The word “intelligence” has thrown a lot of people off when reasoning about AI. Intelligence should probably be defined as “how many examples does it take for you to recognize a pattern.” That’s roughly what humans mean when they call someone intelligent: great pattern recognition, an ability to abstract. That’s in contrast to knowledge, which is how many examples you’ve seen and can remember. Which itself is in contrast to speed of thought, which is how fast you can churn through the examples you’ve seen and the patterns you’ve already deduced in order to come up with new patterns. AI models haven’t really pushed the boundary of “intelligence” upwards, and in fact I’d guess that the smartest humans are basically already optimal in regard to sample efficiency for pattern detection. There’s certainly a hard ceiling on that number. What AI models *have* done is 100000x the boundaries on knowledge and speed of thought. Which is way way more important at the end of the day. You’ll be much less surprised by the future if you keep these distinctions in mind.
Why is it that active management critics of index investing who write papers that point to a recent increase in correlations of stocks returns in the S&P 500 by cap weight never use an equal weight total US stock market index as a proxy? https://t.co/2ZuAcBtkzW
📢CALLING OXFORD📢
#RIHLive
Do you have a historical question for Tom and Dominic? They'll be onstage tonight at New Theatre Oxford, and will be answering your questions in the 2nd half. Send us your questions by replying to this post, or tweet them using the hashtag #RIHLive