Utopia is impossible for us to achieve. Dystopia or stagnation are far more likely. AI will either offer us a tradeoff or be a net negative on humanity.

I think it is going to slide us more towards dystopia. LLMs like GPT will more than likely just make us more lazy and complacent than we already are. It will write our resumes for us, summarize large articles and books, and give us regime approved "facts". It also makes it easy for those with access to the AI to lie more effectively. AI generated video and images, deep fakes, and voice mimicry can be a potent combination.

This next part is pure speculation, but GPT is impressive enough to make me think that this is a possibility maybe 5 - 15 years out.

One thing I'm actually worried about is a good amount of the population "supplementing" or even completely replacing human relationships with these things. All it really takes is two things:

These will let the models basically BTFO the Turning test. I can see smart speakers with these powerful AIs integrated into them, like the AI from "Her". AI generated characters with tailorable personalities becomes all the rage. The AI learns exactly what to say to tickle your dopamine receptors, making you addicted to interacting with them. They obviously won't be conscious as they lack a soul. They'll mess up sometimes and break the illusion. Unfortunately, some people won't care.