The world is consumed with the power and potential of AI. Leading tech names like Altman and Musk have claimed it threatens humanity’s future. While trillions of dollars are being invested. The extraordinary success of Large Language Models, initiated by the breakthrough 2017 transformer/attention coding principles, is seen as just the beginning. More data, more computing, it is assumed, will deliver ever more powerful AI. But a growing number of critics argue these models are at root unintelligent word calculators that are reaching their peak as human training data is exhausted. To be intelligent, they would have to interact with the world, while language models merely find patterns in human text, predicting what comes next.

Does intelligence require an interface with the world akin to a human body, and is the future of AI in robots rather than chatbots? Is the belief that we can scale the attention paradigm profoundly misguided, and does it risk financial collapse as vast investments fail? Or will the language models become ever more effective and take over everything that humans can do?

"The next Einstein" Sabrina Pasterski, post-postmodern philosopher Hilary Lawson, Cambridge philosopher Audrey Borowski, and Oxford AI professor Michael Wooldridge debate whether AI will ever truly think.