Introducing PaperQA2, the first AI agent that conducts entire scientific literature reviews on its own. PaperQA2 is also the first agent to beat PhD and Postdoc-level biology researchers on multiple literature research tasks, as measured both by accuracy on objective benchmarks

It's a bit sad and confusing that LLMs ("Large Language Models") have little to do with language; It's just historical. They are highly general purpose technology for statistical modeling of token streams. A better name would be Autoregressive Transformers or something. They

1/ New paper nature! Discrepancy between human expectations of task difficulty and LLM errors harms reliability. In 2022, Ilya Sutskever Ilya Sutskever predicted: "perhaps over time that discrepancy will diminish" (youtu.be/W-F7chPE9nU, min 61-64). We show this is *not* the case!

Today Professor Dennis Lo Yuk-ming is appointed as the ninth Vice-Chancellor and President of #CUHK for a period of five years from 8 January 2025. Professor Lo is globally renowned as a pioneer, innovator, and leader in liquid biopsy. Details: bit.ly/3zqTTjc

Nearly 50% of researchers quit science within a decade, huge study reveals go.nature.com/3YaQTkr

Congratulations to @UCL alumnus Sir Demis Hassabis on being awarded the #NobelPrize in Chemistry. The Google DeepMind co-founder & CEO, previously based Gatsby Computational Neuroscience Unit and UCL Queen Square Institute of Neurology, is honoured for using AI to predict the structure of almost all known proteins ucl.ac.uk/news/2024/oct/…

"It's a multilingual embedding model"—okay, but what's that mean? Does it mean you can search German docs with a German query, French docs with a French one, all within the same model? Or can you also search French docs using Japanese queries? A lot of these multilingual models

TL;DR: LLMs don’t reason.

Technion, Google Research and Apple recently investigated the problem of LLMs' hallucinations. They looked at how LLMs make mistakes by examining what happens inside them. Here are the key findings:

New preprint! Brains are spatially organized. Most LMs are not. We induce brain-like topography in Transformer LMs using TDANN spatial smoothness. paper: arxiv.org/abs/2410.11516 w/ Hannes Mehrer Badr AlKhamissi Taha Binhuraib 🦉 Nick Blauch Martin Schrimpf

"Large language models surpass human experts in predicting neuroscience results" w Xiaoliang (Ken) Luo and BrainGPT.org. LLMs integrate a noisy yet interrelated scientific literature to forecast outcomes. nature.com/articles/s4156… 1/8