Are you pleased with the current progress/direction of AI?

I think there are a lot of cool AI demos out there but relatively few actually production quality AIs. Most of what passes for AI today seems to me like elaborate smoke and mirrors involving scripted dialogs with chat bots, recommendation bubbles of somewhat dubious quality (e.g. Netflix, Amazon, FB recommending stuff using what I call the more of the same algorithm), general misuse of AI algorithms to solve simple search & normalization problems, and generally a lot of off the shelf stuff being oversold as AI.

Progress by the likes of Google, OpenAI, etc. is very real, of course. Google particularly has a history of productizing things like image search, spearch recognition and generation, etc.

However, social media websites are overselling their ability to determine relevance by a very large degree and consistently favoring ad revenue over relevancy. Basically, they seem to mainly filter stuff out I might want to see in favor of stuff I don't want to see, or have been exposed to repeatedly. This reeks of amateur hour. They are trying to fit a one size fits all solution to a very diverse user base and are expecting sane results. Also a clear sign of a lack of vision. Smart would be recognizing what type of user I am and adapting to that. FB shows no signs of that kind of smartness. IMHO feed quality has degraded continuously since I started using it.

I find it baffling how FB on my phone remains blissfully unaware of the fact that I also have a desktop browser where I see stuff. It regularly puts stuff at the top that is more than a week old filtering out newer stuff in favor of that. It's the most basic signal ever: has the user seen post X; that combined with age should make those posts drop in relevancy dramatically. Any half decent search engine can be trivially configured to do that (no AI required); I know because I build search engines using Elasticsearch for a living.

Twitter seems to have imitated them and devalued their own product. The only time I do searches in twitter is when there's something going on right now. Twitter by default hides all the recent tweets these days in favor of something completely irrelevant that happened hours/weeks/months ago. Smart would be detecting that A) there's something going on right now (spike in related tweets), and B) that I searched for that because it is happening now. That's not how Twitter works anymore and arguably not even something that requires ML. It was more useful when it did not try to be smart.

I love AI. But I'll talk briefly about things I'm not liking:
1- Hype: I hate hype, it can ruin and diverge the most interesting thing and when the bubble bursts and people realize that the hyped thing is not a silver bullet, it goes back to being a niche thing that interests fewer people than it should.
2- Fear: Because people are apparently not understanding the definition of AI and its limitations, they're associating it with some crazy borderline conspiracy theory scenarios.
3- Military use: I think this one is self-explanatory.

Personally I think it's been a miracle for work loads like Map directions, and fronts like NLP and translations. However, I think it's been an absolute disaster when it comes to social media. Particularly in that space it has been used as a route of plausible deniability if the ML acts in an unintended way. "Even we don't know how it works!"