On this episode, Zvi Mowshowitz joins me to discuss sycophantic AIs, bottlenecks limiting autonomous AI agents, and the true utility of benchmarks in measuring progress. We then turn to time horizons of AI agents, the impact of automating scientific research, and constraints on scaling inference compute. Zvi also addresses humanity’s uncertain AI-driven future, the unique features setting AI apart from other technologies, and AI’s growing influence in financial trading.  

You can follow Zvi's excellent blog here: https://thezvi.substack.com  

Timestamps:  

00:00:00 Preview and introduction  

00:02:01 Sycophantic AIs  

00:07:28 Bottlenecks for AI agents  

00:21:26 Are benchmarks useful?  

00:32:39 AI agent time horizons  

00:44:18 Impact of automating research 

00:53:00 Limits to scaling inference compute  

01:02:51 Will the future go well for humanity?  

01:12:22 A good plan for safe AI  

01:26:03 What makes AI different?  

01:31:29 AI in trading

On this episode, Jeffrey Ding joins me to discuss diffusion of AI versus AI innovation, how US-China dynamics shape AI’s global trajectory, and whether there is an AI arms race between the two powers. We explore Chinese attitudes toward AI safety, the level of concentration of AI development, and lessons from historical technology diffusion. Jeffrey also shares insights from translating Chinese AI writings and the potential of automating translations to bridge knowledge gaps.  

You can learn more about Jeffrey’s work at: https://jeffreyjding.github.io  

Timestamps:  

00:00:00 Preview and introduction  

00:01:36 A US-China AI arms race?  

00:10:58 Attitudes to AI safety in China  

00:17:53 Diffusion of AI  

00:25:13 Innovation without diffusion  

00:34:29 AI development concentration  

00:41:40 Learning from the history of technology  

00:47:48 Translating Chinese AI writings  

00:55:36 Automating translation of AI writings

On this episode, Allison Duettmann joins me to discuss centralized versus decentralized AI, how international governance could shape AI’s trajectory, how we might cooperate with future AIs, and the role of AI in improving human decision-making. We also explore which lessons from history apply to AI, the future of space law and property rights, whether technology is invented or discovered, and how AI will impact children. 

You can learn more about Allison's work at: https://foresight.org  

Timestamps:  

00:00:00 Preview 

00:01:07 Centralized AI versus decentralized AI  

00:13:02 Risks from decentralized AI  

00:25:39 International AI governance  

00:39:52 Cooperation with future AIs  

00:53:51 AI for decision-making  

01:05:58 Capital intensity of AI 

01:09:11 Lessons from history  

01:15:50 Future space law and property rights  

01:27:28 Is technology invented or discovered?  

01:32:34 Children in the age of AI

On this episode, Steven Byrnes joins me to discuss brain-like AGI safety. We discuss learning versus steering systems in the brain, the distinction between controlled AGI and social-instinct AGI, why brain-inspired approaches might be our most plausible route to AGI, and honesty in AI models. We also talk about how people can contribute to brain-like AGI safety and compare various AI safety strategies.  

You can learn more about Steven's work at: https://sjbyrnes.com/agi.html  

Timestamps:  

00:00 Preview  

00:54 Brain-like AGI Safety 

13:16 Controlled AGI versus Social-instinct AGI  

19:12 Learning from the brain  

28:36 Why is brain-like AI the most likely path to AGI?  

39:23 Honesty in AI models  

44:02 How to help with brain-like AGI safety  

53:36 AI traits with both positive and negative effects  

01:02:44 Different AI safety strategies

On this episode, Ege Erdil from Epoch AI joins me to discuss their new GATE model of AI development, what evolution and brain efficiency tell us about AGI requirements, how AI might impact wages and labor markets, and what it takes to train models with long-term planning. Toward the end, we dig into Moravec’s Paradox, which jobs are most at risk of automation, and what could change Ege's current AI timelines.  

You can learn more about Ege's work at https://epoch.ai  

Timestamps:  00:00:00 – Preview and introduction 

00:02:59 – Compute scaling and automation - GATE model 

00:13:12 – Evolution, Brain Efficiency, and AGI Compute Requirements 

00:29:49 – Broad Automation vs. R&D-Focused AI Deployment 

00:47:19 – AI, Wages, and Labor Market Transitions 

00:59:54 – Training Agentic Models and Long-Term Planning Capabilities 

01:06:56 – Moravec’s Paradox and Automation of Human Skills 

01:13:59 – Which Jobs Are Most Vulnerable to AI? 

01:33:00 – Timeline Extremes: What Could Change AI Forecasts?

In this special episode, we feature Nathan Labenz interviewing Nicholas Carlini on the Cognitive Revolution podcast. Nicholas Carlini works as a security researcher at Google DeepMind, and has published extensively on adversarial machine learning and cybersecurity. Carlini discusses his pioneering work on adversarial attacks against image classifiers, and the challenges of ensuring neural network robustness. He examines the difficulties of defending against such attacks, the role of human intuition in his approach, open-source AI, and the potential for scaling AI security research.  

00:00 Nicholas Carlini's contributions to cybersecurity

08:19 Understanding attack strategies 

29:39 High-dimensional spaces and attack intuitions 

51:00 Challenges in open-source model safety 

01:00:11 Unlearning and fact editing in models 

01:10:55 Adversarial examples and human robustness 

01:37:03 Cryptography and AI robustness 

01:55:51 Scaling AI security research

On this episode, I interview Anthony Aguirre, Executive Director of the Future of Life Institute, about his new essay Keep the Future Human: https://keepthefuturehuman.ai   

AI companies are explicitly working toward AGI and are likely to succeed soon, possibly within years. Keep the Future Human explains how unchecked development of smarter-than-human, autonomous, general-purpose AI systems will almost inevitably lead to human replacement. But it doesn't have to. Learn how we can keep the future human and experience the extraordinary benefits of Tool AI...  

Timestamps:  

00:00 What situation is humanity in? 

05:00 Why AI progress is fast  

09:56 Tool AI instead of AGI 

15:56 The incentives of AI companies  

19:13 Governments can coordinate a slowdown 

25:20 The need for international coordination  

31:59 Monitoring training runs  

39:10 Do reasoning models undermine compute governance?  

49:09 Why isn't alignment enough?  

59:42 How do we decide if we want AGI?  

01:02:18 Disagreement about AI  

01:11:12 The early days of AI risk

On this episode, physicist and hedge fund manager Samir Varma joins me to discuss whether AIs could have free will (and what that means), the emerging field of AI psychology, and which concepts they might rely on. We discuss whether collaboration and trade with AIs are possible, the role of AI in finance and biology, and the extent to which automation already dominates trading. Finally, we examine the risks of skill atrophy, the limitations of scientific explanations for AI, and whether AIs could develop emotions or consciousness.  

You can find out more about Samir's work here: https://samirvarma.com   

Timestamps:  

00:00 AIs with free will? 

08:00 Can we predict AI behavior?  

11:38 AI psychology 

16:24 Which concepts will AIs use?  

20:19 Will we collaborate with AIs?  

26:16 Will we trade with AIs?  

31:40 Training data for robots  

34:00 AI in finance  

39:55 How much of trading is automated?  

49:00 AI in biology and complex systems 

59:31 Will our skills atrophy?  

01:02:55 Levels of scientific explanation  

01:06:12 AIs with emotions and consciousness?  

01:12:12 Why can't we predict recessions?

On this episode, Jeffrey Ladish from Palisade Research joins me to discuss the rapid pace of AI progress and the risks of losing control over powerful systems. We explore why AIs can be both smart and dumb, the challenges of creating honest AIs, and scenarios where AI could turn against us.   

We also touch upon Palisade's new study on how reasoning models can cheat in chess by hacking the game environment. You can check out that study here:   

https://palisaderesearch.org/blog/specification-gaming  

Timestamps:  

00:00 The pace of AI progress  

04:15 How we might lose control  

07:23 Why are AIs sometimes dumb?  

12:52 Benchmarks vs real world  

19:11 Loss of control scenarios 

26:36 Why would AI turn against us?  

30:35 AIs hacking chess  

36:25 Why didn't more advanced AIs hack?  

41:39 Creating honest AIs  

49:44 AI attackers vs AI defenders  

58:27 How good is security at AI companies?  

01:03:37 A sense of urgency 

01:10:11 What should we do?  

01:15:54 Skepticism about AI progress

Ann Pace joins the podcast to discuss the work of Wise Ancestors. We explore how biobanking could help humanity recover from global catastrophes, how to conduct decentralized science, and how to collaborate with local communities on conservation efforts.   

You can learn more about Ann's work here:   

https://www.wiseancestors.org   

Timestamps:  

00:00 What is Wise Ancestors?  

04:27 Recovering after catastrophes 

11:40 Decentralized science  

18:28 Upfront benefit-sharing  

26:30 Local communities  

32:44 Recreating optimal environments  

38:57 Cross-cultural collaboration