Why did AI do so well?

I can see why people might have expected AI to excel in academia. But people overestimated TTS, and predicted Voice AI would fail. I guess I'm wondering why they predicted it would fail. The inability of Voice AIs to learn and recall anything from training has always been on display to me, as someone who has largely used voice AIs. You can ask them 5 questions, and they won't ever know that you have, because it is not a memory. You can have a conversation with it, and then bring it back up a day later from a different machine, and it'll have no memory of anything you said. <br><br>How did people not see this? Can we make a list of predictions that Voice AI did not live up to?