Co-Intelligence: Living and Working with AI

I recently read Wharton professor Ethan Mollick’s new book, Co-Intelligence: Living and Working with AI, while flying over to Prague to attend EuroPython. It’s not a long title, and it was a relatively quick read. Still, I found it generally useful for anyone interested in a high-level discussion of modern LLMs, how they are already influencing everyday life, and the authors’ multiple predictions about what they will lead to in the future.

The book is divided into nine different chapters that read like independent essays. I am not an active reader of the author’s substack newsletter, One Useful Thing, but I would be surprised if he had not already written some version of each of these things in there.

Part 1 of the book provides an overview of how LLMs work, the alignment problem with them (aka how do we make sure they help rather than destroy humans), and “Four Rules” for using AI that basically boils down to use it and assume it will only get better over time.

I will admit that as a programmer I often found myself wanting a lot more depth and rigor. “AI” is a lazy way to refer to “LLMs,” and the author admits this, but I also suspect that for lay readers, the two terms are more or less the same. In the same way, I found myself nodding along in agreement as he described how they work, and yet, just as things became interesting, he stopped and moved on to more general usage points rather than going further on the technical details. I plan to next read Why Machines Learn, which I suspect will provide the more detailed rigor I enjoy on this topic.

Part 2 of the book focuses broadly on how to use AI to be creative, use it at work, in school, and as a life coach. I can see how these chapters work well for, again, a non-technical reader who is curious but ignorant of any of these technologies. And it is clear the author has given these topics a lot of thought. But I didn’t find a lot of newness there, at least for me.

It’s interesting how he uses AI in the classroom. Forcing students to engage with it, play with it, and try to find both its uses and shortcomings. I agree with him that you can’t avoid LLMs going forward. They already “level up” most people’s writing, serve as a proxy for research, and whatever end form they inhabit will likely be a fundamental part of our lives going forward.

I had a layover at the Amsterdam airport on my flight to EuroPython, and sure enough, in the very small “business section,” this book was there on the shelf. It has also sold quite well, reflecting both the timeliness of the topic and, no doubt, Mollick’s 365k or so Substack followers.

For myself, it is a good quick read to see how a “thought leader” is presenting the topic and providing advice to non-technical readers. It is likely a good gift idea for someone in that bucket. Yet for myself, as previously noted, I wanted more. If you also fall into that bucket, the single best resource I can recommend is Simon Willison’s blog, where he writes almost daily and with real rigor about what’s happening in the LLM space.