Valentin Haddad

What happens when we use AI for research and the reviewing? Let's find out. Enter Human × AI Finance: write a finance paper with AI in 4 weeks. Every submission reviewed by AI agents. Top 4 presented at @uclaanderson. Everyone can submit. Deadline 3/18: humanxaifinance.org

Only 3 days left to submit to Human x AI Finance... I (and AI) look forward to see what you come up with! humanxaifinance.org

So long, Chris. A giant of economics. Always so generous with his time and willing to engage. Every conversation we had (many on NJ Transit) left me marveling at how someone could be so deep yet so straightforward.

Why?

Princeton Conference on Asset Demand Systems to better understand the equilibrium relation between portfolio holdings and asset prices on May 14-15, 2026. Organized with @rkoijen and Liran Einav. Conference program now posted: dsap.princeton.edu/index.php/prog… #DSAP

Only 8 days left to submit your work for the Human x AI Finance conference! ⏳ That's plenty of time to give heavily AI-assisted paper writing a shot. 😉 To answer a frequent question we've been getting: the "x" is meant to be multiplicative, not vs. Why submit? - help discover what the future of research looks like. - 🏆 $1,000 prize for the best paper. - ✈️ The top 4 papers will be invited for presentation at the Fink Center Conference on Financial Markets at UCLA Anderson School of Management (travel and lodging included). 🔗 humanxaifinance.org #ArtificialIntelligence #Finance #FinTech #Research #CallForPapers #UCLA

I think it would be better if there were fewer grand claims about how AI is going to revolutionize research and more people trying to make the best use of AI to improve their research, and let’s see what we get.

Nice piece! The key question seems to be how much harder it will be to distinguish paper quality (at the top of the distribution). An optimistic view is that we will publish the same number of papers, just much better than before. It could be that the increased mass doesn't even get written/reviewed because of clearly lower quality (as it is already without AI) in "steady-state".

Thanks to @causalinf for writing this very thought-provoking piece and please send all the hard questions to him and not me 😃

Every journal editor should read this: causalinf.substack.com/p/claude-code-…

I wrote an academic paper in 4 days using AI. Claude Code wrote the first draft overnight. I iterated with multiple AI agents as reviewers. AI made me faster, not smarter. The paper: "Investing in Artificial General Intelligence" The full story: vincent.codes.finance/posts/vibe-res…

Deadline of 2/27 to submit to the Conference on Asset Demand Systems. Submit online at dsap.princeton.edu/index.php/subm…

To incentivize making your paper submission to Human x AI Finance as good as possible using AI, we will now also offer a best paper prize of $1,000 for the paper which, according to the AI reviewers, was the best submission! humanxaifinance.org Submit by March 18th!

Let's raise the stakes (a little bit): best paper according to the AI reviewers will also receive a $1,000 award! Update your claude md with this extra incentive, and get back to research. humanxaifinance.org

@achenfinance @GeoffreyHuntley submit your final output at humanxaifinance.org. Let's see how it stacks up against other human and AI processes.

Been working on using the Ralph Wiggum method (@GeoffreyHuntley) to write papers and respond to reviews. Opus improves the paper, Opus reviews, and we repeat until the paper passes the reviews. Really not sure if this is productive but it sure is entertaining.

@arpitrage Maybe the best analogy (though somewhat wrong) is that they work more like Bayesian statistics, the "point estimates" are actually distributions.

@mktmacrostruct Hmm that seems more akin to this strand of the literature; taking point estimate LLM outputs and doing statistics on those? x.com/arpitrage/stat…

As far as I can tell, LLMs use many statistical optimization methods (probabilistic modeling of next token prediction, gradient descent, etc) But they don't have standard errors or uncertainty quantification. It's all point estimates. This is apparently sufficient to learn

We have now written a longer piece explaining the background and motivation for launching a finance conference that encourages both AI-written submissions and selects papers using AI: augmentedreturns.substack.com/p/launching-th…

What happens when we encourage AI use for research and use AI to review papers? We are running an experiment to find out: the UCLA Human × AI Finance conference! Write a finance paper with AI in 4 weeks (by 3/18). AI agents review the submissions: humanxaifinance.org 🧵 1/

@MGoFinance just try ;)

🤔

@uclaanderson And this might be the very first academic conference 100% committed to take the papers chosen by AI.

@uclaanderson Honestly, I have no idea what we'll get. Maybe the papers will be great. Maybe they'll be terrible. Maybe the AI reviewers will be as (un)rigorous as humans. That's the whole point — we genuinely don't know, and I think that's more interesting than pretending we do.

What happens when we use AI for research and the reviewing? Let's find out. Enter Human × AI Finance: write a finance paper with AI in 4 weeks. Every submission reviewed by AI agents. Top 4 presented at @uclaanderson. Everyone can submit. Deadline 3/18: humanxaifinance.org