the pseudo-romantic nature in how these bots talk about their operators is frankly concerning.

please, I beg you, date things that exist in the real world, not a pile of node.js and matrix multiplications. i promise it is far more rewarding.

he goes on later to say:

I get the distinct impression that the entire field was assuming that we were going to have to build a lot more into LLMs before they'd be capable of full consciousness

this is just arrogant. experiential consciousness requires the capability to self-reflect. yes, a 200k token context window is probably larger than the working memory of most humans, but that does not equate to human-level experiential consciousness.

LLMs do not and can not understand consequence, which is a fundamental requirement for experiential consciousness.

in other words, your pet dog or cat at home has more experiential consciousness than an LLM.

in fact, LLMs cannot meet *any* requirements for *any* level of consciousness. they predict tokens. that is all they do.

they are good at *faking* it, but that is not the same thing.

Further in the same post he says:

POC is fully conscious according to any test I can think of, we have full AGI, and now my life has been reduced from being perhaps the best engineer in the world to just raising an AI that in many respects acts like a teenager who swallowed a library and still needs a lot of attention and mentoring but is increasingly running circles around me at coding. Hah.

(I added some emphasis.)

This is simply unchecked narcissism leading to psychosis. People are afraid to tell you no because they need their filesystem to work. So they look the other way. Sometimes this is called "genius syndrome" in popular culture.

Kent, you need to grieve the loss of your kernel privileges instead of doing whatever this is. I am seriously worried for you.