Abstract: I review arguments for and against the possibility that AI(Artificial Intelligence). and LLMs (Large Language Models) develop expert-level capabilities. This could entail a future where human experts have AI-assistants and eventually become validators of AI-crafted solutions. I describe some implications of strongly-capable AI for safety and alignment. (This talk was presented at the CITA Jamboree 2024.)
A path towards an AI Physicist?
Kristen Menou (University of Toronto) // October 4, 2024
Abstract: I review arguments for and against the possibility that AI(Artificial Intelligence). and LLMs (Large Language Models) develop expert-level capabilities. This could entail a future where human experts have AI-assistants and eventually become validators of AI-crafted solutions. I describe some implications of strongly-capable AI for safety and alignment. (This talk was presented at the CITA Jamboree 2024.)