At least Elon was honest about the failure of self driving cars.

But saying, "Models are programmed to make stuff up", is a bit of a stretch.

So ai is programmed, its not conscious or thinking or anything like that.

The fundamental problem is that AI models are trained to reward guesswork, rather than the correct answer. Guessing might produce a superficially suitable answer. Telling users your AI can't find an answer is less satisfying.

I really can't wait until we have humanoid (self driving) ai teachers in our wonderful schools, Guessing superficially suitable answers, when our children ask them questions.

Lets teach our kids with ai. YAY!

Yep! thats going to go down well at the parents evenings.