Kevin A. Bryan

@testingham Short and medium run is where the policy relevance is. Let the omnipotent AI consider issues in 2100, no? I mean this nonfacetiously

When economists talk about "post-AGI futures" they mean abundance: 10X growth. When AI people talk about "post-AGI futures" they mean omnipotence: growth until physical limits are hit. A natural question to ask the economists: if AI gets you a century's technological progress in a decade, why will it stop there? Additionally, from the perspective of omnipotence, the economists considerations all start to seem *provincial*. Considerations about redistribution, inequality, democracy, power concentration, dignity, meaning. (This observation comes from me trying to reconcile post-AGI conversations at Lighthaven, and post-AGI conversations at Asilomar)

@erinhengel @JeffreyASachs They also did a large scale AI analysis of language choice - which she objected to because "it is AI" - and when asked about the actual anthropologist writing it - a Harvard professor with a famous book - she pretended she had never heard of him.

@Afinetheorem @JeffreyASachs She says the report is vague and its justification for singling anthropology out poorly reasoned. These are true! I mean using the words of a single scholar—even the president of the field’s principal association—isn’t sufficient reason imo to condemn the entire field!

@erinhengel @JeffreyASachs She was explicitly asked "is it a problem for the head of an academic society to make statements like this". Total layup. And she wouldn't say no. And biologically speaking. "Uncontroversial bimodal sexual characteristics" is more common than "humans have two legs".

@erinhengel @JeffreyASachs The president of their academic assocation argued that the point of the field was fighting capitalist society. If you say the report is wrong you need to least pretend to argue why those two examples are in fact ok.

@erinhengel @JeffreyASachs Because the interview is literally based on her complaint that the Vanderbilt report thought anthropology had been politicized on the basis, among others, that their annual conference banned a seminar for even arguing human sex is a useful analytic category, and that....

@Afinetheorem @JeffreyASachs On 2), I think her point is that, biologically speaking, sex really *isn’t* binary. I’m no expert on this topic, but that was my understanding as well. On 1), those were Gupta’s words, not hers, I believe. I don’t think she even thinks that.

Jesus. Talk about not doing your side any favors. chronicle.com/article/is-ant…

@SBenzell According to your social welfare function weights ;-) (more seriously, the exact answer is one for society at large to answer, not me)

@Afinetheorem Total or average

I see misunderstanding of "We Must Act Now" letter. Avg policymaker thinks AI is 2026 + epsilon. Many think it's snake oil being sold by Sam. Point of letter is that tons of econs/AI policy folks think "most impactful invention ever, like IR on compressed scale", is closer. 2/2

@andreamoro Let me take a look this week. I noticed it also this week so maybe I screwed something up on the last edit.

@Afinetheorem It's not just tikz. I have failed to have it recognize all figures from tbe pdf, tried 3 papers

Great example of the benefit of having your design notes plus a bunch of examples plus your LaTeX draft all in an AI linked folder. Can spin up interactive intro to main results of our paper in a couple hours instead of a couple days. My AIs "know what I like", in a sense. 1/2

@amcafee @lugaricano @BrianCAlbrecht Absolutely. So I think the debate is on "is this because they are too worried about AI danger or because they don't understand its importance in all directions"?

@Afinetheorem @lugaricano @BrianCAlbrecht I hear you, but I think policy folk are currently too eager to put “guardrails” on AI — e.g. opposition to autonomous cars in many cities

I too agree with @amcafee's reaction.

@stuartbuck1 Yes, and 100% I agree that this is not at all what the letter says, and the folks who worried it would be interpreted by some in that way were right and I was wrong.

@Afinetheorem True, but my problem with the statement is more about this: x.com/i/status/20768…

The economists' letter on AI was easily the most unimpressive and useless thing any of the signatories have ever done! (I know many of them and hope they aren't offended....)

@erinhengel @JeffreyASachs Of course! But her claims were that 1) it's not politicized for AAA head to say point of the field is struggle against capitalism & 2) not only is "sex is binary" wrong, but it's not even up for debate scientifically. This is *exactly* the politicization Vanderbilt worries about!

@Afinetheorem @JeffreyASachs And she’s absolutely right that “free speech” rights impose disproportionate burden on certain groups. (IMO, the costs of restricting free speech almost always outweigh the benefits, but I can understand how she might come to the opposite conclusion in many situations.)

@andreamoro I noticed that tikz generated figures are not always being grabbed correctly. Will try to figure out how to fix this - we are using deterministic rules rather than AI to pull the images right now.

@Afinetheorem I'm having some trouble for it to recognize all figures from the pdf. I should probably upload them separately with the tex but they are in a subfolder, hard to select

@bruce_lambert We don't list any specific policies partly because we are also opposed to this - the audience is elected policymakers.

@Afinetheorem It’s the self-appointed savior tone that offends me. No thanks, economists and techbros. I’d rather have almost anyone else design the future than that group.

@STS_News @mattbeane As a self-contained piece, not sure. I have a short one coming soon that is student-readable which may help? Our survey data kevinbryanecon.com/ForecastingAI.… shows the much more modest actual beliefs of both AI lab folks and economists working in this area relative to, e.g., Dario.

Hey, @Afinetheorem and @mattbeane, I was reading this Compact piece "The AI Apocalypse Is Already Here" this morning. It's bad in predictable ways. But it made me wonder: Are there any best pieces by economists or sociologists of work that pick apart GenAI CEO's ideas about how their technologies are going to create mass unemployment? Kevin, you rightly told me that executives are making these claims because they believe them; it's just they don't know what they are talking about because the economics of labor markets are outside their area of expertise. But do we have a nice clear piece that lays that out? compactmag.com/article/the-ai…

@SBenzell That AI is developed to increase the utility of humans.

@Afinetheorem What do you think “complementing humans” in the letter means?

@amcafee @lugaricano @BrianCAlbrecht I prefer Andrew's as well, but not to the extent that I didn't sign. The existing beliefs about AI and the economy and the speed at which capabilities are growing are way too underestimated by policy folks. We have to correct this belief.

@lugaricano @lugaricano, @BrianCAlbrecht, thanks much. Your support for my edits means a lot to me.

(For example, I *don't* think only allowing AI as a "complement to humans" is a good idea. Self-driving cars that save 1.2 million lives, or AI doctors that can diagnose malaria, are clearly worthwhile despite failing that test. Open letter is not the venue for that discussion.)

As for what to do - many of us would disagree on this! Many more would think it presumptuous on epistemic grounds to claim we know. But society, org leaders, govts need to at least know the scale else you get rash policy following events (EU AI Act, recent DOD nonsense, etc). 2/2

@arindube This was why it was open. Too much uncertainty to suggest specific policies, but avg policymaker has WILDLY incorrect beliefs about pace of AI progress, many think it's all marketing hot air, and this is leading to terrible, rash policy when events happen (DoD, EU AI Act, etc)

I have no problems with economists signing this letter, but I personally would probably not have signed it if I were asked. Largely because it's very vague and I don't know what the concrete proposal is (besides that we should learn more, which goes without saying.) To be clear, and contrary to some letter-haters, I think the goal of regulating AI to ensure it is pro-human and pro-worker is *good.* And we should devise institutions that can effectively do so.