I realized that I haven't read or listened much science fiction for a while, so I decided to ask AI for some recommendations. I gave it examples of books and series I've really enjoyed, such as Project Hail Mary, Silo, The Expanse, Bobiverse and The Three-Body Problem.
After the first recommendations, I decided to give it a link to my personal website, where I've kept track of my reading for the past 30 years. I set up my first homepage in the mid-1990s and soon started adding the books I'd read so the site wouldn't be completely static. I still do.
So instead of giving AI just a few examples, I gave it my reading history and asked it to figure out what kind of science fiction I might actually like. That worked surprisingly well. But then I became interested in what else AI could learn from the data.
The basic statistics were easy: books per year, categories, ratings, languages, fiction vs. nonfiction, favorite authors and so on. Nothing particularly special there. I could have imported the data into Excel and done much of it myself. Still, it was interesting to see how much useful analysis AI could produce directly from the raw data.
The more interesting part started when I asked AI to look for patterns that weren't explicitly recorded in the data.
My professional reading seems to have evolved together with my career:
technology → security → management → human behaviour → AI, geopolitics and societal consequences
Roughly 60% of my reading has been nonfiction and 40% fiction. Science fiction alone accounts for about one fifth of all the books, roughly as much as technology, cybersecurity and AI. What I hadn't really noticed before was that my science fiction reading often explores many of the same questions. Technology changes something, complex systems react, people make decisions with incomplete information and eventually something unexpected happens.
AI summarized the common theme rather nicely:
“A long-running exploration of how technology, people and complex systems behave when something changes, breaks or surprises us.”
Another interesting evolution was the way I consume books. Ebooks became increasingly important around 2017 and peaked around 2019. But they turned out to be more of a bridge than the final destination: audiobooks became my largest format in 2020 and now account for roughly two thirds of my reading. These days I mostly listen to books or read physical ones, with just a few ebooks now and then.
AI also made a fairly reasonable conclusion about why. Audiobooks didn't suddenly turn me into a reader, I was already reading roughly the same number of books 20 years ago. But they changed when and where I can consume them. Walking, driving and travelling became potential reading time.
And there's a nice bit of irony in all this: I started by reading books about how digital technology would transform human behaviour. Then digital technology transformed how I read the books about it.
Oh, and the original question wasn't forgotten. AI's current top three sci-fi recommendations for me are:
- Children of Time by Adrian Tchaikovsky
- Spin by Robert Charles Wilson
- Saturn Run by John Sandford
I'll give them a try. Then we'll see how good the algorithm really is.
No comments:
Post a Comment