Embed Notice
HTML Code
Corresponding Notice
- Embed this notice@hakui
>Allegedly, I and others like me anthropomorphize computers and imagine that they are conscious beings that have thoughts and feelings.
The allegations turn out to be true. I don't think he knows much about the nature of consciousness or intelligence; he certainly doesn't understand computers.
Here's a copy of GEB; if you can get past the guy's boomer sensibilities and his tendency to be overly pleased with his own terrible jokes, it's actually a really entertaining book. Hofstadter's big idea is that intelligence requires a feedback loop. Some of his predictions worked out, some did not, but he's spent a lot more time understanding brains and computers . (I do not recommend "I Am a Strange Loop". It is his attempt to--seriously, he put this in the introduction--revise GEB so it's much less entertaining.)
Also here's a copy of "The Quark and the Jaguar". Gell-Mann's big idea is a description of the "complex adaptive system". I have not finished the book but it's interesting so far. (Early in the book he draws a line between real complexity and "crude complexity".)
> Intelligence is the ability to attain goals, such as maximizing user engagement on a social media platform
This is utter horseshit; it is completely useless as a definition for intelligence. Every goddamn program on my computer is "intelligent" by that definition.
Engagement maximization is trivial: interaction_count/(total_interaction_opportunities*((1+age_penalty)^age)). You can tack on k-means (a little more complicated than a single equation; I'd have to look it up, I don't use it a lot) or crude vector similarity if you want it personalized.
You could say it's "intelligence" if it can come up with its own goals without being handed any prior information about its environment and work out how to achieve them. Autonomy is self-direction, intelligence is a measure of the effectiveness of autonomous decision-making. The Twitter algorithm didn't do this, because it can't: it's a statistical pipe, it doesn't really decide anything. "I don't want to promote that because I suspect that I won't like the outcome on broader society" isn't something that equation will ever do. It's not something ChatGPT can do, even: they have to hard-code "AND DON'T SAY THE N-WORD" into it.
GEBen.pdf
quark-jaguar.pdf