Future of Life Institute Podcast

Roman Yampolskiy on Objections to AI Safety

May 26, 2023

01:42:13

Creator website

Episode guests

Roman Yampolskiy

Highlights

AI Chapters

Episode notes

The Risk of Failure for Superintelligent Systems

01:26

Setting High Standards for AI Systems

01:39

Introduction

2min

The Objection to AI Safety Research

2min

The Limits of Intelligence

2min

The Importance of Consciousness in AI

2min

AI's Will Be Tools for Us

2min

The Pros and Cons of Shutting Off AI

2min

The Limits of AI Monitoring

3min

The Pros and Cons of Artificial Intelligence

2min

The Double Standard of AI

2min

10.

The Drive Towards Self-Preservation in AIs

2min

11.

The Evolution of Artificial Intelligence

2min

12.

How AI Is Planning to Do More of What We Want

2min

13.

The Future of Language Models

2min

14.

The Future of AI

3min

15.

The History of AI Safety

2min

16.

The Future of Artificial Intelligence

2min

17.

The Impossibility of Safely Aligning AI

5min

18.

The Impossibility of 100% Accuracy in Cyber Security

2min

19.

The Importance of AGI in Life

2min

20.

The Impossibility of AI Alignment

2min

21.

The Impossibility of Accepting Existential Risk

3min

22.

The Immorality of Consent in Scientific Experiments

2min

23.

The Importance of Long-Term Planning

2min

24.

The Consequences of Increased Intelligence

2min

25.

The Effective Acceleration Is

2min

26.

The Pro-Human Bias

2min

27.

AI and the Dangers It Poses for Humans

2min

28.

The Importance of Proof in AI Safety

2min

29.

The Cognitive Bias in AI

2min

30.

The Problem With Extrapolating Trends

2min

31.

The Evolution of AI Safety

2min

32.

Ray Kurzweil's Predictions of the Future

2min

33.

The Importance of Deep Learning in AI Safety

2min

34.

The Importance of Monitoring AI

2min

35.

How to Be Optimistic About AI Development

3min

36.

The Exponent Curve of AI Accidents

2min

37.

The Proportion of AI Accidents

4min

38.

The Lack of a Journal for AI Safety

2min

39.

AI Safety in the Machine Learning Community

3min

40.

The Future of AI Safety

3min

41.

The Importance of Regulation in AI

3min

42.

The Upsides of Open Source Language Models

2min

43.

How to Improve Mechanistic Interpretability

2min

44.

The Difference Between What Comes Out of a Language Model and What's Happening at the Level of the Actual Weights

2min

45.

The Preference of the Model for Predicting the Next Token

2min

46.

The Importance of Concepts in Language Models

2min

47.

The Implications of Specialization in AI Safety

3min

Roman Yampolskiy joins the podcast to discuss various objections to AI safety, impossibility results for AI, and how much risk civilization should accept from emerging technologies. You can read more about Roman's work at http://cecs.louisville.edu/ry/ Timestamps: 00:00 Objections to AI safety 15:06 Will robots make AI risks salient? 27:51 Was early AI safety research useful? 37:28 Impossibility results for AI 47:25 How much risk should we accept? 1:01:21 Exponential or S-curve? 1:12:27 Will AI accidents increase? 1:23:56 Will we know who was right about AI? 1:33:33 Difference between AI output and AI model Social Media Links: ➡️ WEBSITE: https://futureoflife.org ➡️ TWITTER: https://twitter.com/FLIxrisk ➡️ INSTAGRAM: https://www.instagram.com/futureoflifeinstitute/ ➡️ META: https://www.facebook.com/futureoflifeinstitute ➡️ LINKEDIN: https://www.linkedin.com/company/future-of-life-institute/