This is a pretty good substack post padded with 250 pages of filler. What a frustrating disappointment. If the subject interests you, save yourself a lot of time and start on chapter 14.

The first 200-plus pages of this book are table-setting, long essays with scares like, “What if the Unabomber, but with AI”? It reads like a series of university freshman-level papers, padded with speculation, hyperbole, weak analogies and cherry-picked examples. Typical of so many in the AI space, the author knows tech, for sure, but his grasp on culture, art and history is as deep as a puddle. Be wary of anyone who uses ‘compute’ as a noun.

The last full chapter is really all you need: Suleyman lays out a 10-point proposal to contain the risk of AI and synthetic biology, and there’s lots of merit to it. Even then it’s padded, and I was skipping paragraphs regularly. But it’s original and worth engaging with.

I could nitpick it – he defends bringing a transphobe on an ethics committee with the stunning idea that ‘some might share this view’ (while elsewhere acknowledging that racism in AI is a legitimate problem). In fact he contradicts himself several times about various things, suggesting that this was a bit rushed to press – but it isn’t worth the time.

This is one of those books that makes otherwise smart people think that reading is a waste of time.