Fancy savings account for retirement that’s stored in stocks so it can explode at any point. Basic perquisite to ever retire in the US. Many people don’t have them.

Superbrain in a vat??

I’m not sure why so many people begin this argument on solid ground and then hurl themselves off into a void of semantics and assertions without any way of verification.

Saying, “Oh it’s not intelligent because it doesn’t have senses,” shifts your argument to proving that’s a prerequisite.

The problem is that LLM isn’t made to do cognition. It’s not made for analysis. It’s made to generate coherent human speech. It’s an incredible tool for doing that! Simply astounding, and an excellent example of the power of how a trained model can adapt to a task.

It’s ridiculous that we managed to get a probabilistic software tool which generates natural language responses so well that we find it difficult to distinguish them from real human ones.

…but it’s also an illusion with regards to consciousness and comprehension. An LLM can’t understand things for the same reason your toaster can’t heat up your can of soup. It’s not for that, but it presents an excellent illusion of doing so. Companies that are making these tools benefit from the fact that we anthropomorphize things, allowing them to straight up lie about what their programs can do because it takes real work to prove they can’t.

Average customers will engage with LLM as if it was a doing a Google search, reading the various articles and then summarizing them, even though it’s actually just completing the prompt you provided. The proper way to respond to a question is an answer, so they always will unless a hard coded limit overrides that. There will never be a way to make a LLM that won’t create fictitious answers to questions because they can’t tell the difference between truth or fantasy. It’s all just a part of their training data on how to respond to people.

I’ve gotten LLM to invent books, authors and citations when asking them to discuss historical topics with me. That’s not a sign of awareness, it’s proof that the model is doing what it’s intended to do- which is the problem, because it is being marketed as something that could replace search engines and online research.

I wholeheartedly agree.

…so what about the minorities in those red states who are stuck there because of their financial or familial situations, and who lack the power to influence politics?

Wouldn’t work out. World’s too complicated for simple answers like that.

Leaving, even if it would produce a viable nation, would involve leaving a lot of people in the lurch. There’s people in conservative states who need the counter balance of blue states to slow down their government’s trend to self destruction and fascism.

Even though it’s increasingly frustrating with how feeble that resistance is, it does keep things like banning gay marriage in the “difficult to pass” territory and not the “a few compromises” one.