know AI in 10 years will.

That kind of the main problem: there is no indication that it will. I know one thing: current way LLM works, the chances that the problem of “lying” and “hallucinations”, will even be solved are slim to none. There could be some mechanism that works in tandem with it, but it doesn’t exist yet.
So most likely either we will collectively learn this fact and stop relying on this bullshit, which means there is a generation of kids who essentially skipped a learning phase, or we don’t learn this fact, and there will be a society of mindless zombies that are fed lies and random bullshit on a second-to-second basis.
Both cases are bleak, but the second one is nightmarish.