AI Can Read Books Now—What That Means for You

In 2022, AI researcher Gary Marcus made a bet that by 2029, AI would still not be able to read a novel and answer questions about its plot, characters, and motivations. He thought AI lacked true comprehension. But a new article argues that AI has already surpassed this challenge. Since early 2025, advanced AI models can easily answer both simple and complex questions about stories, even when the answers aren't directly stated in the text.

The author uses an example from Marcus's own book, 'Rebooting AI,' where two children see a cat. A competent reader can infer that Chloe saw the cat because Alexander pointed it out. Modern AI can make that same inference. Marcus recently doubled down on his claim, saying there's no evidence AI can read novels. But the author says this is clearly false, given how well AI performs on reading comprehension tasks.

This matters because AI's ability to understand language is improving faster than many experts predicted. That could mean AI becomes better at jobs that require reading and writing, like customer service, research, and even creative writing. It also raises questions about how we test AI and what we mean by 'understanding.' While AI still makes mistakes, its reading skills are now far beyond what skeptics expected.

The author emphasizes that despite this progress, we still need to regulate AI and solve safety issues. But the debate over AI's capabilities is shifting: the question is no longer if AI can read, but what we do with that power.