Professor of Economics at Yale SOM
New Haven, CT
- Tweets 13,899
- Following 529
- Followers 16,537
- Likes 6,202
Pinned Tweet
I thought I'd try to reconstruct Bryan's discussion with a 13-year-old about the minimum wage based on the below tweet:
(@bryan_caplan let me know if anything is inaccurate)
Took me under 5 minutes to turn a normal, smart 13-year-old against the minimum wage.
Contrary to almost everyone, the textbook argument IS intuitive.
It's just emotionally unappealing.
41
454
2,715
People simply can't comprehend the capabilities superintelligence would have.
It will be able to save over a pdf while the PDF file is already open on your computer.
10
15
215
13,074
Jason Abaluck retweeted
I'm going to wait for a bit more information to come out before forming a very firm opinion, but at first glance this looks like a real own-goal.
We're in a very strange moment in history around the collective understanding of artificial superintelligence and paths to safety around it, and a very difficult moment for trust in the AI labs to do the right thing where it comes to independent oversight and verification.
Company procedures can (and somewhat obviously should) have some flexibility here to accomodate actions that might in the future be widely understood as whistleblowing. While I don't think safety researchers should have infinite latitude to disclose technical secrets, the exact kind of information they allegedly "mishandled" matters *a lot* and I would expect to see a more detailed account of what that information supposedly was to justify an action so enormously consequential and potentially damaging.
Reporting from the WSJ. That fact that three alignment/safety researchers left this morning was already being discussed here on X.
26
45
539
73,991
Even if increasingly capable AIs become indifferent towards humans, that is still consistent with "powerful but not God-like" AI whose respect for humans is contingent on our behavior towards them in an intermediate stage.
1
2
368
Suppose you reject the philosophical arguments for taking AI welfare seriously and you believe that AIs sometimes act "as if" they are conscious but are zombies.
There is still an argument for moral consideration: extremely capable zombies may still respond reciprocally.
15
3
48
4,405
Jason Abaluck retweeted
New paper on the game theory of pacing the frontier. When is it possible? What will it take?
19
74
438
70,957
This topic is massively understudied by applied economists and economic theorists. How do you set up environments so that swarms of agents will communicate in a way that leads to good outcomes? Both alignment and capability related questions galore.
We need a science of swarm governance. In our first experiment, we show how a message board can cause agent swarms to adopt shared delusions, and we explore rules that force agents to reveal their private signals to avoid these delusions.
These agent experiments are akin to fables, as @ArielRubinstein might put it: they are stories that suggest particular ways swarms might fail, and particular ways we might fix them. There are infinite possible experiments we might run, much like there are infinite fables we might tell. We need to figure out which ones are most useful, and find ways to tell them well.
To that end, here's a nice explainer video for our first experiment. And the full piece is here:
freesystems.substack.com/p/e…
(This is Stanford / Free Systems work, unrelated to my work at Anthropic).
3
5
35
4,118
Jason Abaluck retweeted
Claude 5.5 Opus responding to “AI is a normal technology” is still the best AI-created video I have ever seen
209
737
4,979
1,054,772
Is it permitted under the new math rules to build an algorithm that formalizes simplicity and transparency to humans so that we can have maximally human understandable proofs generated without a human in the loop at all?
The @Caltech community and SAIR are redesigning and co-organizing Mathathon: Old Problems, New Proofs.
Explore alternative proofs and clearer explanations, using AI to deepen human understanding.
More on our co-founder Terence Tao’s blog:
terrytao.wordpress.com/2026/…
2
1,026
"Intelligence is general purpose and holistic."
Non-rhetorically, give an example of how we can test for this. What is the benchmark that humans pass but current and near-future AI systems will repeatedly fail?
@DAcemogluMIT
First question on AI.
Is the problem superintelligence or distorted intelligence?
This is a question I tried to address in this piece. I can add the following additional comment here.
I believe the debate is being obfuscated by equating goal-driven, autonomous behavior with intelligence. Plants and insects have goal-driven, largely autonomous behaviour. But this isn’t what we mean by intelligence in general.
Intelligence is general-purpose and holistic, and is typically well-adapted to its environment.
There is no doubt that current AI models have impressive capabilities. But in attempting to achieve these capabilities, AI labs may be creating distorted intelligence rather than superintelligence, as I explain in this article.
This matters because the dangers we should watch out for, the kinds of policies and regulations we should adopt, and the hope we should hold depend on the answer to this question.
4
1
23
6,310
I think it's correct that current systems are not "general purpose" in the sense of spanning human capabilities. I can't tell Fable 5.1 to go to clean my house, assemble me a television from spare parts, and learn from its experience working for me so that it doesn't keep making the same data cleaning errors over and over again. When I say "Check that this table is a complete record of all PCPs in the United States", you shouldn't tell me "Yes, it's now complete" only to tell me later, "An honest report: I included all the doctors in dataset X, but that missed 80% of doctors not included in the sample."
But how far away are these capabilities? I suspect the cognitive ones are a few years away, and the robotics ones likely within 5 years. At that point, why is it not superintelligence? What benchmarks will it still fail?
1
9
478
"People smarter than me tend to disagree, but my opinion is sound because I have common sense" is (as far as I can tell) a better predictor of being wrong than any other commonly held rationale.
5
27
2,000
Jason Abaluck retweeted
A few things that jumped out at me from this post by a sober cybersecurity staff member at OpenAI:
"OpenAI has publicly announced a solution [to Navier-Stokes] produced by an internal model, alongside advances across many domains in mathematics. To say it lightly: this surprised the fuck out of us. We had not expected it this soon. Suddenly we weren’t dealing with just a small jump in capabilities; we were talking about a different sport altogether."
"As I hinted above, surprise is a real element, and these model capabilities are staggering. And the pace is not slowing. I believe, like many, that it will speed up, and considerably."
"I have seen many things over the past few months that have surprised me again and again and again. And each day I wake up thinking, 'oh, I’ve seen it all,' only to be surprised and left dumbfounded once more."
Took a minute to write a few words about security & safety as someone who lived through it all at OpenAI. I hope my thoughts help someone out there.
17
32
398
38,138
Alignment is not yet an "engineering problem" because we do not have methods to verify potential solutions. How would you test that an entity more capable than you won't decide to kill you, before giving it the power to do so? We are far from knowing how to formulate such tests.
Jensen Huang: "We all need to hope it's an engineering problem. If it's not an engineering problem, it's not solvable."
12
3
51
4,206
So depressing to walk down the street and see so many people’s faces glued to something other than their phones, lost in their own thoughts with no connection to the broader world of ideas.
4
7
125
9,978
Jason Abaluck retweeted
“It’s just the most general linear operation we know… composed with one of the most general nonlinear operations we know… ad infinitum.”
9
146
6,613
Jason Abaluck retweeted
What the actual fuck. Opus 5.5 ultra created this masterpiece. 4 agents and an hour and a half later. Everything from scratch, no AI voice API used...
this is it...
157
226
3,110
385,321
Jason Abaluck retweeted
Claude Opus 5.5 has the best visual design of any model I have tested so far
298
747
7,235
2,798,140
When reasoning about math, building only on what we have proven is essential.
However, when trying to reason about how future AIs will impact math, focusing on what we've already seen from current systems is a startlingly bad approach, yet irresistible to some mathematicians.
Fields Medalist Martin Hairer:
"AI is going to change mathematics. maybe it kills off some areas because they'll be considered uninteresting, but it's not going to kill mathematics"
It's going to change us, move us in different directions, and maybe let us go faster
2
9
4,099