Anticipating an Einstein moment in the understanding of consciousness, with Henry Shevlin

Timestamps:
00:00 Introduction: The Central Question Can machines possess consciousness?
01:20 – Dr. Shevlin’s Background and experience
06:00 – Mentalising and Attribution of Mind
07:00 – Defining Consciousness
09:31 – Theories of Consciousness
10:00 – Conceptual Flaws in Defining Consciousness
12:45 – AI Consciousness and the LaMDA Controversy
14:47 – The 20 Percent Chance Hypothesis
15:00 – Distinguishing Consciousness from Sentience
16:05 – Skepticism Regarding Valenced Conscious Experience in AI
17:10 – Moral Consideration and Consciousness
20:00 – High-Level Qualities of Informational Systems (e.g., Boredom)
20:40 – Ethical Implications of Building AI / Antinatalism
22:00 – Negative Signals in AI
22:40 – Hedonic Offsetting Proposal
24:40 – Reliability of Consciousness Testing
27:10 – Deep Realism and Consciousness
29:05 – Behavioural Understanding & Murray Shanahan
30:25 – The Role of Social AI
30:47 – The Danger of Anthropomorphising AI
32:50 – False Positives of Consciousness
33:30 – Challenges in Controlling Conscious AI
35:40 – Superintelligence as the “Einstein Moment”
37:40 – Conscious AGI vs. Non-Conscious AGI
38:10 – ASI Consciousness and Empathy for Humanity

Welcome to the first episode of Understanding Machine Consciousness. This episode is a collaboration with The London Futurists Podcast.

Our guest in this episode is Henry Shevlin. Henry is the Associate Director of the Leverhulme Centre for the Future of Intelligence at the University of Cambridge, where he also co-directs the Kinds of Intelligence program and oversees educational initiatives.

He researches the potential for machines to possess consciousness, the ethical ramifications of such developments, and the broader implications for our understanding of intelligence.

In his 2024 paper, “Consciousness, Machines, and Moral Status,” Henry examines the recent rapid advancements in machine learning and the questions they raise about machine consciousness and moral status. He suggests that public attitudes towards artificial consciousness may change swiftly, as human-AI interactions become increasingly complex and intimate. He also warns that our tendency to anthropomorphise may lead to misplaced trust in and emotional attachment to AIs.

Note: this episode is co-hosted by David and Will Millership, the CEO of a non-profit called Prism (Partnership for Research Into Sentient Machines). Prism is seeded by Conscium, a startup where both Calum and David are involved, and which, among other things, is researching the possibility and implications of machine consciousness. Will and Calum will be releasing a new Prism podcast focusing entirely on Conscious AI, and the first few episodes will be in collaboration with the London Futurists Podcast.

Find out more: https://www.prism-global.com/podcast/...

Selected follow-ups:

Henry Shevlin - personal site
Kinds of Intelligence - Leverhulme Centre for the Future of Intelligence
Consciousness, Machines, and Moral Status - 2024 paper by Henry Shevlin
Apply rich psychological terms in AI with care - by Henry Shevlin and Marta Halina
What insects can tell us about the origins of consciousness - by Andrew Barron and Colin Klein
Consciousness in Artificial Intelligence: Insights from the Science of Consciousness - By Patrick Butlin, Robert Long, et al
Association for the Study of Consciousness
London Futurist Podcast
https://www.prism-global.com/