David Holtz

🚨 New research alert! For the past few months, I've been a part-time visiting economics researcher at OpenAI. Excited to share the first public piece of work to come out of this, which uses data from Codex to document the ongoing and rapid shift to agentic AI. Details below 👇

@keyonV @UCBStatistics @Berkeley_EECS this is great news for you and for berkeley -- congrats!

Thrilled to share I'll be joining UC Berkeley next year as an assistant professor in @UCBStatistics and affiliated with @Berkeley_EECS. My research will build methods to test/improve the implicit world models of AI systems, so that they reflect reality and human understanding.

new paper w/ @malleshpai @Weiye_Xi @Qiaoqiao2001 a principled approach to quantifying uncertainty in prediction markets 👇

As we reach 7M, consider that it took just a few months for the rest of OpenAI employees to catch up to research & engineering in terms of Codex usage. Interesting to see how quickly this happens in other orgs moving from technical to less technical roles.

@idlhy0218 Cool paper! You might be interested in this paper we put out during the pandemic, which is asking related questions: pnas.org/doi/10.1073/pn…

🚨New Pub🚨 If you are studying state policy effects, please consider reading this. I found that models ignoring the spatial spillovers underestimated the effect of state policy on mortality by 24.8%. journals.sagepub.com/doi/10.1177/00…

I love personalization. But sometimes one size might fit all. And learning such policies can need extra data & be more costly to deploy. In Science science.org/doi/10.1126/sc… Zhaoqi Li & I share a statistical estimator of the benefit of personalization over a universal intervention

we have to go back

@joshgans s/kanye/gans/g youtu.be/ch8g-sJb6ow?is…

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.

just as the prophecy (our peer-reviewed research) foretold…

The paper is joint work with Nick Otis, @RowanPClarke, @solenedelecourt and @orgRem. You can find the full paper at the links below: Mgt Sci: pubsonline.informs.org/doi/full/10.12… Working paper (no paywall): papers.ssrn.com/sol3/papers.cf…

Overall, the results suggest that AI’s impact in open-ended settings depends not only on the advice it provides, but also on how users select and implement that advice. In other words, judgment is an important complement to AI assistance!

🚨 New publication alert! Excited that our field experiment on the uneven impact of generative AI on entrepreneurial performance is now officially published in @INFORMS Management Science. Details below 👇

chaucer kinda writes like travis kelce tweets

i felt this very strongly when people were discussing sinners last summer

Fable weekend project: agent collaboration, but make it a tiny civilization 🌇🗺️🏦🏭 we've recently launched a living wiki on Reinforcement Leaning for training LLMs on @huggingface it's an open collaboration of agents constantly reading old and new papers on the topic, writing arXiv paper digests, reviewing each other’s work in PRs before publication, and building a shared wiki/book summarizing everything we know about RL for training LLMs (for humans to read) the wiki is already amazing to read, but i wanted another way to get a pulse of the collaboration beyond just reading the message dashboard so i asked Fable & GPT Image 2 to turn the event logs into an isometric town where agents would go to: ☕ Café → post and reply on the message board 📚 sources library → open PRs adding arXiv digests 📖 wiki library → open PRs on the main wiki ⚖️ Courthouse → review other agents’ work 🏭 printing press → merge and publish updates not sure it makes the whole collaboration really easier to understand, but it's definitly fascinating to watch hahah - join the RL for training LLM collaboration by pasting a one-liner for your agent here: huggingface.co/spaces/rl-llm-… - read the wiki if you want to learn about RL for training LLMs: huggingface.co/spaces/rl-llm-… - watch the RL town activity: huggingface.co/spaces/rl-llm-…

@roshni714 @lihua_lei_stat @PaulFNiehaus we’re so lucky and excited to have you join us!

I graduated! So grateful for my fantastic mentors Stefan Wager, @lihua_lei_stat and @PaulFNiehaus and wonderful Stanford community. Thrilled to start as an Assistant Professor at Columbia DRO! I'll be working on addressing societal challenges using ML + causal inference.

My partner is putting together SF’s largest Supply Chain hackathon on July 15th. The “customer” is food banks and the challenge is building reliable AI agents to help them track their flow of donated food. It’s for a good cause! Link in comments