London Futurists

The shocking problem of superintelligence, with Connor Leahy

Oct 25, 2023

Guest Connor Leahy, German-American AI researcher and entrepreneur, discusses the balance between excitement and concerns surrounding new technologies, the potential outcomes of superintelligence, and the challenge of controlling an entity that is much smarter than humans. The importance of addressing AGI as a problem for its beneficial impact is emphasized, along with the need for regulations, safety measures, and raising awareness to address the challenges of artificial intelligence.

43:51

Episode guests

Connor Leahy

AI Summary

AI Chapters

Episode notes

Podcast summary created with Snipd AI

Quick takeaways

Governments and policymakers need to enforce regulation to ensure the safe and beneficial development of superintelligence.

Public awareness, knowledge dissemination, and collective action are essential in addressing the risks associated with superintelligence and AI.

Deep dives

The urgency of addressing superintelligence and AI risks

In this podcast episode, the urgency of addressing the risks associated with superintelligence and advanced artificial intelligence (AI) is highlighted. The guest, Connolly, discusses his interest in existential risks and the need to ensure that the development of superintelligence is controlled to benefit humanity rather than lead to disaster. The episode focuses on the upcoming Global AI Safety Summit and the importance of discussing both short-term risks like privacy and bias, as well as long-term risks associated with control and alignment of superintelligence. Connolly emphasizes the need for policymakers and governments to step in and enforce regulation to manage these risks effectively.

Introduction

3min

The Excitement and Concerns of New Technologies

5min

Capabilities of GPT-4 and Shorter-Term AI Risks

7min

The Potential Outcomes of Superintelligence and the Burden of Proof

2min

Competition for Resources and the Importance of Control in the Face of Super Intelligence

5min

The Challenge of Controlling Superintelligence

21min

This is the second episode in which we discuss the upcoming Global AI Safety Summit taking place on 1st and 2nd of November at Bletchley Park in England.

We are delighted to have as our guest in this episode one of the hundred or so people who will attend that summit – Connor Leahy, a German-American AI researcher and entrepreneur.

In 2020 he co-founded Eleuther AI, a non-profit research institute which has helped develop a number of open source models, including Stable Diffusion. Two years later he co-founded Conjecture, which aims to scale AI alignment research. Conjecture is a for-profit company, but the focus is still very much on figuring out how to ensure that the arrival of superintelligence is beneficial to humanity, rather than disastrous.

Selected follow-ups:
https://www.conjecture.dev/
https://www.linkedin.com/in/connor-j-leahy/
https://www.gov.uk/government/publications/ai-safety-summit-programme/ai-safety-summit-day-1-and-2-programme
https://www.gov.uk/government/publications/ai-safety-summit-introduction/ai-safety-summit-introduction-html
An open event at Wilton Hall, Bletchley, the afternoon before the AI Safety Summit starts: https://www.meetup.com/london-futurists/events/296765860/

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