Oh, I agree with that. I’m not arguing at all that AI will be “done” in 2–3 years. It turns out there is still fundamental database research happening, so I think the research roadmap for AI will continue to be fruitful for decades. All I’m arguing is that every important piece of software we write will need to use AI techniques to realize its full potential.
Machine learning will get into all important software we write, so totally makes sense for Apple to give iOS developers a set of APIs to add ML into their apps.
Andrew: thanks for the comment. Yes, let’s focus on building useful and good products (in both senses of “good”) as the distinctions between machine intelligence and human intelligence get blurrier and blurrier.