The details
In the interview with Nilay Patel, Roose ranks Anthropic as the lab most committed to AGI, with OpenAI close behind and Google divided: chief executive Sundar Pichai avoids the term, while Demis Hassabis and Sergey Brin use it. He argues belief can pay off, saying OpenAI’s big bet on computing power has left it the only major lab not severely short of it. Patel questions whether the labs can reach AGI before the money runs out. Roose thinks large language models remain the most promising path, perhaps with changes such as continual learning.
Why it matters
Whether or not AGI arrives, belief in it is steering huge bets on computing power and deciding which products survive. Roose’s point that conviction drives these choices helps explain moves that look odd as pure business. Patel’s scepticism is the counterweight: missionary zeal doesn’t pay for chips if revenue lags behind.