Joe Carlsmith is a writer, researcher, and philosopher. He works as a senior research analyst at Open Philanthropy, where he focuses on existential risk from advanced artificial intelligence. He also writes independently about various topics in philosophy and futurism, and holds a doctorate in philosophy from the University of Oxford.

You can find links and a transcript at www.hearthisidea.com/episodes/carlsmith

In this episode we talked about a report Joe recently authored, titled ‘Scheming AIs: Will AIs fake alignment during training in order to get power?’. The report “examines whether advanced AIs that perform well in training will be doing so in order to gain power later”; a behaviour Carlsmith calls scheming.

We talk about:

• Distinguishing ways AI systems can be deceptive and misaligned
• Why powerful AI systems might acquire goals that go beyond what they’re trained to do, and how those goals could lead to scheming
• Why scheming goals might perform better (or worse) in training than less worrying goals
• The ‘counting argument’ for scheming AI
• Why goals that lead to scheming might be simpler than the goals we intend
• Things Joe is still confused about, and research project ideas

You can get in touch through our website or on Twitter. Consider leaving us an honest review wherever you're listening to this — it's the best free way to support the show. Thanks for listening!

Eric Schwitzgebel is a professor of philosophy at the University of California, Riverside. His main interests include connections between empirical psychology and philosophy of mind and the nature of belief. His book The Weirdness of the World can be found here.

We talk about:

• The possibility of digital consciousness
  • Policy ideas for avoiding major moral mistakes around digital consciousness
  • Prospects for the science of consciousness, and why we likely won't have clear answers in time
• Why introspection is much less reliable than most people think
  • How and why we invent false stories about our own choices without realising
  • What randomly sampling people's experiences reveals about what we're doing with most of our attention
• The possibility of 'overlapping minds'
• How and why our actions might have infinite effects, both good and bad
  • Whether it would be good news to learn that our actions have infinite effects, or that the universe is infinite in extent
• The best science fiction on digital minds and AI

You can get in touch through our website or on Twitter. Consider leaving us an honest review wherever you're listening to this — it's the best free way to support the show. Thanks for listening!

Sonia Ben Ouagrham-Gormley is an associate professor at George Mason University and Deputy Director of their Biodefence Programme

In this episode we talk about:

• Where the belief that 'bioweapons are easy to make' came from and why it has been difficult to change
• Why transferring tacit knowledge is so difficult -- and the particular challenges that rogue actors face
• As well as lastly what Sonia makes of the AI-Bio risk discourse and what types of advances in technology would cause her concern

You can get in touch through our website or on Twitter. Consider leaving us an honest review wherever you're listening to this — it's the best free way to support the show. Thanks for listening!