Elon Musk and Nick Bostrom were mocked for warning about the dangers of powerful future AI. I've said this here before, the primary use cases of machine learning right now, today, are hostile. That does not bode well for our future. We have very manipulative advertising, mass surveillance used to put ethnic and religious minorities in concentration camps in China, and in YouTube's case, recommending a lot of totally junk science to people who just want to learn new things. I'm afraid these four things probably impact way more people than translation systems and self driving vehicles right now.

I am completely in favor of the internet as an uncensored communications platform. However, in a war of ideas, there is a giant chasm between making a propaganda video of any sort available to watch in an archive and showing the stickiest propaganda video to those who are most likely to be retained watching it.

When I was at NIPS last year, the only person I heard get really angry about something, from an ethical standpoint, was relating to deep fakes. If you are a researcher or a project developer, you need to think really hard about what you are building, because soon you won't be the one building it.

edit: I don't think YouTube should solely be singled out for bad recommendations. Netflix has a lot of junk science on their platform and my assumption is their algorithms are bumping it to the top. In their case they have to proactively go out and pay for licensing for those videos, which is really inexcusable.