You can always make an argument that some new thing is bad by pointing out it is solving problems that were already solved or solving problems no one cares about.

That's not the issue. I'm not a luddite. The issue is that you can't rely on its answers. The accuracy varies wildly. If you trust it implicitly there's no way of telling what you end up with. Human learning process normally involves comparing information to previous information, some process of vetting, during which your brain "muscles" are exercised so they become better at it all the time. It's like being fed in bed and never getting out to do anything by yourself, and to top it off you don't even know if you're being fed correct information.