Nick Allardice joins the podcast to discuss how GiveDirectly uses AI to target cash transfers and predict natural disasters. Learn more about Nick's work here: https://www.nickallardice.com  

Timestamps: 

00:00 What is GiveDirectly? 

15:04 AI for targeting cash transfers 

29:39 AI for predicting natural disasters 

46:04 How scalable is GiveDirectly's AI approach? 

58:10 Decentralized vs. centralized data collection 

1:04:30 Dream scenario for GiveDirectly

Nathan Labenz joins the podcast to provide a comprehensive overview of AI progress since the release of GPT-4. 

You can find Nathan's podcast here: https://www.cognitiverevolution.ai   

Timestamps: 

00:00 AI progress since GPT-4  

10:50 Multimodality  

19:06 Low-cost models  

27:58 Coding versus medicine/law  

36:09 AI agents  

45:29 How much are people using AI?  

53:39 Open source  

01:15:22 AI industry analysis  

01:29:27 Are some AI models kept internal?  

01:41:00 Money is not the limiting factor in AI  

01:59:43 AI and biology  

02:08:42 Robotics and self-driving  

02:24:14 Inference-time compute  

02:31:56 AI governance  

02:36:29 Big-picture overview of AI progress and safety

Connor Leahy joins the podcast to discuss the motivations of AGI corporations, how modern AI is "grown", the need for a science of intelligence, the effects of AI on work, the radical implications of superintelligence, open-source AI, and what you might be able to do about all of this.   

Here's the document we discuss in the episode:   

https://www.thecompendium.ai  

Timestamps: 

00:00 The Compendium 

15:25 The motivations of AGI corps  

31:17 AI is grown, not written  

52:59 A science of intelligence 

01:07:50 Jobs, work, and AGI  

01:23:19 Superintelligence  

01:37:42 Open-source AI  

01:45:07 What can we do?

Suzy Shepherd joins the podcast to discuss her new short film "Writing Doom", which deals with AI risk. We discuss how to use humor in film, how to write concisely, how filmmaking is evolving, in what ways AI is useful for filmmakers, and how we will find meaning in an increasingly automated world.   

Here's Writing Doom:   https://www.youtube.com/watch?v=xfMQ7hzyFW4   

Timestamps: 

00:00 Writing Doom  

08:23 Humor in Writing Doom 

13:31 Concise writing  

18:37 Getting feedback 

27:02 Alternative characters 

36:31 Popular video formats 

46:53 AI in filmmaking

49:52 Meaning in the future

Andrea Miotti joins the podcast to discuss "A Narrow Path" — a roadmap to safe, transformative AI. We talk about our current inability to precisely predict future AI capabilities, the dangers of self-improving and unbounded AI systems, how humanity might coordinate globally to ensure safe AI development, and what a mature science of intelligence would look like.   

Here's the document we discuss in the episode:   

https://www.narrowpath.co  

Timestamps: 

00:00 A Narrow Path 

06:10 Can we predict future AI capabilities? 

11:10 Risks from current AI development 

17:56 The benefits of narrow AI  

22:30 Against self-improving AI  

28:00 Cybersecurity at AI companies  

33:55 Unbounded AI  

39:31 Global coordination on AI safety 

49:43 Monitoring training runs  

01:00:20 Benefits of cooperation  

01:04:58 A science of intelligence  

01:25:36 How you can help

Tamay Besiroglu joins the podcast to discuss scaling, AI capabilities in 2030, breakthroughs in AI agents and planning, automating work, the uncertainties of investing in AI, and scaling laws for inference-time compute. Here's the report we discuss in the episode:  

https://epochai.org/blog/can-ai-scaling-continue-through-2030  

Timestamps: 

00:00 How important is scaling?  

08:03 How capable will AIs be in 2030?  

18:33 AI agents, reasoning, and planning 

23:39 Automating coding and mathematics  

31:26 Uncertainty about investing in AI 

40:34 Gap between investment and returns  

45:30 Compute, software and data 

51:54 Inference-time compute 

01:08:49 Returns to software R&D  

01:19:22 Limits to expanding compute

Ryan Greenblatt joins the podcast to discuss AI control, timelines, takeoff speeds, misalignment, and slowing down around human-level AI. 

You can learn more about Ryan's work here: https://www.redwoodresearch.org/team/ryan-greenblatt  

Timestamps: 

00:00 AI control  

09:35 Challenges to AI control  

23:48 AI control as a bridge to alignment 

26:54 Policy and coordination for AI safety 

29:25 Slowing down around human-level AI 

49:14 Scheming and misalignment 

01:27:27 AI timelines and takeoff speeds 

01:58:15 Human cognition versus AI cognition

Tom Barnes joins the podcast to discuss how much the world spends on AI capabilities versus AI safety, how governments can prepare for advanced AI, and how to build a more resilient world.   

Tom's report on advanced AI: https://www.founderspledge.com/research/research-and-recommendations-advanced-artificial-intelligence   

Timestamps: 

00:00 Spending on safety vs capabilities 

09:06 Racing dynamics - is the classic story true?  

28:15 How are governments preparing for advanced AI?  

49:06 US-China dialogues on AI 

57:44 Coordination failures  

1:04:26 Global resilience  

1:13:09 Patient philanthropy  

The John von Neumann biography we reference: https://www.penguinrandomhouse.com/books/706577/the-man-from-the-future-by-ananyo-bhattacharya/

Samuel Hammond joins the podcast to discuss whether AI progress is slowing down or speeding up, AI agents and reasoning, why superintelligence is an ideological goal, open source AI, how technical change leads to regime change, the economics of advanced AI, and much more.   

Our conversation often references this essay by Samuel: https://www.secondbest.ca/p/ninety-five-theses-on-ai   

Timestamps: 

00:00 Is AI plateauing or accelerating?  

06:55 How do we get AI agents?  

16:12 Do agency and reasoning emerge?  

23:57 Compute thresholds in regulation

28:59 Superintelligence as an ideological goal 

37:09 General progress vs superintelligence 

44:22 Meta and open source AI  

49:09 Technological change and regime change 

01:03:06 How will governments react to AI?  

01:07:50 Will the US nationalize AGI corporations?  

01:17:05 Economics of an intelligence explosion  

01:31:38 AI cognition vs human cognition  

01:48:03 AI and future religions 

01:56:40 Is consciousness functional?  

02:05:30 AI and children

Anousheh Ansari joins the podcast to discuss how innovation prizes can incentivize technical innovation in space, AI, quantum computing, and carbon removal. We discuss the pros and cons of such prizes, where they work best, and how far they can scale. Learn more about Anousheh's work here: https://www.xprize.org/home  

Timestamps: 

00:00 Innovation prizes at XPRIZE 

08:25 Deciding which prizes to create 

19:00 Creating new markets 

29:51 How far can prizes scale?  

35:25 When are prizes successful?  

46:06 100M dollar carbon removal prize 

54:40 Upcoming prizes 

59:52 Anousheh's time in space

Mary Robinson joins the podcast to discuss long-view leadership, risks from AI and nuclear weapons, prioritizing global problems, how to overcome barriers to international cooperation, and advice to future leaders. Learn more about Robinson's work as Chair of The Elders at https://theelders.org  

Timestamps: 

00:00 Mary's journey to presidency  

05:11 Long-view leadership 

06:55 Prioritizing global problems 

08:38 Risks from artificial intelligence 

11:55 Climate change 

15:18 Barriers to global gender equality  

16:28 Risk of nuclear war  

20:51 Advice to future leaders  

22:53 Humor in politics 

24:21 Barriers to international cooperation  

27:10 Institutions and technological change