As far as I know she doesn't have any amazing formal prediction record, but neither do I or lots of other rationalists. She just writes well, thinks clearly, and has good opinions.
She's a replacement-level controversial blogger (who already often get some following) who's also if roughly the 90-95th percentile of both "attractive woman" and "willing to get naked online". This makes her more famous both for the direct effect (some guys will follow her for that) and the schelling point effect - if you've got a dozen similar mid-decent Hanson types but want to coordinate on everyone following the same one, might as well focus on the one who also stands out for an unrelated reason.
This assumes there are many good bloggers and we need to find some reason to prefer one over another. This whole conversation is predicated on "she's just a person who writes well and consistently, why do we care about her so much?"
At the risk of saying self-interested arrogant things: basically nobody writes well and consistently. I write moderately well and consistently, and make more money as a blogger than I ever did practicing medicine. Writing well and consistently is so rare that it it fully explains all of her fame and then some.
I don't think we have "a few dozen mid Hanson types". We have Hanson himself - who has 80,000 followers on Twitter - Aella - who has 176,000 - and a handful of others who are also popular. If there were other people as good as those people, they would also have tens of thousands of Twitter followers and be famous.
I actually sometimes occasionally read Aella for fun when I'm bored. This is a really high bar; anyone who reaches it will be famous no matter how pretty they are or aren't.
To put my original claim a different way*: I think aella is about, say, a +2sd blogger in terms of writing skills/edginess. It's a nontrivial and fairly impressive level of skill, but not a massive outlier, and is thus insufficient to explain her fame (which is +3-4sd). However, being +2sd on two other things that are mostly uncorrelated (but help with fame) combines to make her a significant enough outlier to get this level of attention.
*I'm tempted to get into a long aside about how to estimate the exact number of similar-skill bloggers and the ability distribution, but while I think we'd still have different opinions on the exact distribution and her placement in it I think my main claim should survive them.
It often seems the other way around: People assuming Aella is "fascinating" (as you say) to others only because of her attractiveness seem to be the ones who can't decouple her appearance from her other work. Or at least overemphasize it. She's prolific in very different ways that can equally explain the fascination.
That comment would be much more convincing with some examples of these different ways in which she is prolific.
Her kink research, obviously
Willingness to experiment on herself, like taking the poop pill for gut health or talking about the outcomes of extreme no-shower durations
Unerring talent for stumbling onto scissor statements that drive people into a fury
I agree. I have paid virtually no attention to her naked activities and find her totally fixating. She writes in that brilliant rationalist way that emphasizes plain language to avoid your thought process from getting lost in complex language and tripped up by assumptions built into it
This goes with her attractiveness to set up a road block for people who aren't engaged by her work to assume it's all that. I'm currently avoiding her because my inner Chesterton's Fence is chronically annoyed but I can't think of anyone else particularly comparable to the plasticity of her curiosity
Well, a survey with a sample size of thousands has a value in and of itself, I'd say. You really can't discount that.
There's also the fact that she's willing to be un-PC, which obviously a lot of rationalists like.
The fact that she's the sort of girl nerds dream about and rarely get obviously has a lot to do with why she gets those huge sample sizes. But if it lets her do research academia wouldn't dare do, well...I might not think Musk deserves his billions, but I like his work with electric cars, and I want humanity to go to Mars.
IMHO, Aella is curious about interesting stuff, so her writing is often interesting.
(Specifically, I have her in a bucket with Gwern, Hanson and Tabarrok as "usually interesting when I have time to look, but I usually don't.)
> She gets treated like she's some sort of intellectual, but watching the Lex Fridman podcast episode that featured her, it's very clear that she's...not.
I came away with the exact opposite impression. She was having none of Lex's tendency towards imprecise and poetic descriptions/questions, and rightly drilled him on exactly what he was trying to convey.
I'm tickled that Destiny is the kind of guy who gets invited to speak at conferences these days. The last time I remember hearing about him, he was "that Starcraft II streamer who keeps getting in hilarious fights with Deezer and CombatEX over Skype."
Just gotta try hard and believe in yourself, I guess.
It's honestly a more respectable inviation than somebody whose claim to fame is being a sex worker with a twitter account.
Aella's done more genuine data science work than the supermajority of this comment section. (And probably contributed more to the corpus of human knowledge, tbh)
I don’t think her invite is important enough to argue over. She’s a rationalist internet celebrity; that’s all you need to know.
But I do think the idea that she has done “genuine data science” work should be pushed back against. She can’t have done more genuine data science work because she hasn’t done any genuine data science work. Her work is both hopelessly compromised by methodological issues (no controls and poor/non-random sampling) and does not utilize the scientific process (opinion essays that use rationalist rhetoric to present personal experiences). This is not to say that she is incapable of doing some but to this point she has not done any so she cannot possibly have done more than anybody else in this comment section.
She posts informal polls that are entertaining and occasionally suggestive about the opinions of people who pay attention to her, which should be enough to justify further interest anyway if you must justify it.
I just think you're dead wrong about her work having serious methodological issues.
It has issues. But quantity has a quality all its own.
And I'm not sure 'rationalist-leaning internet geek' is all that much more unrepresentative than 'college student who has time on their hands'.
I would not be surprised if many of her critics would have equally skeptical stance to studies of college students who walked through the wrong door in a library or internet surveys.
In many ways, humans are quite similar to each other, so it is easy to make case for it when studying mechanisms that are shared by all humans. As a point on the critics' favor, the social and psychological phenomena, everything that can concerns the mind and behavior in fact is the dimension where the humankind shows most variety. (To educated layman, like me, I should add). And the internet has the weirdest niches of it all: easily available, easy to self-sort into bubble networks. It makes so much intuitive sense why any finding with such sample would fail to replicate in some other population due to some bizarre selection effects. But on the other hand, sometimes intuitive common sense is wrong. Though I have no expertise in psych-related, from what I understood, the replication crisis didn't mean all such studies failed to replicate.
Tangential thoughts:
+ Any sample, including the college student / random internet study sample, becomes is more interesting if you can perform randomized controlled experiments. Causal effects are still causal effects, especially if they replicate, even weird causal effects that replicate in weird niche populations.
+ There are many things to dislike about scientific peer review as part of publication "stamp of approval" process, but all comments I ever received were useful. (If the reviewer just didn't understand something or had ultimately inapplicable advice, still thinking it through was useful experience.) I don't know if she has tried to publish, but that would be a way to get feedback (and occasionally, the stamp of approval).
+ Suppose one has a study and set of results. A question of some kind of bias arises. Various, increasingly clever sensitivity analyses are presented to address it. A fact that I find weird is that it is surprisingly rare to see a numerical exercise showcasing exactly how much bias there would be needed to observe any particular result.
https://aella.substack.com/p/my-kink-survey-design-methodology
She does way more than twitter polls. What part of her methodology do you take issue with?
As for non-random samples, there's a dozen reasons why that's not a valid critique (not the least of which is that her samples are already much better than a huge portion of published research). But I'll just link her post on it here:
https://aella.substack.com/p/you-dont-need-a-perfectly-random
Anyways, are YOU a data scientist? Have you been paid to do data science? Because I see multiple actual data scientists telling her that what she does qualifies.
I will fully admit to having a higher standard for rigor than many so-called scientists of different stripes. I don’t think I’ll be lowering that standard in the name of inclusivity.
I am not a data scientist, though my job involves certain elements of the prototypical skill set. It would be bizarre if the world of ‘data science’ was so subjective that it could only be evaluated from within, as if it was the qualia of an acid trip.
You were asked a specific question and I'm still curious about it: which part of her methodology do you take issue with?
>It would be bizarre if the world of ‘data science’ was so subjective that it could only be evaluated from within
No, but presumably people with the title should have more say in the boundaries of the category then someone who is not a data scientist. The criteria you've implied to fail her for would also exclude many, if not most, data scientists.
But this is all moot, since the criteria you initially gave makes it seem like you either aren't actually familiar with her actual survey/analysis work (since you claim it's all personal experiences, which is false, and lacks controls, which is either false or not applicable) and are refusing to update on the information you've been provided with, or are arguing in bad faith by using an esoteric definition of "data science" and refusing to provide that definition.
"Make your claims clear, explicit, and falsifiable, or explicitly acknowledge that you aren't doing so (or can't)."
Doubt. If you say the median instead though, maybe.
He was the first (and so far only) streamer to be interested in prediction markets, his fans got into it, they created a lot of Manifold markets on who he would get in beefs with or whatever, and now I guess he is part of history.
Other big political streamers are commies allergic to anything market-related. Outright rightist can't get big in the current climate, and Destiny is a weird contrarian always balancing on the verge of being completely cancelled, but until that happens he's the best we're going to get.
I'm curious what it is about this format that a rightist can't get big in it. I'm not familiar with political streamers, really, and hadn't heard of Destiny. Maybe I'm too old for this stuff? (Older Millennial). Or maybe I'm just allergic to spending my time listening to unscripted ranting, which is what this sounds like to me.
Following the pattern of other formats, you would normally think that even if the market is dominated by leftists, there would still be room for one rightist to grab some market share by sole virtue of being the ONLY rightist voice.
Well, the way that people get big in streaming nowadays is through networking and collaborations, with all the attendant echo chamber effects. If some lone rightist threatens to get big in spite of that, it's easy to find an excuse to ban him, because "trust and safety" structures are also staffed with leftist activists.
The main streamers and their fans are kids or recently kids. Kids are generally very lefty since they haven't had a real job yet.
Speaking as someone who had a real job for decades, that describes a lot of rightists. I remember when Ayn Rand was the "in" thing. Generally among people with no experience.
Kids tend to get swayed in various different ways. One of the characteristics is that the way they are swayed is ... unpleasant ... to their respected elders.
Yes, I was going to say, Martin Blank's comment doesn't ring entirely true to me. As you point out, Ayn Rand is someone who appeals disproportionately to the young. As does, to a lesser degree, libertarianism more broadly. I thought it was the young that catapulted figures like Jordan Peterson and, dare I say, Andrew Tate.
I'd say one of the most salient and enduring political tendencies of intellectually-inclined young people is a preference for unnuanced idealism over a more pragmatic or nuanced politics. Marxism is still a leading brand for unnuanced idealism, but there are others.
This was my thought as well, I keep seeing him pop up in rat-adjacent spaces and assuming it can't be the slightly-edgy former SC streamer, but it is.
The Economist now regularly cites Metaculus in exactly that way!
Can you share some links? I haven't seen it, but I don't read it religiously
I’ve just noticed it popping up in articles recently. I’ll see if I can find some examples.
Here's an example: https://www.economist.com/finance-and-economics/2023/05/23/what-would-humans-do-in-a-world-of-super-ai
Key quote: "Last summer forecasters on Metaculus, an online prediction platform that is a favourite of many techies, thought it would take until the early 2040s to produce an ai capable of tricking humans into thinking that it was human after a two-hour chat, had good enough robotic capabilities to assemble a model car and could pass various other challenging cognitive tests. After a year of astonishing ai breakthroughs, Metaculus forecasters now think that this will happen by the early 2030s."
The Economist first cited Metaculus on January 1 2021 in an article headlined, "What To Expect In 2021 According To Prediction Markets".
Thank you! Though I wouldn't say that's the Economist citing it in exactly that way, even though it's on the way to it. First, it has to introduce what Metaculus is (of course, at this point a necessity), which it others with the "that is a favourite of many techies", biasing those against tech against it more than they already would be. And, the "AI Astrology" in the graph really rubs me the wrong way.
I wish metaculus had a more general LK-99 replication question. It seems silly and short-sighted to restrict to only the first replication attempt, knowing how slow and difficult the process can be.
re: "why would any agency give a no-action letter now". Because it forces the targets to engage in expensive, time-consuming legal and regulatory activities to fight back, and a lot of times targets won't or can't. Then the precedent gets established for free. Basically it's an end-around the regulatory process. The PredictIt case isn't exceptional. Agencies often do this kind of thing (they all have their own mechanisms) before initiating regulatory proceedings, just to see if it works. It costs them very little, especially compared to the normal regulatory process. Worst case, a judge tells them to do it better, so they do. Worst-worst case, they step back and follow the process they should have followed to begin with.
Obligatory not a lawyer, but I have spent a lot of time working with regulatory data and processes. Things like no-action letters are already considered legally binding commitments to a degree (I don't think this is a novel ruling), and agencies are held to something resembling normal standards of justification. Biggest difference is that they don't need to go through the public comment process this way, which is tedious and annoying and which they mostly hate. Because, as you note, a bunch of wingnuts spam them and they have to (are supposed to, at least) prove they've responded to every comment.
"Because it forces the targets to engage in expensive, time-consuming legal and regulatory activities to fight back, and a lot of times targets won't or can't"
I'm confused - I thought a no-action letter was when an agency said something was okay. Who is fighting back? Can't they just never tell anyone that what they're doing is okay, and just leave things up in the air?
It's kind of weird that they ever decided on an official policy of selective enforcement, right? Like a no action letter is basically saying "we have a rule against this, but you guys seem pretty nice and harmless, so I guess we won't shut you down right now. Maybe later," and then if you decide to shut them down later they can sue you because... what were you thinking. The whole idea seems arbitrary and capricious.
https://en.wikipedia.org/wiki/No-action_letter
Wiki says a no-action letter is given when an behaviour is technically, legally NOT ok but the agency won't do anything right now because common sense.
The wiki article is rather bare bones and has a [citation needed] for the above claim, so there's that.
Also:
https://www.investor.gov/introduction-investing/investing-basics/glossary/no-action-letters
"In addition, the SEC staff reserves the right to change the positions reflected in prior no-action letters."
So I suppose "fighting back" against a no-action letter doesn't make much sense except if you want a clear-cut, legally binding approval of your proposed actions. Then again, why would you request a no-action letter in the first place.
You think there are "lizardman" problems with play-money prediction markets?
He's joking
I don't know, with no money on the line I could definitely see people making wacky answers just for fun (which is what 'lizardman' originally refers to, 7% of Americans answering 'yes' to 'lizardmen controlling the government' because they think it's funny).
The caveat is that silly answers will probably reduce the amount of play-money they have, reducing their impact, unless they spend real money to increase it again.
This works if the number of new accounts increases slower than the rate at which they lose play money by betting on silly things. It would be nice to model that
I would just like to formally register that there is rather little steel in either the 757 (American Airlines Flight 77 and United Airlines Flight 93) or 767 (AA Flight 11 and UA Flight 175).
Oh, so now you want a mystical prophet who is also an aeronautical engineer? You will be insisting that they make clear and unambiguous predictions next, I suppose?
I kind of thought that, had missiles struck the US somewhere, those would be considered the steel birds. Or maybe if a crowd was attacked by a bunch of drones. Or a big plane crashed accidentally where a lot of people were.
Looking at it, I considered the prophecy too vague to be dis-proven.
Yeah, I was thinking it's a poor psychic that says "steel birds" when it's gonna be "aluminum birds".
Or even just "metal birds". Reduced precision is better than provable inaccuracy.
Amazing that LK-99 has 25~30% probability. When I first read it in Korean news I obviously thought it would be some ridiculous scam. Then I read that superconductivity related stocks are soaring. And now the prediction markets. If this resolves true, I would be seriously impressed, in both ways.
Also, last week I commented that X links cannot be viewed by unregistered people - looks like that changed. Not sure how long this will last though.
I still think it is a scam/mistake, but the soaring stocks and prediction markets are exactly what I would expect, and why I don't think prediction markets would/will be a great way to make decisions on Issues Of Note. 'If this is true we'll be rolling in the dough' is enough to make people rush out and buy stocks in the hopes of money money money, and then prediction markets see the stocks rising and go likewise because surely if all this money is being poured out, there must be a basis for the optimism! As if bubbles never happen.
True, but there are also people doing the exact opposite - https://manifold.markets/QuantumObserver/will-the-lk99-room-temp-ambient-pre#pzRcUUcSbNZoM5kH805s
There's more than one group. One of the groups is almost certainly either a fake or a useless product. (Needs extremely high pressure, etc. Specs to reproduce it aren't clear, etc.)
The other group is claiming essentially STP for the final product. (Actually boiling water is supposed to still be stable.) And they also provided enough info to allow others to reasonably try to replicate it. That's the one that's 25-30% probable.
It's just happenstance that both groups made announcements at about the same time.
I did not know that - but no, the one I initially read about is the second one. Thanks though.
The STP one has a strong Pons & Fleischmann flavour to it.
Speaking of questions, I'd like to pose something of a metaquestion to everyone here:
Of the following two questions, which do you think is more important?
A) Whether Trump's legal troubles prevent him from running or serving as president again or not?
or
B) Whether the "LK-99" superconductor really pans out as an ambient pressure, room temperature superconductor or not?
Which do you think is more important? :-)
My view is (B). Technology over politics.
B, but if as seems likely it turns out to be yet another "oops, sorry" dud, then A because Trump running or not for the next presidential election will affect the USA,. the lives of those in the USA, and by the ripple effect of influence the rest of us.
That's fair, Many Thanks!
The Trump thing has already mostly done its damage by setting a precedent that hounding high-profile opposition by endless lawsuits is now okay.
How do you avoid that, without setting a precedent that high-profile political candidates can commit crimes with impunity?
In Italy we generally accept the second one :-D
I would think that if the precedent set is that you can hound high profile opposition by endless lawsuits then you would have seen a lot against Joe Biden. Instead the precedent seems to be that you can hound high profile criminals by endless lawsuits, which is why you see a lot against Hunter Biden.
> Instead the precedent seems to be that you can hound high profile criminals by endless lawsuits, which is why you see a lot against Hunter Biden.
No. Comparing Trump with Hunter Biden is absurd. The precedent is if you commit endless crimes, you will face endless lawsuits.
https://edition.cnn.com/interactive/2023/06/politics/annotated-trump-indictment-dg/
Hunter Biden doesn't have 40 counts of charges against him unlike trump.
Right - but he does have some charges against him, unlike Joe Biden. This all seems to me to suggest that, while politically prominent people might be slightly more or less likely to be prosecuted for some things than other random people, there doesn't seem to be open license to persecute any politician one wants, and there also doesn't seem to be total impunity.
Trump got his hush money payment sliced and packed into 26(?) different charges.
Hunter got his three explicit Federal gun felonies consolidated into one misdemeanor. And I have no idea how many state/local gun crimes weren't charged at all.
> Trump got his hush money payment sliced and packed into 26(?) different charges.
If the charges are not valid then Trump can challenge them in court. He has appointed 234 Federal Judges and we all know he is a great judge of character and we can be sure none of those judges are part of deep state.
Besides Trump is not a stranger to courts, Trump & Friends filed around 63 lawsuits challenging various aspects of 2020 election but unfortunately lost in all of them. This must be due to the deep state. Hopefully if this goes upto supreme court they will support him and make him the president. All Hail president Trump.
Pretending that the number of charges is some sort of indicator of relative badness is dishonest in a world where Lavrenti Beria and Tish James exist.
Pretending that you were not using the number of charges as an indicator of relative badness between Trump and Biden is also dishonest.
That's fair, Many Thanks!
> The Trump thing has already mostly done its damage by setting a precedent that hounding high-profile opposition by endless lawsuits is now okay.
What about precedent set by Trump that you can try to overthrow the government or you can steal top secret documents?
Please stop spreading bullshit, Trump committed crimes and he is facing the consequences.
The document precedent was already established by Hillary Clinton.
Yeah. The document precedent was already set by Hillary Clinton, she didn't return the top secret documents when ordered by Court and later tried to delete video footage.
Return what?
She made electronic copies, and kept them on an insecure server where they could have been (and were unless you're pretending the Chinese are incompetent) accessed by anybody. And yes, she NEVER turned over that server.
But yeah, paper copies are magically worse. Go with that.
Hey! Let me ask you a honest question, are you mentally challenged? Cause that's the impression I'm getting from your comments.
> Return what? She made electronic copies, and kept them on an insecure server where they could have been (and were unless you're pretending the Chinese are incompetent) accessed by anybody.
First of all Hillary Clinton used a private email server just like Jared Kushner and Ivanka Trump (https://www.washingtonpost.com/politics/2019/03/21/their-emails-seven-members-trumps-team-have-used-unofficial-communications-tools/). Unlike Hillary who was investigated by FBI and House Oversight Committee Jared and Ivanka were never investigated.
> And yes, she NEVER turned over that server.
Stop spreading bullshit dude. Before you make absurd claims like this at-lease read the Wikipedia https://en.wikipedia.org/wiki/Hillary_Clinton_email_controversy As part of the FBI's Midyear investigation (code name "Midyear Exam"),[141] at the request of the IC inspector general, Clinton agreed to turn over her email server to the U.S. Department of Justice, as well as thumb drives containing copies of her work-related emails. If you are mentally challenged and can't comprehend Wikipedia articles maybe you learn how to read and then you can think about commenting in astral codex.
> And yes, she NEVER turned over that server.
Trump didn't return the documents when ordered by court and tried to delete the video recordings. And being incompetent he failed to delete the video recordings.
Re journalists, I think it's pretty simple. English/history kids who are afraid of math and numbers become journalists who are afraid of AI taking our jobs, and numbers. The intersection of people who write articles for mainstream publications and people who read publications like this/know much about prediction markets is vanishingly small. Not to mention, there's a bandwagon effect: no one wants to be the first one to start using this weird probability thing when no one else is doing it -- especially considering the hate the weatherman, Nate Silver, etc. get for "being wrong". There's little incentive to do it, and plenty not to.
Hmm, I disagree. As a sort-of-journalist-adjacent-person, I'm constantly being asked to have opinions on things outside my area of expertise, and prediction markets are a great way to do that ethically.
A big part of journalism seems to be about trying to come as close to the brink of expressing an opinion as you can, while still seeming objective. So if two countries are getting close to war, journalists know their readers want some kind of clarity about "will they go to war, or won't they?", but if they interject their personal opinion, they'll be breaking norms plus lose trust if their opinion turns out wrong. So they try to launder their opinion by eg asking experts, especially experts on their side of the aisle who they know will agree with them. Then they can have an article like "HARVARD INTERNATIONAL RELATIONS PROFESSOR: WAR INEVITABLE". Prediction markets would be one more way to do this.
You're not exactly a central example of a journalist-adjacent person in terms of caring about objective reality vs vibes, though.
Journalism is dominated by people-people, like most establishment types. They don't like or intuitively understand numbers, data, the scientific method, or abstract distributed systems of knowledge processing. They keep these things as far away as possible. They feel sure that knowledge, moral purity, wisdom and so on are rare and highly concentrated in a small number of excellent individuals. That's why so many of them hold markets in disdain, talk about algorithms as if they're magic spells, and many are deeply unsure about democracy (being as it is a system for collecting distributed wisdom) - frequently preferring terms like "populism" and "fascism" which are understood to be left-coded ways to signal disagreement with the foundational assumptions of markets and democracies.
Instead they work in terms of individual people and their brands. A friendly local academic is right up their street: you can phone them up at any time and they will talk for hours if you let them, so you can just ask them a question, they'll give you an opinion, and nobody will blame you if you repeat whatever they say and it turns out to be wrong because how could you possibly have known that a Smart Person was wrong? Like, that whole expectation would be completely unreasonable! They have a title! They work at a famous institution! The brands are impeccable!
Quoting some obscure prediction market runs counter to all these instincts. Firstly, it's a market i.e. assumes that wisdom is distributed instead of concentrated. Secondly, none of the markets have any well known brands, and none of the traders are even identified at all. Thirdly, they yield outcomes in the form of numbers. Finally, if it turns out the market was wrong, you can't deflect blame onto the market itself because none of your peers respect this form of knowledge either.
Democracy may occasionally function as a means of collecting distributed wisdom, but that's not it's purpose. It's purpose is to keep people moderately satisfied with their rulers, so there are fewer insurrections and civil wars.
Note that there are other associated functions whose purpose is to ensure that nobody with sufficient power to start a civil ware feels sufficient motivation to do so. That's not directly part of democracy, but it's been associated with it since the Magna Charta.
Democracy is a terrible way to determine truth but, since preferences are inherently subjective, it's better confined to aggregating those than wisdom.
Yeah and it feeds on itself, since there are people who got quoted once as an "expert". They pretty much advertise themselves as having a reliable opinion.
When a journo wants an expert to back up their slant they call the relevant person and ask for another quote. Voila: expert-backed, so it's not opinion it's news.
I think that news stories are often written to suggest that some kind of interesting outcome is possible, or that an event which occurred has possible interesting implications. On a lot of these topics, the prediction market will often be pretty sure the interesting thing won't happen, and including something like "Metaculus forecasts a 5% probability of China invading Taiwan by the end of 2024" would undercut the rationale of printing an article on a potential invasion.
I also think newspapers would rather print "July 2023 is on track to be the hottest month ever" than a percentage chance of 2023 being the hottest year ever, for example.
To be fair, 5% is unacceptably high for something so high-cost as an invasion of Taiwan. Of course, most people probably don't realize that.
Journalists definitely look dumber than they are a lot of the time because they're reporting on something they know little about, but on the other hand, this seems like a good guide to how seriously you should take an article on, say, fusion power or gain of function experiments or AI, written by someone who knows approximately nothing about these fields except that he briefly interviewed a couple experts whose statements he mostly didn't understand.
And at the same time, as best I can tell, most journalists are innumerate and scientifically illiterate, at least compared to anyone I'd want to hear talk about science and technology. (They're probably mostly less so than their readership, though.)
Sure, but as you mention, they already have a way to do it -- by asking "experts". In fact, the current way, from their perspective, is probably better, since "war inevitable" sells much better than "64% chance of war". It also allows them to package the same opinion multiple ways "Harvard IR prof" vs "Stanford historian" vs "geopolitical expert" vs ..., might all say slightly different things, and there you get plenty of articles out of it, whereas "the markets say X" is about all you get out of that style of reporting. It's boring, and more importantly, it's illegible. No one knows what 23% means, really. They just round down to 0%, and then get mad at you if it ends up happening if you talked about the 23% too much (cf. 2016 election). It's opinion laundering, but in a way that makes the end product really lame to read and might get people mad at you, and definitely will confuse some people, if not the journalist themselves (which I think is a larger part than you addressed in your reply).
I think it’s much simpler. Regardless of who a *journalist* is, if they want to cite a source to back things up to their *audience* then they need to appeal to a source whose expertise is legible to the audience. I’d the audience doesn’t know what a prediction market is, then it’s better to just cite a random government official or university professor.
Update: the Manifold market is WAY up, based on this paper: https://arxiv.org/pdf/2307.16892.pdf
Mild grumble:
"Apatites are materials with the general formula
A10(TO4)6X2±x, where A = alkaline or rare earth metal;
M = Ge, Si, or P; and X = halide, O, or OH. The
name ‘apatite’ derives from the Greek apat¯e meaning
‘deceit’ as a result of the diverse range of forms it can
take [17].
Here I consider the lead-phosphate apatite
Pb10(PO4)6(OH)2."
a) "TO4", then "M = Ge, Si, or P"
1) I assume TO4 should have been MO4
2) Ge and Si are in the same group, but not the same as P. Maybe As was intended?
b) "X = halide, O, or OH" Yes, halide and OH can substitute (F/OH substitution
is why fluoride helps tooth enamel) but O has an extra charge...
c) "where A = alkaline or rare earth metal" And lead is neither - though it _does_ have a +2 charge in many compounds, as do the alkaline earth metals, so that works chemically.
These are all nits, since the compound of interest is explicitly specified, but I wonder about the proofreading of the paper...
I'm surprised that people are taking a purely computational paper so seriously. I don't know any DFT myself but I talked to some colleagues who do and they seemed rather indifferent to it.
Well, they almost certainly didn't read the paper, just the tweet: https://twitter.com/sineatrix/status/1686182852667572224.
Lol thanks for featuring my extremely uninteresting comment but unfortunately my name is not Muhammed Wang, it is indeed just a pseudonym that I use sometimes
It would be pretty funny if your name were actually
dfkjgshdtvjksl.
Besides, shouldn't it be Wang Muhammed? ;)
Sure enough, the Metaculus superconductor question just resolved No.
Which is funny because both the Polymarket and Manifold ones are both spiking sharply up in the past few minutes (I assume some new positive news must have come out)
Not surprising though, for the reasons Scott pointed out.
Manifold looks to me like it's trading 99% on vibes
The Metaculus question has just resolved to No: https://www.metaculus.com/questions/18090/room-temp-superconductor-pre-print-replicated/
Gaia says: "Resolves NO, since neither of two replication attempts published on the ArXiv today by teams in India and China, respectively, were successful:
Synthesis of possible room temperature superconductor LK-99: Pb9Cu(PO4)6O
Semiconducting transport in Pb10-xCux(PO4)6O sintered from Pb2SO5 and Cu3P"
It's mildly amusing that, due to the manifold markets being a live, updating link, at the time I'm reading this, rather than being within a few points of each other, the manifold market is up to 50% while the static polymarket view is still showing 24% (although it has also increased to a little over 30% at the time of writing...must be some news for them both to increase!)
Baba Vanga is famous in post-Soviet countries; so much that in Russian her name turned into a verb meaning "to predict baselessly/ironically".
Although I'm of the opinion that she was a KGB asset (https://www.vagabond.bg/russia-brings-vanga-3460) to influence leaders who base their decisions based on mystics, astrology and such; and nowadays a common psyop method.
Noting as well some Metaculus questions with broader resolution criteria are opening soon:
Any LK-99 replication by 2025: https://www.metaculus.com/questions/18177/room-temp-superconductor-replicated-by-2025/
Date room-temperature and -pressure superconductor discovered: https://www.metaculus.com/questions/18169/date-for-room-temp-superconductivity/
Maximum critical temperature for ambient pressure superconductors by certain years: https://www.metaculus.com/questions/13668/high-temperature-superconductor-progress/
Commercial application for room-temperature + ambient pressure superconductor by 2025: https://www.metaculus.com/questions/18170/commercial-use-of-room-temp-superconductor/
Re: election betting, maybe there’s a legitimate fear that already-partisan people may exaggerate their positions more if their own money is directly tied to the election. If a media editor has $250,000 riding on an election, they might be less motivated to resist their outlet’s biases in pursuit of truth. That seems like a really high limit to me.
Naive question: Which *fundamentally new capabilities* has current-generation AI got which were not already present in expert systems, Markov chains, etc? I see the ChatGPT responses, and all I can think of is "cute, that's a bit like the dozens of articles back in the day about how we can generate fake Italian-looking names, except with whole articles."
This is a weird question to me. If you abstract it away enough then the LLMs aren't doing anything "fundamentally different" than a simple python script that outputs "hello world!" The dramatic increase in quality over what was previously available is what is novel and surprising. Yes, earlier AI was able to do some of the same things that these LLMs can do (although nothing was able, as far as I'm aware, to do everything that these LLMs can do in a single AI), but GPT-4, Claude 2, and others are _dramatically_ better at almost everything than previous AIs were.
First, there is a question of scale: Eniac was not "fundamentally different" from my laptop: both are in principle Turing complete (at least given infinite external memory), so they are capable of any computation (in principle), but in practice we all see rather massive difference. Also, current LLM's in particular, contain a fundamental innovation in the form of transformer architecture, compared to earlier recursive networks which had very short attention span and poor parallelism. Thirdly, deep learning in general, compared to earlier statistical methods, including Markov chains, but also shallow networks represents a kind of "fundamental change" because of its ability to form and use abstractions. And, finally, the fact that even simple statistical methods like Markov chains can generate text which can sometimes plausibly look like a text in a chosen human language with some traces of emergent grammar and sense tells us a lot about ourselves. Does this mean we are close to AGI (whatever it means) or at least human level intelligence? I am deeply skeptical, I think there are still many pieces of the puzzle missing (frankly, we do not even understand well the dynamics of learning in deep networks), and that we may soon see again blog posts discussing whether progress in AI has stalled (you might find amusing to read some discussions in the recent past about stalling of AI research, I found one from 2018!) but it is not all hype by Open AI and other big AI enterprises. The current progress is real, relevant, and quite fundamental, also for understanding of our own nature
I mean, the absolute biggest one is true parsing of natural language input.
It's weird to even include expert systems in this question. Expert systems are essentially just a static set of (usually) yes-or-no questions, hooked up to a bunch of "if" statements (every outcome of which is hand-picked). It's fancy version of a uQuiz. Like, not even similar.
Markov chains *are* sort of similar, but they're just... not good? They don't display even a facsimile of reasoning or recognition of the content of the input.
Also I think the year we've had with them undersells just how much of a leap chatGPT was towards passing the most rote kinds of humans. Just yesterday I had an interaction with a tech support agent who, based on our conversation, acted *less* capable of cognition and problem solving than my experience with CGPT.
Tech support agents have a script to follow and for a lot of them English is a second language (at best). So what you're seeing there is the speed at which the human can follow the script plus the number of off-script answers you make.
Rather like the previous answer saying that ENIAC was not fundamentally different than your laptop. That's true, but ENIAC was not able to compute fast enough to e.g. produce video output in real time. I don't know if you would consider this capability a difference in kind though.
Writing grammatically correct sentences in meaningful paragraphs about a broad range of topics was way beyond anything that expert systems or Markov chain models could do.
Code synthesis. Before GPT-3, code synthesis was essentially a pipe dream - there's an "old" (2015) article¹ by Gwern where he attempts to train a pre-Transformer architecture RNN to output CSS (which isn't even a Turing complete language), and it's... bad. "Success" in that case meant "it produces random output that sort of looks like CSS". GPT-3+ can synthesize syntactically and semantically correct code in every popular programming language just from an english description. This is fundamentally new.
1: https://gwern.net/ab-test#training-a-neural-net-to-generate-css
Taking about the use of Kalshi as a way for businesses to hedge against world events (like elections), does this distort the market as a pure prediction market? I have no economic knowledge about how hedging impacts say stock or commodity futures markets, but obviously hedging is a big part of the trading that goes on there and those tend to be accurate. But there the “bet” made when trading is directly tractable to the outcome being predicted, i.e. the future value of a company or commodity adjusted for risk, and hedges represent a form of risk adjustment that is directly related. It seems to me like substantial hedging on events that resolve to a yes or no would change the dollar value of a position in the prediction market from purely representing the market’s predicted likelihood of that outcome occurring to “the likelihood of this event occurring weighted by financial risk to the participants in this market if the event does occur” so there’s information being represented about things beyond the event itself. Maybe that balances out when market participants have risk on either side but for your example if Bernie Sanders we’re elected I would think all of those Wall Street betters are assuming the same direction and a very similar magnitude of the financial risk it represents for them. Of course if the market gets too skewed someone comes in to take the other side, but is the balance of the seesaw still weighted toward market-wide directional risk like this? Or am I not understanding how hedging impacts a prediction market (would be easy for me to do)?
It comes down to volume. If the volume is high enough, these hedges won't move it, because to whatever extent a hedge _does_ move the market, someone else sees that movement as an opportunity to make money (which it is, to whatever extent it moves the market away from a correct prediction), and so more money comes in and it corrects.
If the volume isn't high enough though, then yes, these kinds of hedges would indeed reduce it's quality of prediction. But it's hard to imagine that state of affairs lasting very long. the EMH is by no means absolute, but it doesn't _consistently_ leave the _same_ $100 lying on the ground.
>the EMH is by no means absolute, but it doesn't _consistently_ leave the _same_ $100 lying on the ground.
I mean sure it does, sometimes for years.
Citation needed. I'm skeptical you have an example that is similar in the ways that matter to the above hypothetical. This is the kind of example that is one of the very few ways where the EMH would actually operate mostly as expected in the real world. There are lots of cases where the EMH doesn't work in the real world for all kinds of reasons. I just don't think that "hedgers move the prediction market away from the consensus prediction" is one of those.
Yes, if there's a systemic factor causing one side of the hedge to be more desirable (e.g. there's a presidential candidate everyone thinks would be bad for the economy), this can happen. It already does happen in the stock market - put options on the S&P500 are worth more than you'd expect purely based on the distribution of actual S&P returns because they're used as hedges.
(This does in principle mean you can make money by selling SPY puts, but it's picking up pennies in front of a steamroller, so not enough people are willing to do it to overcome this imbalance).
Right, which is exactly the point: There is a limit to how far hedgers can move a market. If they move it too far, those pennies become dollars and someone dashes out in front of the steamroller until all that's left is pennies again. So in an extremely pedantic sense, yes, hedgers can and will move the market. But they will only move it so far that no one thinks it's wrong enough to be worth correcting.
lol at "a couple of times a day I check the prediction market on [superconductors]" as if it were the stock market or expert opinion or anything except the agglomeration of what some random retards online think
Boy do I have some news for you about the stock market and expert opinion.
But those are *selected* retards, Shaked!
(1) " Nobody worries about guys who have put $1000 on Labour at the local Paddy Power."
Because Paddy Power's is a bookies, we know how bookies work, and people absolutely do try to rig race results:
https://en.wikipedia.org/wiki/2002_Breeders%27_Cup_betting_scandal
"It’s always “We Are Various Prestigious Economists, And Have You Considered That Economics Tells Us That Prediction Markets Are Good?”
Feck it, now I am *doubly* against prediction markets if the economists are out there telling us we should have them.
Look, the Smart People want their fancy betting markets because they're too high class to step inside William Hills and put 50p each way on the outsider in the Gold Cup? Sure, let them, but let's not kid ourselves this is anything other than "I want to use my Giant Brain to make money" and not a way of "we will find out the best way to decide 'is arsenic a health supplement?' by using the forces of the market".
(2) Robert Miles will be at this bunfight, eh? Maybe he can give you all a selection of his hits:
https://www.youtube.com/watch?v=DvyCbevQbtI
(Yes, I'm sure he's never heard *that* joke before).
(3) I do absolutely want to get some informed opinions on the superconductor. My priors are that this is going to turn out to be a hoax, fraud, or over-enthused reporting of minor results like the previous efforts. If anyone can give me a semi-solid "nah, this looks sorta legit" rundown I'd appreciate it.
(4) "I ran into some finance people at the NYC meetup this week and asked why they weren’t using more advanced forecasting technology - prediction markets, superforecaster tournaments, calibration training, that kind of thing. A common reply was “who says we aren’t using it?”
That's the part that makes me repose even less confidence in "Allowing election bets makes Wall Street less interested in elections, not more!" As you say, knowing the outcome of elections is worth hundreds of millions to them. If they have their own internal markets, which live or die by being correct, and that market is telling them "X is likely to win the next election", it may well be in their interest to try and make sure Y is selected instead of X during the primaries, or to help out Z to get elected instead. Maybe they throw money at Z's campaign, or they run advertising, or they put out subtle influences to influence *other* prediction markets, or they get polling companies to do polls crafted to return "Z is the people's choice!" and then release those results.
My take on it is that if there's a lot of money riding on the result, then hell yeah somebody will try and influence the outcome:
https://www.theguardian.com/politics/2005/apr/13/uk.localgovernment
New fears over postal vote fraud
https://researchbriefings.files.parliament.uk/documents/SN06255/SN06255.pdf
Electoral fraud since 2010
https://www.youtube.com/watch?v=L1QkUndT4Zs
PC Returning Officers Scandal
https://www.deccanherald.com/state/top-karnataka-stories/returning-officers-to-be-questioned-over-votegate-1163985.html
Returning officers to be questioned over votegate
https://www.premiumtimesng.com/news/top-news/548443-kenya-decides-electoral-returning-officer-missing.html?tztc=1
Kenya Decides: Electoral Returning Officer missing
I'm as cynical about the Smart People as you are, but Predictit has a much wider reach than Paddy Power.
I am very confused about the US government's overregulation of prediction markets. Is there any rhyme or reason for them to come down so hard on this compared to other countries, when the general rule (or general perception anyway) is that the US regulates most things much less than most other western countries?
Also, is this extreme hostility to gambling (political betting is downright illegal, which I was very surprised to learn; it's a completely normal part of political campaigns in Australia) primarily seen as a left-wing socialistic thing, or a right-wing moralistic thing? I'm guessing the former, based on the tone of all the rhetoric, but if it were the latter that would at least (sort of) fit better with common preceptions of US governance. I'd still find it very weird, though.
I think the perception of the US not regulating stuff much is mostly wrong (especially financial stuff). Specifically on gambling and securities there's some pretty strict regulations.
Well, the USA is the country that did the Prohibition. It was founded by Puritans, so whilst it may be more capitalist than most countries in general, when it comes to the regulation of moral vices ....
Left vs right wing morality is just a confusing concept here, because "right wing" is usually meant to mean conservative in the US context and the US is historically very religious, but left wing morality and conventional Christian morality are just two sides of the same coin. The actual opposite to left wing socialism would be something like libertarianism or anarchism, neither of which are big on regulation of gambling.
Left or right is hard to define, particularly if you're defining "left" as "socialistic" and right as "moralistic".
The historical core of opposition to gambling in the US is religious, rooted in the teachings of the churches most affected by the early 20th Century "Social Gospel" movement of American Protestantism (the United Methodist Church still officially denounces all gambling to this day).
And the "Social Gospel" movement was the religious wing of the early 20th Century Progressive movement in the US. (Other causes of the "Social Gospel" movement included alcohol prohibition, the elimination of child labor, the living wage, and political corruption.)
Assuming that the US is just “less regulation” is just as bad as an assumption as assuming that Western Europe is just “more left”. These might be useful shorthands for certain central policy debates, but not across the board.
Historically there may have been hostility towards gambling, but I dont think thats been true for quite some time. It also depends on what type of gambling. Horse and Dog racing have long been legal in many many states. Jai alai is a game popular in miami that has long had gambling associated with it.
For long periods only Nevada and New Jersey (and some river boat casinos which i am not very familiar with) allowed gambling such as cards or dice or slots, but that has changed a lot. Native American reservations have long been able to open their own casinos. And some states started allowing horse and dog tracks to open casino related operations.
The big changes recently have been to allow sports betting in more places as well as remove requirements for animal racing to be attached to casino operations. Now many, if not most, states allow sports betting.
Changes to gambling regulations over the past ~15 years have been similar to the changes to marijuana regulations over the same period. States saw how much money they could make by allowing it but heavily taxing it. So more and more states are removing regulations. There has been some opposition from religious groups or other moral objection; and at first New Jersey and Nevada based casino operators and the states of New Jersey and Nevada tried to prevent other states from allowing gamblings for fear their monopolies would be lost. The casino companies just adapted while Las Vegas became more about entertainment than just gambling (new jersey had been on a long slow decline for a while already).
I suspect the difference in attitudes to gambling between Australia and the US might mostly come down to the roots of Australia's political culture being way more Catholic compared to the mostly Protestant US political class of the past.
Thanks for linking to https://fatebook.io! Interested to hear people's impressions
Also - the "who says we aren’t using it" anecdote about using Metaculus in finance is surprising, I'm curious whether this extends beyond the intersection of ACX readers and finance people.
Does anyone know of a similar site or tool for implementing group prediction markets (multiple people can make predictions, aggregate score, etc). I’d like to implement it at my work, but wouldn’t be able to use a public market like manifold.
On Fatebook you can share questions with a list of email addresses who can all make predictions. I'll probably add forecast aggregation soon, maybe today!
Fatebook is for registering predictions (e.g. 70%), it's not a prediction market (e.g. "Buy YES for $0.58"). I think this works better for very low numbers of participants and is more intuitive, but you miss out on nice properties of prediction markets.
On Manifold you can now create private groups, which might fit your usecase?
"the 'who says we aren’t using it' anecdote about using Metaculus in finance is surprising"
Why is that surprising? Finance is full of number-heads (aside from those who are farther from the numbers and make money primarily due to being born rich and/or schmoozing). They're exactly the types I would expect to read this blog and/or be interested in prediction markets.
Also, the incentives for them to use every source of information possible to determine the right financial bet to make are...huge. That's the whole deal with the stock market. So it doesn't surprise me at all, tbh. The only reason I can think of for this being surprising to someone is if they didn't think predictions markets were a source of meaningful information distinguishable from noise. Which, if that's your opinion, fair, though I don't agree.
Surprising only because I'd spoken to someone who'd spent some time pitching superforecasting-style forecasting consulting work to finance (and other) companies and had concluded that they largely weren't interested. It might be that I'd over-updated on this and a couple of other anecdotes.
Ah I see, interesting
Depends on how they pitched it too, I imagine. "This is the best way to figure out the best way" - yeah, nice, see you later. "This will make you gazillions and put you ahead of the competition!" - tell me more!
Yeah, in house proprietary software for...prediction markets? What super secret, hard to explain design elements would that have? My best guess is that some people misinterpreted the intent of Scott's question and are referring to tools that scrape Twitter/other sources of big data for consumer sentiment to predict earnings.
https://twitter.com/iris_IGB/status/1685265405386878977
For those betting on LK-99 - russian anime girl(?) anon claims reproduction on her kitchen (yes, we are in weird timeline)
There's also a rather dodgy Chinese video, and now allegedly the US Navy. Or something.
So it's starting to look like there might be something there, but nobody really has come out to say 'yes/no' as yet.
Oh, sweet lord. I don't know this Sinclair Chen, but good grief: we have no god but Mammon?
"Capital markets today serve a function not just of allocating investments, but also of aggregating information. They incentivize smart ambitious professionals to work hard, grinding 60 hour weeks in their glass towers, striving be the best predictors of the value of assets - and yet, any American from any background, whether black or white, straight or gay, christian or atheist, female or male, can beat the market so long as they see or infer something that the NY suits don't see. And by beating the market they improve it. It is the purest meritocracy and the purest truth-finding machine and there's nothing else quite like it in the world."
All hail the blessed and divine market! Let none gainsay the market, for in Mammon our pelf there is neither gay nor straight, black nor white, or male nor female! There is nothing like it in this world for it is not of this world but comes from the one true holy source of all motivation - getting stinking rich*!
No wonder they go for "you have blood on your hands" over the top J'Accuse! How many ordinary Americans read the stock market or base their behaviour on "Ah, I see Disney stock is down"? Yet we are supposed to give full credence to "a bunch of math nerds playing with a toy market would be followed, understood, and believed by the man in the street, who would then have changed his behaviour about masks and Covid"?
Lady, I wore a mask. I got vaccinated. I isolated during the lock-down. I still got Covid, and no bunch of smart maths predictions would have prevented that.
*If those smart ambitious professionals got paid in "here's an apple and a pat on the head", how many grinding 60 hour work weeks for the pure truth-seeking sake of it would they put in?
I'm reminded of this excerpt from "A Cynic's Guide To Fintech", particularly the first few sentences of the second paragraph:
"Fintech business model #5. Assuming that the regulators will be more inclined to listen to your whining than to the incumbents'
"Usually a bad idea in financial services. Regulators basically don't like small financial services companies. There are severe diseconomies of small scale in supervising them, they are more prone to blowing up and they don't do very much for your career. And financial services is an intrinsically regulated industry where consumer protection is often very rigorous for a good reason. So the whole Uber idea of just blatantly breaking the law and then sending out a press release about how uncool and obstructive everyone is being is not going to go down well. Several fintech startups have already found out that there is no exemption for tech companies from the money-laundering or consumer finance laws, and that regulators usually don't care if they've driven someone they regard as a rule-breaker out of business. The existence of financial regulations also tends to mean that fintech companies need to have a lot more capital lying around than they would if they didn't need a financial services licence."
To be honest the superconductor thing is a big embarrassment for the prediction markets. I am a physicist, and among people who are experts in superconductivity (myself included) I can't find anyone who is remotely optimistic. The data looks way too fishy and does not really resemble the usual data for confirmed superconductors, it has logical inconsistencies, the authors are clearly ignorant of relatively basic concepts in superconductivity, and this report is part of a long history of dozens or perhaps 100 "unidentified superconducting objects". If this forecast had been made only by experts in superconductivity there's no way it would have gone above 1-2%. To see the prediction markets at 50% makes me inclined to dismiss the whole approach as anything that can remotely be described as "predictive".
The bar of “a successful replication” is fairly low though. If this thing is really just a powerful diamagnet, then some group will replicate that and get the same misinterpretation, and it will be “a successful replication”.
Disagree: the original data seem inconsistent with just a powerful diamagnet, so if somebody showed that it was a powerful diamagnet, it would not be a successful replication.
FYI, this market has already resolved NO: https://www.metaculus.com/questions/18090/room-temp-superconductor-pre-print-replicated/
(Not intended as a gotcha, just sharing useful additional information from elsewhere in the thread)
FWIW, given the openness of the original researchers to offer their original material for others to test independently, I think intentional hoax/fraud is unlikely, and most likely it’s just researchers making mistakes.
I've also bet against on Metaculus. I hope some foreign physicists are cleaning up on Polymarket.
I really would like an "Idiot's Guide" to what is going on right now, because my initial reaction to this was "oh yeah, yet another hype that is going to explode when poked", but I'm seeing claims that the results have been replicated elsewhere, so I don't know what to think.
I'm still very dubious, but it looks a bit like there *might* be something in the theory there? Help?
Question for the physicists on here: if it turns out to be just a powerful diamagnet, not a superconductor, is that novel or useful or interesting?
It has been replicated. The theory behind why it works has been simulated. It's basically an engineering problem in the developed world at this point. Maybe the US will crank out a Nature article by the end of the year but really who cares about them. The "Rationalists" as per yoozh are behind the curve and have a Western-centric conservatively-slow-to-update bias. They, as a whole, don't even accept UAPs despite the vast evidence in favor of their existence. The less time you spend in Flat Earth level communities like this one the better.
(1) Allegedly replicated on somebody's kitchen table in Russia and via a dodgy video from China. I want something a teensy bit more rigorous than that
(2) Dunno about "Rationalists", being slow to accept extraordinary claims without good supporting evidence just seems sensible. We've been at this rodeo before, remember?
(3) I don't belive in UFOs, even if they have been rebranded to UAPs. Though according to the guy telling all and sundry that they're real and here, the Vatican knows about them too. So until Pope Francis issues some kind of pronouncement, yeah I remain to be convinced.
(4) Why so prejudiced about the Flat Earth?
On a sidenote, I love that the 'USO' line is also a Japanese pun for 'lies: https://gwern.net/doc/science/1997-cava.pdf#page=5
> Although I like the result of this decision, I’m worried about the ruling that no-action letters constitute binding commitments whose amendment or cancellation requires careful agency action with every t crossed and every i dotted.
What is it that you like about the decision? It seems to me that the bad thing you describe in the second clause here is the primary result of this decision, so I don't know what there is to like about it.
I just meant I like the fact that PredictIt can keep operating.
The LK-99 market on Manifold is going absolutely crazy right now. It's a weird dynamic, where hundreds of newly registered users are throwing in 500 mana on yes (that's the initial play money allocation), sometimes even buying mana for this purpose, while established users already feel overleveraged and don't have the liquidity to correct the market.
"My my, whoever would have thought that the prospect of EASY MONEY even if it's only play money would attract hordes of treasure seekers and the SENSIBLE RATIONAL WISE ONES would be overwhelmed so that the market was not, in fact, self-correcting", she said drily.
Replying only to say that this -- "although 180 of those were by a person named Chris Greenwood who somehow submitted the same message 180 times. This is probably a metaphor for something" -- made me LOL.
I'm surprised you didn't mention that the Salem prediction market contest ended yesterday.
> Why would any government agency ever give a no-action letter now?
https://en.wikipedia.org/wiki/No-action_letter
> A no-action letter is a letter written by the staff members of a government agency, requested by an entity subject to regulation by that agency, indicating that the staff will not recommend that the agency take legal action against the entity, should the entity engage in a course of action proposed by the entity through its request for a no-action letter.
This makes a no-action letter sound like something that should exist (i.e. you ask an agency if something is legal, rather than guessing and fighting a long legal battle later) but if I remember, this case makes it seem like they don't actually do that? You wrote last year (https://astralcodexten.substack.com/p/mantic-monday-81522?):
> On August 4, the CFTC reversed itself, saying the PredictIt had “not operated its market in compliance with the terms of the letter” and that it had to shut down by February
And then wrote that it wasn't clear why this reversal had happened since they stuck to the stated rules, but that it seemed like some sort of regulatory capture happened, or the agency just changed its mind. If these things can happen, then what's the point of such a letter? It doesn't actually reduce uncertainty. Rather than engaging in legal bone-reading around what laws say and what (if any) precedent exists, now companies have to engage in psychological bone-reading around how binding the commitment is. Making them binding is the only way for such a letter to be of any use.
Lemme introduce you to an organization called the BATFE, ADA compliance, and the definition of a "machine gun."
Oh, I'm aware of that fiasco as well.
The Manifest speakers & special guests page has added Robin Hanson and the CEO of the Forecasting Research Institute (Tetlock's new organization).
Man, this story is wild. I can't imagine how much of a field day Matt Levine would have had with it if it had happened in a real-money market.
https://news.manifold.markets/p/manifold-predicted-the-ai-extinction
This seems to confirm that contra what Scott says about "Allowing election bets makes Wall Street less interested in elections, not more", prediction markets can indeed be rigged.
Maybe you *want* insider trading for prediction markets, because that's based on insider info because the Platonic Ideal of a prediction market is laid out in that piece:
"Prediction markets allow people who lack expertise about certain topics to form precise models of what the future could look like thanks to the live-updating probabilities that are generated by traders."
But uh, in that case you could just go directly to the experts and people in the know and ask them "are the aliens in contact with the Vatican little green men or Greys?" instead of going the extra step with a betting market.
Insider trading is about making a profit, not adjusting the market, and the story shows this is what people were doing, and that despite all the lofty rhetoric, if A puts a ton of money on "YES" then people will be reluctant to buy "NO" even if they know better, or think they know better. So it is demonstrably possible to influence public sentiment and change opinion. The big Wall Street firms might be very happy to dump a million or so into prediction markets, even if they lose it, because it shifts public sentiment towards Bob instead of Billy, and they want Bob because he is not going to Sarbanes-Oxley them if he wins the election.
"Users started commenting on the market, with weak speculation that there could be insider trading. But because this was Quinesweeper’s first market, others feared he was planning to load up on YES shares and fraudulently misresolve the market to yes regardless of the outcome. The Manifold team assured users that we would fix the resolution if it was resolved incorrectly so that users could trade on their true beliefs, and not have to factor into their probability the chance of fraud."
But they don't seem to have been able to control the insider trading; they didn't want to delete the market, and the efforts to bury it only made it *more* conspicuous and got *more* people interested and jumping aboard.
"A lesson I would recommend learning is that many AI markets on Manifold are heavily insider traded"
I would imagine if prediction markets do get off the ground/are allowed use real money, this will spread to other markets. When there is the chance of money being made, there is always the incentive to manipulate or rig the market in your favour, even if it is 'only' using insider info.
"Firstuserhere has also informed me that a few YES insiders agreed to not buy the market above 55% so they could conceal their continued accumulation of cheap shares. They really were working together to maximise how much profit they could make off of the ignorant NO bettors."
"From what I can tell, turning the market into one big meme worked surprisingly well, and some users genuinely believed there was no statement, and that the only reason the probability was so high was “for the memes”.
And these are prediction market users who presumably have a good idea of what way it all works, and yet they could be fooled into thinking the entire market was a fake, or that some outcome other than the reality was true.
I don't think this shows as hopeful a way forward for prediction markets as the author thinks; forget predicting AI risk, work on "if we do manage to get prediction markets mainstreamed and if we do get permission to use real money, how do we avoid manipulation, scams, and profiteering if what we really want is to get the best information, not just make money?"
I think the biggest problem with Scott and the other pro-prediction market and pro-crypto people is that they think that markets just magically work in a state of nature. The reason that our real-money markets work so well is *because* of the regulation from SEC and friends, after a century of lessons won the hard way. But then when the regulators say "hold on a moment, this seems prone to manipulation and insider trading", they scream that the government is holding back innovation.
Obviously, not *everything* the SEC does is beneficial or necessary, but on a scale from "no SEC" to "status quo", I'd guess that it is at least 95% necessary. We've seen what happens without regulation from all the financial panics and scams of 100 years ago, and then the crypto world helpfully speed-ran the entire history of scams and fraud in a decade for the benefit of anyone who forgot. And now we have demonstrations like this.
>We've seen what happens without regulation from all the financial panics and scams of 100 years ago
I can't remember the exact details but I read a novel from the late 19th century which involved, amongst other things, this kind of financial scams - not quite crimes because, as you say, regulations hadn't been imposed then. It involved the representative of a large American financial combine who worked in London and was making trades on the stock market; using such advanced technology as THE TELEGRAPH he was able to get early information from America about shares and buy cheap to sell high. Fair enough, sez you, insider trading maybe but not illegal at the time.
But what he *also* did, on instructions from the combine, was manipulate share prices so they could pluck pigeons. He might buy up a lot of shares in a particular company, and either he'd 'let slip' to selected investors that this was a Good Thing and he was letting them in on the ground floor early before it became public knowledge, or others would see this and think "Hmm, if the XYZ Combine is buying this, it must be about to boom" and the pigeons would put their shirts on buying shares, thus driviing the price up.
At which point the combine would sell up, the price would crash, and the pigeons would be ruined, while the combine either knew the company was going to collapse and used this way to maximise their returns, or they could then buy back the shares cheaply in the company which was going to grow naturally. No regulations, no regulatory bodies around to stop this kind of sharp practice.
The main plot was about a guy who invested this way, was ruined, and his scheme to get his revenge (and his money back). The difference between Then and Now is also shown in the Sherlock Holmes story "The Stockbroker's Clerk" (where between you and me, I often wonder exactly how stupid was the client? But then again, the con artist tricking him knew exactly what to say to a young, desperate, and not too bright guy to push all his buttons):
“Well, the whole thing hinges upon two points. The first is the making of Pycroft write a declaration by which he entered the service of this preposterous company. Do you not see how very suggestive that is?”
“I am afraid I miss the point.”
“Well, why did they want him to do it? Not as a business matter, for these arrangements are usually verbal, and there was no earthly business reason why this should be an exception. Don’t you see, my young friend, that they were very anxious to obtain a specimen of your handwriting, and had no other way of doing it?”
Nowadays we don't let "business matters" be "usually verbal" for the reason that we no longer do things on a handshake and 'a gentleman's word is his bond'.
The BBC broadcast a radio adaptation of the 60 canon stories in the 90s and here is a link to the shows, including The Stockbroker's Clerk:
https://fourble.co.uk/podcast/sherlockholm
I really like what Bert Coules did with The Lion's Mane; he took the bare bones of the plot and did a completely different story around it, with bonus "The Man from U.N.C.L.E." and Star Trek references if you can catch them!
Cf. the Buying Things From A Store FAQ: https://astralcodexten.substack.com/p/the-buying-things-from-a-store-faq , especially part 10.
The problem isn't that regulation can't fix an issue. The problem is when you constantly complain about said regulation, *including in this very post*.
I think there should be low enough regulation to allow things to exist, but high enough to allow them to be good. I don't know exactly what that level is, but I think a good first step would be allowing it to exist, then seeing whether it is good or not and what steps would be necessary to make it so.
I do hope at least some regulation can be at the market level - ie some market decides insider trading is bad and it will ban anyone who commits it. Compare to eg casinos regulating cheating. That way we can have more flexibility and diversity in level of different regulations.
Well, just right there - the market exists, and we've seen someone (a couple of someones, indeed) hopping in to Win The Game. I think we have a good general idea of how human nature is going to work out, even with the nice civil rationalists (and that last is not a sneer, you lot generally do try hard to be good people). But even good people succumb to temptation.
See 4.1 at https://astralcodexten.substack.com/p/prediction-market-faq#%C2%A7what-are-the-most-common-objections-to-prediction-markets
Yes, prediction markets would be unnecessary if every human could identify the people who know the true information, go to them, and trust them to give a true answer. But we can't.
It looks like this prediction market ended somewhere between the accuracy level it would have had if there had been no insider trading, and the accuracy level it would have had if every insider had disclosed all the knowledge they had publicly. I think this is good for prediction markets (though not for AI orgs who want to keep their information secret)
yeah, but you had people using inside information to make money (even if it was only play money) so I don't think that prediction markets will be free of the problems that plague ordinary markets unless and until the prediction market bods lay down the law with some heavy regulation.
And the piece says that Metaculus was reluctant to do anything much in the name of transparency and infosec, so the market-riggers were enabled to go right ahead.
Imagine this was a wider world, public, market with real money and some immediately vital topic (like an election). I have no confidence the same wouldn't happen.
The question: "Will Israel pass the judicial reform bill?" is not very well formed, since "Israel" is a diffuse entity. The parliament (Knesset) might pass the bills, but the Supreme Court (Bagatz) might rule them unconstitutional, which is weird, since Israel doesn't have a constitution. Then... well, forecast that.
Some people say Civil war.
Dating apps + prediction markets makes sense culturally. The kind of microeconomists who care about matching algorithms also love prediction markets.
Most hedge funds, index fund managers, and mutual fund providers have theirs filthy green hands in the pockets, arseholes and mouths of both major presidential candidates by Election Day.
Since these organizations are soon going to own all the property in America and make us rent our lives away, I think it’s fair to say no entity can exceed their investment and capacity for over handed and underhanded force should the situation call for it.
Rigging the election might be possible with either large Russian or Chinese intelligence investment, but really only with the explicit compliance of a major town square owner like Musk or Zuck. But why on earth would they stick their neck out for a state that just experienced a minor coup, and a state with its own private internet where they don’t stand to profit much. Especially not Musk who would lose SpaceX overnight if he was ever discovered to being letting the enemies of the US tamper with ... twit... twi.... X ‘gag’.
Suffice it to say, it would need a trial run with oversight, but it’s prob fine to bet on the election. Maybe a little funny too since both parties seem to be competing to see who can jam their own foot in their mouth harder.
Would a rationalist content search engine be something you, reader, would pay a small amount for? Say $2-6/month. Imagine it has all the features you want.
I don't get it. Who are these people who are anti-prediction market and how do they profit from blocking them?
>"Why would any government agency ever give a no-action letter now?"
Cynically, perhaps because it'll be binding(ish) on your successors who might otherwise be inclined to more energetic enforcement than you would prefer?
> Anyway, Vanga’s other predictions for 2023 include:
> ...
> A powerful solar storm
Article I'd read earlier today:
> Giant Solar Storm Struck Earth, Moon And Mars Together For First Time In History: Study
This marked the first time that a solar event was measured simultaneously on the surfaces of Earth, the moon, and Mars.
https://www.ndtv.com/science/massive-solar-storm-struck-earth-moon-and-mars-together-for-first-time-in-history-study-4267181
Just came across this interesting article profiling several highly successful traders on PredictIt. It's amazing how much trouble people will go to to win money on PredictIt.
https://washingtonmonthly.com/2022/04/03/the-art-of-the-pump/