Interestingly, I thought Historian Mr Bret Bevereaux covered it better than any other places I’ve seen do a non-technical deep dive:
And I think where this tech has gotten is really impressive(despite being basically just a scaled up version of what I was learning about 20 years ago) and there are a lot of really excellent applications that will actually benefit us from something being able to pick up non-obvious patterns from giant data sets. And like you said, it’ll be good to assist real thinking humans but won’t replace them without disastrous results.
Because 90% of those 90% have the capability of doing it, they just choose not to because the internet lets them turn their brains off and be awful or they have some other motivation to bullshit. But the point is, I trust the output of one of these AIs about the same as a random Tweet somebody posts. Maybe it’s true, maybe it’s not, I’ll need to dig deeper to validate it.