AI Hype
Over the past few months, there has been a ridiculous amount of hype about AI from people who don’t understand the basics of the subject: the easy stuff is hard and the hard stuff is easy. They remind me of my grandmother, who thinks texting is more astonishing than videophones. Most attention has been concentrated at computers plagiarizing answers to text questions and coming up with very simple -indeed, 2000s or late 1990s level- models of computer art.
The places where AI would genuinely be impressive -music and computer vision- are getting surprisingly little attention. Needless to say, there have been some impressive steps in computer vision over the past five years -the Discord content filter is the most notable example I know of -test it out yourself by making your own nudes server.
For AI to actually be impressive, it has yet to pass two tests: the Turing test and giving an accurate analysis of Piknik lyrics. Bryan Caplan has already pointed out actually existing AI is, despite possessing all the information in the world, is still bad at answering actual questions posed to economics students.
The goal of humanity should not be to get GPT-3 to type like humans - it is to get humans to speak like GPT-3.