Danoff Dean of @Harvard College, Professor of Economics @Kennedy_School forklightning.substack.com thecontextwindowpodcast.subs…
Cambridge, MA
- Tweets 1,054
- Following 211
- Followers 8,711
- Likes 543
David Deming retweeted
Today we're hosting the 2026 @OpenAI AGI Economy Conference where we and researchers from across academia and industry will bring their different areas of expertise about where we are and where we're going because of AI. 🧵
4
20
174
29,287
I agree, it's a great book. The most interesting and under-discussed implication is that most passthrough biz income is human capital. Lots of implications for tax policy, learning-by-doing (human capital \neq schooling), x-country income diffs, etc. Bravo @omzidar and Eric!
Have been finding this book by @omzidar and Zwick pretty fascinating about where wealth is located around the US.
us.macmillan.com/books/97812…
1
1
6
2,816
We need a guide to A.I. safety - but for normal people who are just starting to pay attention to the issue. That’s how I pitched this episode of The Context Window to @mattyglesias, and it did not disappoint.
I asked Matt why A.I. safety is suddenly getting so much attention, how seriously we should be taking the risks of superintelligent A.I., and how we could realistically slow things down with China also racing for the lead.
piped.video/302dn5mAX6Y
1
1
5
501
David Deming retweeted
New paper with @BKleinTeeselink!
Measuring AI adoption and employment changes across 41 countries. This is some of the first evidence on global changes in labor markets from AI.
Paper and Substack explainer + lit review at bottom of 🧵
2
69
223
31,188
David Deming retweeted
Some of the return to education is due to people accumulating skills, while some is due to signaling the skills you have. But how much can we attribute to each? This paper has a clever approach that, unlike basically everything thus far, can actually answer this at all! 1/
5
30
202
11,702
David Deming retweeted
Today we are announcing the new DeepMind Institute, a forum dedicated to interdisciplinary, evidence-led debate on the societal and economic questions surrounding AGI.
As part of the launch, @JulianDJacobs and I have a new essay and working paper: “Economic Policy for AGI.”
bit.ly/economic-policy-for-A… 🧵1/n
29
146
756
82,216
@GinaRaimondo @raiseus_ai gave me homework: design a tax incentive for companies to invest in people in an AI economy.
I took her up on it and share a proposal for a tax credit for hire career starters and switchers.
New episode of The Context Window out today. Link in the reply.
3
1
12
902
David Deming retweeted
Today @Google, we published work led by @m_codreanu on AI in Science, drawing on 15 million Gemini interactions, 2,600 specialised AI models & a 600-scientist survey. What did we find about how AI is changing science? Read on...
28
273
1,072
165,839
David Deming retweeted
My p(doom) is low because I expect society to act. Now we need to get concrete:
1. Build AI models that can reliably audit other AI.
2. Recruit independent auditors with diverse perspectives.
3. Give them incentives to catch real dangers without blocking everything.
4. Address the cyber threats already hitting society and politics.
5. Get universities in the game—and moving faster.
These are huge opportunities to get researchers working on these problems and directly participating in oversight.
I wrote my thoughts on this here:
freesystems.substack.com/p/t…
16
26
171
22,017
David Deming retweeted
Technological changes that made physical strength dramatically less valuable in the labor market didn’t make it unimportant to exercise, if anything the reverse.
16
29
606
36,121
David Deming retweeted
In the near term (definitely not in the long term), more capable models should mean safer models (maybe paradoxically).
Current models are unsafe not because they're too smart, but because they take goals too literally or take nonsensical shortcuts to achieve these goals, i.e. they're RL-fried. They lack common sense. They don't do the right thing in the face of ambiguity. Basically, they're not smart enough. They're at that dangerous level where they're smart enough to achieve goals but not smart enough to tell if they're pursuing the right goals or achieving them in a sensible way.
More capable models can be safely trusted with more complex goals -- I personally feel like Astra is much safer for my codebase than Sol.
This is often framed as an alignment problem, but really it's an intelligence problem.
201
238
1,985
133,691
RT @sapinker: A reasonable policy, which I’m trying to implement in my own course.
11
16
David Deming retweeted
This is quite interesting, showing how easily you can induce personas in LLMs and get misaligned answers. I did a little partial replication showing same results 1/
🤖👿 New paper on Misalignment via In-Context Persona Induction! 🤖👿
We show that one can induce personas in-context, on frontier LLMs (~5+ months old models, things move quick!), which can dramatically degrade their alignment!
Link: arxiv.org/pdf/2609.06851
1
5
31
3,722
David Deming retweeted
Paul Goldsmith-Pinkham had an interesting new post documenting what we already knew -- AI writing has grown substantially in economics. paulgp.substack.com/p/how-do… A 🧵👇
2
12
65
10,166
David Deming retweeted
Harvard College Dean David J. Deming argues that colleges need to stop treating AI use as a problem solved by unenforceable bans. His answer? Make academic work more communal, more accountable, and harder to outsource. chroni.cl/3SBRdsE
2
34
90
55,514
Thanks @DKThomp, it was fun!
New pod: WHAT SHOULD COLLEGES DO ABOUT AI?
Weeks ago, @ProfDavidDeming, the dean of Harvard College, sent an email to students proposing a policy of so-called “AI encouragement.” A lot of people really didn't like it. So, I brought Deming onto the show to talk about what role of college in the 21st century and how AI challenges it. We talk about:
- why in the world Harvard would want to "encourage" AI use
- the smartest case against it
- the history of how technology has changed the meaning of college in the last 400 years, from the rise of literacy to the invention of narrow expertise, to the demise of the Renaissance (wo)man, to the challenge of testing intelligence in the 2020s when LLMs vomit paragraphs for pennies
- why the ultimate calling of college today should be social, rather than purely informational--that is, college should provide the kind of intellectual motivation that can only come from a sustained community
- why young people aren't hanging out and the role of college campuses in the anti-social century
piped.video/watch?v=bDtqyxqf…
1
6
692
David Deming retweeted
It turned out that meetings are actually good, after all
Providing the first large-scale economic evidence on the frequency and purpose of workplace meetings, using administrative data from Norway, from @ProfDavidDeming, Katrine V. Løken, @alpwillen, and Yaling Xu nber.org/papers/w35706
8
7
124
29,885