I’ve been a bit disappointed with the debate about the future use of AI in science. Either it has been short-sighted: “Wow, let’s try this thing we’ve always been doing, but with chatbots instead” or pessimistic without any real arguments. So, in remedy, I wrote an essay, channeling my inner Marvin Minsky.
Download! 👏🏻 It! 👏🏻 Here! 👏🏻 SSNR:5131737.
Central to the essay are the limits human language puts on knowledge acquisition. For example, how we will never take the time to get away from the messiness of the vernacular language (see below) in favor of logical and systematic language that could support our collective intelligence project called science. However, ML/AI could do, in principle, but would also have to overcome several language barriers.
An illustration I found floating around the Internets, lauded for its clarity and precision. This might well be the case for the illustration, but not so for the concepts of aggregations of humans and animals depicted. Imagine if we rather had nonoverlapping concepts, filling out the entire space; how much more useful wouldn’t the language be (assuming the two axes of meaning are the most important, etc.).
Petter Holme 