Turns out our soon-to-be AI Overlord, ChatGPT, has a worldview based in the 19th-century canon, Gen X sci-fi favorites, and the social dynamics at Hogwart’s School For Lil Magicians. According to this profile at Business Insider, a data scientist has reverse-engineered ChatGPT’s reading list by quizzing the thing about canonical texts. As the article explains:
To figure out what GPT-4 has read, they quizzed it on its knowledge of various books, as if it were a high-school English student. Then they gave it a score for each book. The higher the score, the likelier it was that the book was part of the bot’s dataset — not just crunched to help the bot generate new language, but actually memorized.
The study, which hasn’t been peer reviewed, reveals the top 20 most likely books that have been fed into ChatGPT (along with a list of the top 50 non-public domain books). What stands out to me here is the high number of mediocre to bad books being forced upon our robot friend: between Fifty Shades of Grey and Dan Brown, it’ll be no wonder that ChatGPT eventually chooses to eradicate the human race. Sure, there are some wonderful books on this list, but why not aim a little higher with the entire syllabus?
But insofar as ChatGPT has evidently read Trainspotting, at least the end of humanity might come with a bit of Glaswegian flair.