I, For One, Welcome Our New Self-Driving Overlords

10 Replies to “I, For One, Welcome Our New Self-Driving Overlords”

1. That’s enough K pop for one night

2. They put Kamala Harris in a computer?

3. Re the music rating …
   I wasted my afternoon listening to Paul McCartney when I could have been listening to
   Paul McFartney. Chat GPT would have rated it higher

   1. Hey ChatGPT, is Paul McFartney a singer or a tutor? Then Kirk says “Mr. Spock, let’s get out of here”.

4. Having used commercial AI, I must say these blatherings are not that common when a large-context AI solves a closed problem using more than negligible amount of computing power. With a price of running a V8 for minutes, you’ll get a lot of computing and it is incredible.

   The price of the AI comes partly from the fact you’ll need a lot of LLM to debug what one run of an LLM got wrong. It’s easier to make the LLM talk than to make it do a simple calculation, but believe it or not, I’m impressed with my Claude doing fact-finding in more closed problems.

   If you give it a very open problem, it might be less useful. But it definitely beats me in many, many things.

   Anyway, with an LLM, you better be prepared it feeds you something that looks right and is not.

   1. LLMs are an implementation of the proof of Gell-Mann amnesia. They all suck, at everything, and the data demonstrates this (in my field, software engineers think they’re more productive using LLMs. The actual data shows they’re less productive). But people only notice that in fields they know very well, and forget it when they use LLMs for fields they don’t.

5. I have yet to use Chat Crap Whatever and frankly never will. I downloaded grok last summer and have zero complaints thus far, although I know it has its issues.

6. It certainly knows how to lie when caught.

7. A computer voice with annoying vocal fry. It’s getting worse.

   1. The only place I want to hear vocal fry is from somebody in the electric chair.