New working paper by Benjamin Manning & Kehang Zhu that & I’m super excited about benjaminmanning.io/files/rs.pdf

How well can we understand an LLM by interpreting its representations? What can we learn by comparing brain and model representations? Our new paper highlights intriguing biases in learned feature representations that make interpreting them more challenging! 1/

Don’t Believe the AI Hype by Daron Acemoglu - Project Syndicate Summary of my new paper on macroeconomics of AI project-syndicate.org/commentary/ai-…

A sequential experimentation model with endogenous feedback, from Roland Bénabou and Nikhil Vellodi nber.org/papers/w32483

General intelligence is *precisely* learning -- the ability to efficiently learn new things, beyond what your genes and past experiences prepared you for. Current ML has near zero intelligence because static inference with a curve only yields local generalization, with zero

New #RLC2024 paper Three Dogmas of Reinforcement Learning joint w/ Mark Ho and Anna Harutyunyan | Աննա Հարությունյան! arxiv.org/pdf/2407.10583 We reflect on where our scientific paradigm needs adjustment, and suggest three departures from previous conventions. Curious to hear what folks think! 🧵

This changed everything for me Arthur, W. B. (2021). Foundations of complexity economics. Nature Reviews Physics, 3(2), 136–145.

Does astrology work? We tested the ability of 152 astrologers to see if they could demonstrate genuine astrological skill. Here is how the study was designed and what we found (including a result that really surprised me): 🧵

A good heuristic about whether something can in principle be solved with deep learning: can it be done by an expert human 100% intuitively, without having to consciously think about it? If yes, then pure pattern recognition is an appropriate approach to the problem, and given

Hi, fdf left us. Irish goodbye. Gone for now, but you never know. His factor model implementation is here: github.com/0xfdf/toraniko I am still weakly in touch (i.e., can email, not sure at the other end there's a human, a AI or a hyper-intelligent dog). I am grateful he

🚨Thrilled to share our new paper "Causal Claims in Economics"! 🚨 Thiemo Fetzer 🇪🇺🇺🇦 - same handle elsewhere and I analysed over 44,000 economics papers using AI to create a knowledge graph of economics and map out causal relationships. Here's what we found 🧵👇

When we develop AI systems that can actually reason, they will involve deep learning (as one of two major components, the other one being discrete search), and some people will say that this "proves" that DL can reason. No, it will have proven the thesis that DL is not enough,

From prediction markets to info finance vitalik.eth.limo/general/2024/1…

Worth checking out:

Great explainer of some complex yet important topics:

“This work provides a foundation for new tools that can help investigate individual and collective behavior”

We still have a relatively poor understanding of the relationship between research and policy. Program evaluation in particular is often motivated by a desire to make policy better. But how effective is program evaluation itself? Michelle Rao's JMP tackles this question.

Some of the most important lottery anomalies from the behavioral risk literature (e.g., loss aversion) actually have nothing to do with risk. They also arise in perfectly deterministic settings.

Super cool

vibe coding seems great on a first date. but does the relationship ever last?