Seminars

I'll describe the principles underlying the development of Aristotle, an AI theorem prover trained via reinforcement learning on Lean proofs, that acheived a gold medal equivalent score on this years IMO problems, has been used to solve open problems and autoformalize new papers in real time.

*Abstract*

Abstract: How do you redesign a home screen beloved by millions, without losing the simplicity that defined it for over a decade? How do you introduce new ideas on a platform where small UI changes shape how 100 million people navigate their TV?

Abstract to be confirmed

Link to join virtually: https://cam-ac-uk.zoom.us/j/89473073451

*Abstract*

Memorization in LLMs has long been perceived as undesirable, associated with privacy risks, copyright concerns, and wasted capacity. In this talk, I argue for a complementary perspective: memorization is an intrinsic property of LLMs that can be leveraged to build a better LLM ecosystem.

In 2013 I wrote a book chapter on an imagined world where P = NP. A world with advances in medicine, translation, video recognition and generation, and much more.

Abstract: Almost everything you know about university coding projects – short duration, small teams, stable requirements and clean data – has no relevance to the majority of IT work in commercial organisations.