Kevin Roose book examines whether AI is smarter than humans yet

In short: A new book by journalist Kevin Roose looks at the push to build AGI and says today’s AI can outperform people in some areas, but it is not consistently smarter than humans.

What happened

Kevin Roose’s book, The AGI Chronicles: The Inside Story of the Race to Create an Artificial Superintelligence, was released on October 6, 2026. It describes competition among major AI companies, including OpenAI, Anthropic, and Google DeepMind, to build what many call AGI.

AGI stands for artificial general intelligence, meaning an AI that can handle many different kinds of thinking and problem solving, not just one narrow task. Roose’s publisher says the book is based on more than 150 interviews, plus company documents.

Roose focuses on a simple question: could AI already be matching or beating people in important ways? He points to systems that can do some specialized work extremely well, like solving difficult math and science problems very quickly. That is like a calculator that can beat any person at arithmetic, while still not being able to run a household or hold a reliable conversation.

Roose also stresses the limits. Performance changes depending on the task, and today’s AI can still be unreliable. In a CBS interview, he suggested a practical bar for AGI: an AI that is at least as intelligent, honest, and reliable as an ordinary person.

Why it matters

Many headlines talk about AI as if it is simply “smarter than humans.” This book argues that the real question is what task you mean, and whether the system can be trusted day to day. That matters for schools, workplaces, and public services where mistakes or made-up answers can cause real harm.

Source: NYTimes