Of course I would like to be able to use the latest and greatest generative AI / machine learning / LLMs and the like1 to make my life and work easier. To that end, I periodically take to ChatGPT (or an alternative) to test out its powers as an educational tool - either to support learning or aid administration.

I have set a pretty low bar for success, mostly because the results have not been especially impressive, but even so I have usually been disappointed.

Today, we were talking about general knowledge in a class, inspired in part by a pair of old academic studies on the topic.2 It was a fun discussion, and as part of the prep for this I had taken to technology once again, asking the question:

What is the easiest question to answer that the fewest people know the answer to?

I chose the option to have a ‘reasoned’ response rather than a regular query - and so it reasoned:

The user’s question seems philosophical and a bit tricky to interpret. I’m thinking it could mean: "What's an easy question that few people actually know the answer to?"

Trying not to feel like a character in a generic 90s BBC sci-fi show, “the user” read on:

It might also be a meta-question, where its simplicity makes the answer elusive. There are a few ways to interpret this!

So far, so good. But today the ghost in the machine impressed me like it had not managed to before; after a little meander it landed on a final answer:

Some might respond with a self-referential twist, saying that the easiest question of that description is simply:

“What is the easiest question to answer that the fewest people know the answer to?”

Genius.

Although the class still managed to best the machine, with the delightfully simple:

What is a woman?

And the beautifully metaphysical:

Are you okay?

Thank you, Chat Gippity. You were close.

  1. I am still hunting for a term I am happy with for this stuff - it’s not intelligence, artificial or otherwise, or at least I am not prepared to make that commitment as things stand. ↩︎

  2. Tauber, S.K., Dunlosky, J., Rawson, K.A. et al., General knowledge norms: Updated and expanded from the Nelson and Narens (1980) norms, Behavior Research Methods 45, pp 1115–1143 (2013). https://doi.org/10.3758/s13428-012-0307-9

    Nelson, T.O., Narens, L., Norms of 300 general-information questions: Accuracy of recall, latency of recall, and feeling-of-knowing ratings, Journal of Verbal Learning and Verbal Behavior, Volume 19, Issue 3, pp 338-368 (1980). https://doi.org/10.1016/S0022-5371(80)90266-2 ↩︎