By 2025, will most well-educated people expect AI to within 10 years be better at intellectual work than 99% of humans?
➕
Plus
158
Ṁ45k
Jan 1
20%
chance

Prediction is from James Miller: https://twitter.com/JimDMiller/status/1741989384881058050

James Miller: Prediction for 2024: Most well-educated people will think that within 10 years AI is going to be better at intellectual work than 99% of humans, and this expectation will have profound effects including reducing the importance parents put on their kids doing well in school.

This resolves to YES if a majority of college educated people would answer this way on a survey, or we would reasonably expect them to do so.

Resolves to NO otherwise.

If the answer is sufficiently unclear and market price reflects this, market will be resolved via a poll asking whether the worded claim is true.

Get
Ṁ1,000
and
S3.00
Sort by:
opened a Ṁ500 YES at 25% order

Raising my limit to 25% on the o1 model release, anyone want to bet against?

@JamesBaker3 The part that I'm having trouble modeling here is how much this is penetrating the awareness of the average college educated person. Perhaps 25% odds isn't wrong?

But I'm wondering if mostly people don't think much about AI and if you were to sample randomly from college educated people they'd be more like "What??", rather than "Oh yeah, that's what I'd expect." People certainly aren't acting like they think humans will be obsolete in less than 10 years?

@ChrisPrichard I wonder if there have been any existing surveys that would be relevant to this one?

I model people more going "10 years is far away, sure!" without pushing that belief into their life

opened a Ṁ1,000 YES at 13% order

Thanks @ChrisPrichard , I put another 1K on yes for you or anyone

Just wondering out loud: I keep thinking this question should be priced way higher, and am wondering about the perspectives of NO holders. I wonder if there's a focus split here, with the "people will think...99%" folks betting yes and the "profound effects" folks are betting no?

I think people will be confused and contradictory about the effects in their own lives, but that they'll readily believe the premise (in survey form) even if they don't know what to do about it practically. NO holders, what's your story?

I don't think people are very forward-looking in this way. Unless it actively unemploys half of them they'll "cope" and imagine it'll never get better.

Real impact of AI in peoples life today is limited. Most people never heard of scaling laws, but they heard of "hallucinations". So on top of the unknown crowd, there is a big portion of people who are sceptics.

The question description is what matters not the title.

  1. Would this be based off survey responses or revealed preferences? Aka if people say they do believe in A(G)I but 529 plan contributions don't meaningfully budge, would that count as a Yes?

  2. Is this for the US alone, for developed countries, or for the world as a whole?

  3. Does "most" mean 50.01% or, say, 60%?

  4. Do we know what percent of well educated people believe in this today?

Rephrase the misnomer euphemism "well-educated" to the neutral "college graduates". If you're feeling pessimistic, try "over-schooled"

Zvi already did exactly that: "resolves to YES if a majority of college educated people would answer this way on a survey"

I'm talking about the title. I only made my comment because the description didn't match the title. I made a similar suggestion on another market about Argentina, where the title and description were mismatched. If correcting the title doesn't add any extra verbosity, it should be required

But the source (and therefore the spirit of this market, I presume) is a quote, from a prediction made by a 3rd party.

Yeah I noticed that too. Even then, I took issue with the wording, it's not that important but I'm someone who wants to change the perception that attending 25% less school converts you from an enlightened "well-educated" (or simply "educated") person to an "under-educated" (or even worse, "uneducated") person

bought Ṁ1,000 NO

It's irrelevant. Even the mediocre well-educated sheep will no longer believe that by then.

I appreciate the practical take, but my point is transparently symbolic, so I still want to see it changed

bought Ṁ50 YES

Anyone who denies that AI won't be by 2030 is deluding themselves, let alone 2035

Feels like a question that will get haggled over.

If you define the question and the poll type now, may be able to convert this into a futures contract over that later poll.

@ZviMowshowitz Does the resolution here depend at all on the "reducing the importance parents put on their kids doing well in school" part? (Which seems oddly specific and non-load-bearing to the central claim, to me)

To the extent that would shape people's responses.

Thanks, I'm hearing that as agreeing that it's not an additional conjunctive clause (that people must believe "reducing" vs other effects) and that the resolution depends squarely on people's responses to "within 10 years AI is going to be better at intellectual work than 99% of humans"

I believe that if a well-educated person utilizes artificial intelligence in their work, the results will correspondingly be of high quality. If a person with lower intelligence attempts to generate something, then accordingly, their result will be primitive. I also want to add that there are numerous resources like https://customwriting.com/ where intelligent individuals work and produce quality work, which looks much better than what CHatGPT produces.

This hinges a lot on what constitutes "intellectual work" ...in many domains AI already exceeds humans. But Miller's explanation that it would reduce the importance parents put on their kids doing well in school suggests, at minimum, that "most human variation in intellectual skills will be economically unimportant." I think this unlikely.

I think it's more unlikely that a majority of college educated people act on this, although I'm less certain they wouldn't say so on a survey.

© Manifold Markets, Inc.Terms + Mana-only TermsPrivacyRules