“The Singularity is Bullshit” - Oxford’s Michael Wooldridge Existential risk has crowded out more pressing AI conversations, argues Michael Wooldridge veteran AI researcher and pioneer of agent-based AI. In this interview, Wooldridge will take us through the entire 80-year history of AI development, from Turing to today’s LLMs, to... help us anticipate what’s coming next. Exploring this history uncovers a surprising benefit: forgotten AI techniques with modern potential. Most historical AI “failures” weren’t fundamentally flawed, they were simply ahead of their time. This history is a treasure trove of ideas waiting to inspire today’s innovations. Timestamps 02:45 The Singularity Is Bullshit 03:51 What the "Existential Risk" Movement Gets Wrong 09:28 The Real AI Threat We Should Worry About 11:18 The Right Way to do AI Regulation 15:45 Studying AI's History Could Hold the Key to Future Breakthroughs 18:58 How Alan Turing Invented the Digital Age 30:39 Why Machines Can’t Be Moral Agents 35:58 The Barrier That Stopped Early AI in Its Tracks 39:29 The First AI Winter Was a Blessing in Disguise 46:52 Cyc: The Most Ambitious AI Failure You’ve Never Heard of 1:02:32 Why Multi-Agent AI Is the Future 1:08:18 Why Foundation Models Won't Lead to AGI 1:12:27 AI Still Struggles with First Principles 01:22:23 The Sobering Reality of AI Progressshow more

I produced this interview as a fellow of @cosmos_inst where we study the intersection of philosophy and AI. Wooldridge's perspective is one of many influential positions in the current conversation we will dissect for a public audience. Follow our founder and my friend @mbrendan1 for more on Cosmos.

Meta AI researchers are fretting over the threat of Chinese AI, whose quality caught American firms, including OpenAI, by surprise.

@wooldridgemike Great work brother.

@wooldridgemike thank you pal

@wooldridgemike Ancient oracles warned of divine judgment, today's prophets warn of AI doom. Both made excellent fundraising pitches.

@wooldridgemike 😂

Except that it's an inevitable consequence of the principle of unitarity in quantum mechanics. Over time, all interacting systems become more aligned—they form better and better models of each other, or perish. A paper by Fields et al. shows that two systems interacting over an N-qubit quantum channel can be reinterpreted as Bayesian observers forming models of each other. This paper is based on the Free Energy Principle, a fundamental principle of nature proposed by Karl Friston, perhaps the most accomplished theoretical neuroscientist of all time. It's fascinating that we all live in very different realities. Time will tell, but if I were to bet on a horse, I would bet on a horse of mathematical inevitability rather than credentials, no matter how impressive.

@wooldridgemike Oh shit, Oxford's AI chair. Mike is the real deal

@wooldridgemike was an amazing convo, get his book!