The Future of AI Prediction

I think the future here is something similar to GitHub Co-pilot, where you have a machine learning system that is trained on lots of code. And we've taught it how to, when the user is in a partially complete state, which you can get training data for by taking a valid file and cutting out pieces of it. Then train it to predict where the closing paren should go. That's interesting. So then it becomes somewhat of a special case of autocomplete, where you're just trying to use training data to predict what do they probably mean here? And hopefully it's right most of the time. And if it's not, you can fix it.