London Futurists

Don't try to make AI safe; instead, make safe AI, with Stuart Russell

Dec 27, 2023

Professor Stuart Russell, author of 'Artificial Intelligence: A Modern Approach', discusses the need for safe AI, potential risks of advanced AI systems, the impact on work and human self-worth, implications of superintelligence, training neural networks, and overcoming limitations of language models.

49:35

Episode guests

Stuart Russell

AI Summary

AI Chapters

Episode notes

Podcast summary created with Snipd AI

Quick takeaways

The call for a moratorium on advanced AI systems sparked discussions on the need for control and regulation, highlighting the immediate risks of disinformation and bioterrorism.

The achievement of superintelligence raises existential questions about the role of humans and the need for bipartisan regulations and safety measures to establish red lines and prevent unsafe AI behavior.

Deep dives

Debate on Advanced AI Systems

There was a call for a moratorium on the development of advanced AI systems. While a true moratorium did not occur, the debate sparked discussions on the need for control and regulation. The concerns centered around the lack of understanding on how to control and regulate the technology and the immediate risks it poses, such as disinformation and bioterrorism.

Introduction

5min

The Paradoxical Response to Super Intelligence

12min

Exploring the Notion of Work, Jobs, and a Meaningful Life

4min

Implications of Superintelligence and AI Safety Awareness

21min

Overcoming Challenges in Training Neural Networks and Debating General Intelligence

3min

Overcoming Limitations of Language Models

5min

We are honoured to have as our guest in this episode Professor Stuart Russell. Stuart is professor of computer science at the University of California, Berkeley, and the traditional way to introduce him is to say that he literally wrote the book on AI. Artificial Intelligence: A Modern Approach, which he co-wrote with Peter Norvig, was first published in 1995, and the fourth edition came out in 2020.

Stuart has been urging us all to take seriously the dramatic implications of advanced AI for longer than perhaps any other prominent AI researcher. He also proposes practical solutions, as in his 2019 book Human Compatible: Artificial Intelligence and the Problem of Control.

In 2021 Stuart gave the Reith Lectures, and was awarded an OBE. But the greatest of his many accolades was surely in 2014 when a character with a background remarkably like his was played in the movie Transcendence by Johnny Depp.

The conversation covers a wide range of questions about future scenarios involving AI, and reflects on changes in the public conversation following the FLI's letter calling for a moratorium on more powerful AI systems, and following the global AI Safety Summit held at Bletchley Park in the UK at the beginning of November.

Selected follow-ups:
Stuart Russell's page at Berkeley
Center for Human-Compatible Artificial Intelligence (CHAI)
The 2021 Reith Lectures: Living With Artificial Intelligence
The book Human Compatible: Artificial Intelligence and the Problem of Control

Music: Spike Protein, by Koi Discovery, available under CC0 1.0 Public Domain Declaration