🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

AI models appear to mimic the real world. But how can we tell if they truly understand it? Excited to announce the ICML 2025 Workshop on Assessing World Models! Working on related questions? Submit a paper (max. 4 pages) by May 20!

“In the U.S., unfortunately, most people cannot afford a lawyer. What I'm currently working on is legal natural language processing and NLP tech, which would increase access to justice.” A new Meet the Researcher w/Dominik Stammbach is now up ➡️#CITP Blog: blog.citp.princeton.edu/2025/04/16/mee…

Why is bureaucracy so complex and regulations so hard to navigate? Why are uni administrations making things ever more wasteful? The answer is that complexity can be good, but this fact itself creates perverse incentives to over-complicate things. 👇 1/n

Delighted to host Dominik Stammbach as a Postdoctoral fellow in my group! He's being doing some amazing work on real-world AI for the public interest, particularly in the legal domain!

In our new ETH Center for Law & Economics vlog episode, Daniel Hemel NYU Law & Luca Baltensperger ETH Zurich discuss Prof. Hemel’s study "Law and the New Dynamic Public Finance", which challenges conventional wisdom in classical tax theory & offers novel perspectives. Watch it here: youtu.be/pgXqP1ky2qU

Wikipedia should implement a global LLM audit. Have LLMs review every page for errors, propose corrections, and estimate Q = P(correction is better). Initially, human editors review any change. If Q is validated by humans, automatically update for Q close to 1. Jimmy Wales

As expected, that was popular. Here is my attempt at consolidating all the answers into a list. - Prenorm: normalization in the residual blocks before the attention operation and the FFN respectively - GQA (Group Query Attention): more Q than (K, V)

Reminder: The Swiss Journal of Economics and Statistics invites submissions for the collection #Textual #Analysis in #Economics and #Finance (link.springer.com/collections/cd…), edited by Daniel Kaufmann, Elliott Ash, Markus Leippold and Leif Anders Thorsrud. We welcome original research

1/6 Can AI-powered “enforcement bots” substitute for class actions? It turns out yes likely leading to major disruptions in the legal system.

Want to study the economic impact of AI and influence the policy choices a frontier lab makes? I'm building a team to advance Anthropic 's Economic Index & other ~special projects. Lots of fun! Economist: job-boards.greenhouse.io/anthropic/jobs… Data Scientist: job-boards.greenhouse.io/anthropic/jobs…

1/6 Are robot judges perceived as less fair than human judges? Yes, but minimal human oversight can eliminate this human-AI fairness gap. Check out our new Working Paper “Mitigating the Judicial Human-AI fairness gap.”

Using machine learning to detect corruption Last month on VoxDev, Elliott Ash (ETH Zurich), S Galletta (ETH Center for Law & Economics) & Tommaso Giommoni (UvA Amsterdam) discussed how machine learning can improve corruption detection and prevention in Brazil: voxdev.org/topic/methods-…

Wow! Check out this cool working paper. "War Causes Religiosity" The authors scraped web images of hundreds of thousands of gravestones of deceased Americans. They found that people who were randomly drafted into war are at least 20% more likely to have religious gravestones.

There are so many hallucinated citations in court nowadays, that I'm starting to put together a tracker. Check it out and feel free to send ones that I've missed along. New tabs coming for more categories of AI+Law cases!

In India, judge in-group bias by gender, religion, & caste is minimal. Just Accepted new paper by Elliott Ash, Sam Asher, Aditi Roy Bhowmick, S. Bhupatiraju, D. Chen, Christoph Goessmann, Paul Novosad, Bilal Siddiqi zurl.co/NXXnA