Swarmpicker

I'm saying people are stochastic parrots with a ratchet algorithm, and so are LLMs. And obviously AI can replace people for lots of tasks. And there will be more in the near future.

A "model swarm" of LLMs can refine its accuracy by iteratively interacting. Which is what academic papers are.

So a convicted child molester is getting threats in prison. Doesn't that happen every day? Why is the federal govt protecting this particular felon?

bit.ly/4ou3GJK

What makes you think so?

Nobody would ever call a blues lick a "melody."

Start with Heider's balance theory, which led indirectly to neural-net theorizing in psychology. I'm sure you can find a summary on the web; go from Heider to Cartwright and Harary... to the PDP "bible".

It's okay of course to fuck over consumers, but the tariffs are going to make it hard for corporations, too, and that's going to generate blowback.

There is no empirical evidence that a "mind" exists. It can't be observed or measured. This is basic psychology. A theory of mind is a belief that people have minds, but there is no evidence that they do. LLMs are going to teach us this, real fast, when they do everything our "minds" do.

It's a belief schema. If there is certainty then it is illusory, since there is no evidence that a "mind" exists. Still, we *feel* certain, that really, what else is there?

We have social aptitudes a horse does not have, and language, but we read cues just like a bloody horse, and respond to them without knowing what is "behind" them.

There's no evidence that a human TOM is anything different from what Clever Hans did. We don't *know* if other people have minds. This is a tougher philosophical problem than some. Start with: there is no empirical evidence that a mind exists. Then build up to our "knowledge" of other minds.

It has a theory of mind - they can predict human behavior very well. It's not alive, obviously, and there is currently no substitute for a lifetime of experience. But surprising things have happened.

A brain is not a computer housed in a body, they are one thing. An LLM has neither native senses nor motor abilities, feels neither pain nor hunger. We live in a consensually defined "reality" that is defined by our abilities. The LLM lives in its input data.

If you mean "embodied intelligence," it may be possible someday to simulate embodiment (not a fucking brain in a vat) but obviously LLMs are going to be limited to the experiences included in their training data.

Important to remember that these algorithms are in the "Kitty Hawk" stage of development. They can converse credibly and that's a big deal, like flying a few hundred feet was a big deal. Let's not imagine that they will stay at this stage forever.

The same principles of coherence apply to affect as well as cognition, and the two are intertwined. Sometimes affect wins.

Scrolling I got to this after your other comment. Of course Chomsky has a "universal grammar" theory to defend, and cannot understand that a neural net extracts grammar from examples, like any other statistical pattern. He's not the only linguist who hates this. This article is just kinda sad.

The explanations are justifications. Self-reports of cognitive processes are not considered valid data.

Do you know how many volumes of writing Chomsky produced? If there is something in all that that you think is relevant, then just say it, or say where to look for it.

"Read Noam Chomsky" is a pretty big assignment. If there is something in particular then, uh, I don't like being behind curves, y'know?

Look at driving cars, which we think AI is not good at. How many accidents per day do you think they have, and how many accidents do humans have? We have millions. But we think we're good at it and they aren't. Pure species chauvinism.

An yet you believe there is a "mind" in a human? What is your evidence?

If you think reality is anything more than a Chinese room then you're deluded.

And honestly people don't do these things very well.

AI is going to surprise us by revealing that there is nothing special about being human. "Understanding" isn't any different from what an LLM does, except for implementation in a physical body.

Humility will be the long-term takeaway.

We laugh now but this technology is very very new - it's like laughing at a baby trying to talk.

Conservatism is based on the social psychology of small groups, where norm formation, group dynamics, and social identity help groups cohere and reduce social sources of dissonance. Liberalism aspires to consider various small groups simultaneously.

The big worry about tariffs is that consumers will have to pay more. But the other thing is that companies will make less, if they shave off the profit margin to keep prices down. And the CEOs are not going to like that.

One of the factors working in Trump’s favor is that if you lie persistently enough, eventually the fact checkers either get bored & exhausted or afraid of seeming “biased” because they keep pointing out every other sentence is a lie, and just let it slide without comment.