@AdamThierer Yeah but that's really too reduced form for specifics. The current AI regulation being pursued ad hoc has not been analysed properly.
English
Joshua Gans
43.5K posts
Joshua Gans
@joshgans
Professor at the Rotman School of Management, University of Toronto. Chief Economist of Creative Destruction Lab https://t.co/a9ZbnBauCF
Toronto, Ontario Katılım Kasım 2008
753 Takip Edilen20.3K Takipçiler
well, for starters, here's a link to one by an economist I think highly of. 😉
there are others, but need to classify what fits in the range of "regulatory economics of AI," because much of the best work is going to be more issue / sectoral-focused (labor economics, pricing, transportation econ, fintech & AI, etc) which make sense given the breadth of AI. I'd welcome more broad-based work, but the real action is going to be in that more targeted literature on "The Economics of AI & [fill in the issue / sector]"
nber.org/papers/w32741
English
Many very impressive names on this letter to “act now” to better understand the transformative effects of AI (as all the signatories on it are already doing!) But I don’t see any specifics here. Just aspirational STUDY-IT-EVEN-MORE! vibes.
Do we really have a shortage of folks looking into AI these days?? In my 35 years of doing tech policy, I have never witnessed so many people, pundits, policymakers, philanthropists, and professors investigating any other issue as much as the transformative effects of AI are being studied currently.
Look, I’m all for more study, but the “ACT NOW” part of this implies policy activism of some sort “to steer AI in a direction that complements humans and benefits society.” Someone needs to define for me what they mean by that before we can have serious conversation.
Erik Brynjolfsson@erikbryn
Here's our statement on AI and the economy. We Must Act Now A Statement on AI’s Transformation of the Economy 1. AI may become radically more powerful over the next 10 years. 2. This could drive an unprecedented transformation of our economy, larger than the Industrial Revolution, but unfolding over a vastly shorter time frame. It could bring risks, including large-scale job displacement, as well as opportunities such as major gains in living standards. 3. Economists, policymakers and technology leaders must act now to understand the economics of transformative AI and to build the incentives, guardrails, and institutions needed to steer AI in a direction that complements humans and benefits society.
English
Joshua Gans retweetledi
This specification from @joshgans is going to dominate everything. The task of orchestrating human inputs and evaluation, alongside embodying it in an orchestration of agents, is enormous. There are surely, surely, very few pre-AI organisations that are up to it.
English
There is lots of criticism of the wemustactnow.ai letter on AI to which I am a signatory to because of lack of specifics. You know why? As far as I know there has not been a single paper in regulatory economics on the details of regulating AI to counter potential risks.
English
we will soon have a research field I am calling "AI Management" where we will study how AI systems manage other intelligent entities - including other humans, models, robots etc.
if these results translate to AI managers, then it suggests that a lot of the gains from "smarter" models will come from task allocation and matching rather than improved performance on the task itself.
We need benchmarks for this! We're launching something this week that is somewhat in this vein, but not explicitly.
Would be curious to see more takes on this topic!
Alexander Berger@albrgr
Very cool evidence on the role of good managers in improving productivity
English
Lots of papers on whether we should stop AI or not. Lots on broad governance ideas. But literally none on who should get access to frontier models and why. This is why this letter is important. I just wish it had pointed this out more clearly.
English
@joshgans Isn't it functionally equivalent to Ultra in Codex? Maybe @reach_vb or @Dimillian can clear up
English
This is actually the founding principle of All Day TA. When professors provide content, it is used to train their TA and nothing else. We don't even store it (even when some profs want us to), and we can't use it to train our own backend.
Satya Nadella@satyanadella
English
@ahall_research I think that people are concerned more about reputational damage than we think AND those that aren’t produce stuff that is more obviously easy desk rejects. In my experiments, the optimal amount of AI is well below zero cost paper production.
English
So interesting -- I wonder what forces are helping us avoid the slopocalypse in journal submissions so far?
Ben Ansell@benwansell
@ahall_research Great points Andy. As an editor of a major-ish polisci journal with a very international submission base I would say the slopapocalypse is not yet nigh. Clearly some AI created papers - very obvious and very poor. Submissions up by maybe ten percent.
English
Good to have you back! But does this mean I may get ‘old-Gans’ as my R2 at some point?
Joshua Gans@joshgans
After a year of getting the AI to write more of the words in my academic papers and not simply help with the maths, I have had feedback that people actually miss the 'old-Gans.' Given decades of R2s not liking my writing style this was quite heartening. Changing my approach now.
English
Joshua Gans retweetledi
Looking forward to reading in detail - but lots of great insight here!
One key point most evals miss / if it’s cheap to verify, it’s ok for lots of variance in output quality (you might even want it!) - at the moment our assessments of AI “quality” are completely divorced from how humans will use it.
Kevin A. Bryan@Afinetheorem
I have a fun new paper today w/ @joshgans: what makes an AI valuable? We noodled on this literally since 2024. Answer: AI is used by humans. They can extend, verify, get a second opinion, etc. AI is therefore part of what decision theorists call a "composite experiment". 1/10
English
After a year of getting the AI to write more of the words in my academic papers and not simply help with the maths, I have had feedback that people actually miss the 'old-Gans.' Given decades of R2s not liking my writing style this was quite heartening. Changing my approach now.
English
Joshua Gans retweetledi
I was clearly wrong about Anthropic. They are obviously currently the leader in AI. No company has released a model as good as Mythos/Fable and they will undoubtedly have Mythos 2 ready soon.
And I would never cut them off in a way that hurt them badly, even as a competitor. That’s not my style.
Tesla open sourced its patents and we made the Supercharger network available to all competitors, even though we could have made it a walled garden.
SpaceX launches competing satellite systems with no increase in price or use of unfair terms.
Even my worst enemies can attack me on this platform.
…
English
Here is a nice interactive explanation of my latest paper with @Afinetheorem Kevin coded this one up. I hope we can find a way to submit it with our paper. kevinbryanecon.com/AIandhumanchoi…
English