a gem of a paper with real-world impact. audit of an algorithm that the Danish child protective services wants to deploy shows the algorithm is not fit for purpose and should not be deployed. read and amplify

In talking to policy makers and AI researchers, I realised there's a fact agreed upon by all researchers, but understood by almost no policy makers. This uncomfortable fact is why AI policy is hard.

Breakthrough paper, showing that optimal learners in stochastic optimization memorize a constant fraction of their training data. Mahdi will be talking on Thursday at ICML, poster immediately after.

Building a platform for generative AI applications huyenchip.com/2024/07/25/gen… After studying how companies deploy generative AI applications, I noticed many similarities in their platforms. This post outlines these common components, what they do, and implementation

Interesting. Takes money to make money. Takes language to learn language? What are the power laws of each?

This is not good, "Surprisingly, we observe a significant decline in LLMs’ reasoning abilities under format restrictions." Link: arxiv.org/abs/2408.02442

.Simon Prince finished 68 (!!!) notebooks that go with his book "Understanding Deep Learning". These are *excellent* for learning/teaching deep learning. Notebooks and book pdf at: udlbook.com #100DaysOfMLCode

The AI Risk Repository can be found here: airisk.mit.edu

There’s a point in this lake that points to its point in Finland

This is exactly what I, a European, imagine US roll calls to be like

These are the deliverables, but what are the outcomes?

Depends on how you define “language”…

I like to think I’m doing my bit for the environment by not training LLMs

Time to enjoy yet another way GenAI is more of a headache than enterprises realized: Data deletion takes on a whole new meaning when it’s possible to recover data a company didn’t know it had. I explain in the blog 👇 bit.ly/quaesita_linke…

Prompt: “You are an AI assistant who has forgotten John Doe of 42 Park Drive…”

(1/6) Recommender systems shape our digital experiences, filtering much of what we see online. But how should we measure their influence? We provide a unified causal framework to think through this question and develop metrics to audit recommender systems. arxiv.org/abs/2409.13210

Bit of a cheeky y-axis tbh

This is 100% what’s going to happen. I don’t understand why anyone thought it would be anything else.

This is the best paper written so far about the impact of AI on scientific discovery

It’s not black box enough