Microsoft released first version of VS in 98 and the version I used first was 2003. I feel like MS has missed on capturing user’s development behavior. They do capture a lot of telemetrics. This is all internal. From what I know, It’s used to find out how and which features of VS are you using and plan their roadmap accordingly.
VS knows how developers are writing code, the mistakes they are making and eventually the changes they are applying to fix them. VS could have built a knowledge graph so that if a dev makes a mistake again after a few months then it can suggest - a solution again, instead of the developer googling one more time.
ML (machine learning) has opened up a lot of new frontiers. We can build baseline models and extend them. A new way to cut down on plumbing so that we can focus on providing business solutions. I would keep an ML model but use it to provide code recommendations for SOLID and OOPS principles. And, imagine the code quality after that. Also, the code reviews would also improve substantially in terms of review time and the quality of suggestions. Intellicode is a very nimble step in that direction but has a very long way to go.
Finding a car or a human in a picture is very easy because we are just comparing the pixels but doing the semantic analysis on text is really complicated. There are off-the-shelf ML models available in the market, that can be leveraged for this purpose but they have limitations in terms of what you can do.
Definitely, interesting times for writing code and it’s getting more exciting every day.