Thought Leaders in Artificial Intelligence: Steve Scott, CTO of Cray (Part 5)

Sramana Mitra: Do you have any thoughts on this problem that is being discussed nowadays? AI is a bit of a black box and all these biases that are creeping into AI are going to drive society in the next several decades. We don’t really have a very good understanding of what really the AI is doing in a lot of domains and a lot of AI applications. How does the world deal with that?

Steve Scott: That is a super good question and it’s very real. I mentioned before that the marriage of AI and traditional simulations can help address that. People are very leery to take what was done in simulations and just replace it wholesale with a deep neural net because you don’t really know if you can trust what’s going on and you lose some insight into what’s happening. You can’t peer inside and understand it.

If you marry the two together and do things like preconditioning a traditional simulation with AI, you can address some of those issues with traditional modeling. We’re very aware of that in the high-performance computing space. It’s much more pernicious in these areas that have an impact on society where you’re making policy decisions. Who should get health insurance or home insurance? It’s really concerning to do that in a black box way.

There are a lot of researchers that are trying to work on better understanding what’s going on inside the neural net and finding ways to query it and validate that it’s doing something reasonable and have it explain its action so to speak. Other than that, this is an area that’s going to require concerned scientists and ethicists to be involved. You fast forward of course.

The even larger concern is, we are likely some decades away from having extremely broad capable general artificial intelligence. There you have to start worrying about the existential risks to humanity itself. There are people that have raised alarm bells about them. I think it’s very real. We’re decades away from that. It’ll provide a lot of good in the meantime, but it’s something that we’re going to have to pay real good attention to.

Sramana Mitra: We’re decades away; not centuries away.

Steve Scott: Exactly. We’re getting near the end of CMOS technology, which is the foundation for almost all general-purpose computing. There’s not any guarantee that we’re going to find something to replace CMOS when Moore’s Law peters out. It’s possible that computational performance will substantially slow down. I tend to be optimistic. I think we’re going to continue to get this exponential improvement of performance. It’ll happen then.

We don’t understand how the human brain works fully. There are more questions than answers. If you look at the trajectory to the point where we get to have orders of magnitude of more computational performance than the human brain, we’re still orders of magnitudes less. You fast forward several decades. I think it’s very likely that we’re going to have this general artificial intelligence. The future could either be really good or could be really, really bad.

Sramana Mitra: If you’re interested, you may want to look at this series on my blog called Man and Superman. It explores all these issues in quite a bit of depth. If you want to contribute to that series, you’re very welcome to do so.

Terrific. It was nice talking with you.

This segment is part 5 in the series : Thought Leaders in Artificial Intelligence: Steve Scott, CTO of Cray
1 2 3 4 5