Avoiding Negative Side Effects: Concrete Problems in AI Safety part 1

Channel:Robert Miles AI Safety · Length:9:33 · Published:2017-06-18 · Views:161080

Watch on YouTube

Summary

Robert Miles introduces the first of several videos on the paper 'Concrete Problems in AI Safety', focusing on how AI systems can cause accidental negative side effects. It explains the problem in accessible terms.

Why HN recommends it

HN recommends Robert Miles' channel as a clear starting point for people curious about AI safety.

For those curious about AI safety I heavily recommend Robert Miles channel on YouTube. — HN, 2023

AI safety, AGI & impact

Subtitles (excerpt)

Hi, I just finished recording a new video for Computerphile where I talk about this paper, Concrete Problems in AI Safety. I'll put a link in the doobly-doo to the Computerphile video when that comes out. Here's a quick recap of that before we get into this video. AI can cause us all kinds of problems and just recently people have started to get serious about researching ways to make AI safer. A lot of the AI safety concerns are kind of science fiction sounding, problems that could happen with very powerful AI systems that might be a long way off. This makes those problems kind of difficult to study because we don't know what those future AI systems would be like, but there are similar problems with AI systems that are in development today or even out there operating in the real world right now. This paper points to five problems which we can get started working on now that will help us with current AI systems and will hopefully also help us with the AI systems of the future. The Computerphile video gives a quick overview of the five problems laid out in the paper and this video is just about the first of those problems, avoiding negative side effects. I think I'm going to do one video on each of these and

On Hacker News

For those curious about AI safety I heavily recommend Robert Miles channel on YouTube. He has lot of good content describing out the challenges we face when trying to create safe AI. Here is one potential starting point: [VIDEO] — jononor, 2023-01
This highlights an interesting general point - in many situations, there is no simple safe fallback policy. On a highway, an emergency stop is not safe. This is a general problem in AI safety and is covered nicely in this youtube video, as well as the paper referenced there - [VIDEO] — nmca, 2018-05
Re: alignment, see the "Where are we" part of this talk: [LINK] See also the AI safety problems for concrete problems that people are currently working on - as in with code and real simulations, not "just" math and thinking: [VIDEO] Re: scaling up. That's called intelligence explosion, there are a lot of write ups about it, but the gist of it is just that by simply increasing working memory size,… — pas, 2018-11

Related videos