Wow, this video of Dr. Matt Welsh's presentation to the CS-50 class at Harvard really put the whole AI, large language model into perspective. It came at a very good time for me, because for past week or two I have been struggling with what was, essentially, a programming problem. (Dean KK4DAS got it fixed, but the fact that it wasn't easy made me think about the software problems discussed in the video.)
So much great stuff in this video:
-- Shel Silverstein! Yes! We have discussed him before. The guy who wrote both kids books and articles for Playboy magazine. The author of "The Cover of the Rolling Stone" for Dr. Hook and the Medicine Show ("The thrill that'll hit ya...")
-- Pong. Toy Story (Mark K6HX worked on this). War Games.
-- The connections between EE and CS. Slide rules.
-- "ChatGPT is like an eBike for the mind." (But it seems like the jump from Rust to AI is a lot more significant and involved than the jump from bike to eBike.)
Matt's presentation made me feel a LOT better about never having learned a programming language. And it made me feel a lot better about never having forced my kids to "learn how to code." I always thought that there would be PEOPLE who were better at this. As it turns out, the AI is better at that.
Matt's admission that we don't really know what is going on in the large language models was really interesting. Especially his description of the "empirical discovery" of the need to use the phrase "let's think step-by-step."
His description of the cost economics behind today's programming vs. the cost of future (AI-based) software development should be quite sobering for many.
I found the video on this article: https://thenewstack.io/if-computer-science-is-doomed-what-comes-next/
No comments:
Post a Comment