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. >>>

Thought Leaders in Artificial Intelligence: Steve Scott, CTO of Cray (Part 4)

Steve Scott: With the advent of GPU computing, deep neural nets started to become enabled to the point where you could get good enough performance so that you could really do useful things with them. GPU computing is the application of the processors that were designed for highly parallel tasks of painting triangles on the screen for rendering graphics in real time.

It turns out you can use all those parallel functional units for doing normal computation. GPUs are the first processors with a single processor level powerful enough to do meaningful deep neural net. We could previously do it on very large systems, but that limits you to a small number of people who have these large systems. GPUs were the first ones that could do it on a single desktop basis at a useful scale. >>>

Thought Leaders in Artificial Intelligence: Steve Scott, CTO of Cray (Part 3)

Sramana Mitra: Can you give an example?

Steve Scott: If you think about deep neural networks in particular, there’s training and there’s inference. Training is the learning part where you take a bunch of data and based on that, you train a model to be able to provide some function. Inference, of course, is using that model that has done the learning already to make decisions on new data. The inference problem is sort of a throughput problem.

Once you’ve got the model designed, you can run lots of data through the model and make ad decisions very quickly. The training problem itself takes a lot of compute and a lot of communication. This is the sort of thing that a Cray system >>>

Thought Leaders in Artificial Intelligence: Steve Scott, CTO of Cray (Part 2)

Steve Scott: The way people have used Cray and other high-performance supercomputers is, you have a bunch of equations that present a model for the natural world whether that’s equation of airflow across an airplane wing or equations dictating the molecular dynamics involved in drug discovery.

You iteratively solve these equations spread across these points in space. You’re figuring out what goes on by solving these large numbers of equations that represent the real world. In AI, we have a different way of calculating results. Deep neural networks are more of data-driven versus mathematically-driven techniques where you have these layers of of artificial neurons. They’re taking inputs in. >>>

Thought Leaders in Artificial Intelligence: Steve Scott, CTO of Cray (Part 1)

Steve takes us deep into the field of high performance computing and how AI is impacting it.

Sramana Mitra: Let’s start by having you introduce yourself a little bit as well as Cray’s activities in the domain of AI currently.

Steve Scott: I’m the Chief Technology Officer at Cray. I’ve been with Cray most of my career. I actually was a summer intern here for two years during grad school and have been here ever since except for a three-year walkabout I took a few years back. I went and spent two years as the CTO of the Tesla GPU computing Group at Nvidia and spent a year in the platforms group at Google and then came back in 2014. >>>

Thought Leaders in Artificial Intelligence: John Roese, Global CTO of Dell EMC (Part 7)

Sramana Mitra: The domain that I feel very bullish about is healthcare. There are vast swaths of the world that have no access to good healthcare. I think AI can turn that around.

John Roese: Absolutely. In the mid 2000’s, I was the CTO of a company called Nortel in Canada. I was on the Board of Directors of an initiative called One Laptop Per Child. The vision was not about building a cheap computer. The vision was, if we can get, at least, computer access to the developing world, we can bring the transformation.

We drove down the cost of computers. We completely created environments where we could start to think about having computers >>>

Thought Leaders in Artificial Intelligence: John Roese, Global CTO of Dell EMC (Part 6)

Sramana Mitra: I wrote a series called Man and Superman. It makes a point that we’re going to a society where people who have the higher-order skillsets are going to thrive and everybody else is going to struggle. That gap is going to grow. The obvious question is how do you create more superman kind of skillset.

John Roese: I’m a bit more optimistic than some people are because I can see all of the areas where AI is having a positive impact on job creation, but there is a bias towards high-skill, high-function jobs. It has a very significant impact on more manual labor service-type of roles. If you look at it holistically, we probably are going to create more jobs than we’re going to lose. It’s just biased towards the high functioning areas. >>>

Thought Leaders in Artificial Intelligence: John Roese, Global CTO of Dell EMC (Part 5)

Sramana Mitra: How big is this pool?

John Roese: There’s an unfortunate statistic that we tracked from many years. The pool is not growing in terms of the total number of potential people who can do this work in many markets. In the United States, there were about two million professional computer scientists about 20 years ago. Even today, it’s about two million which is disturbing.

Sramana Mitra: Why? >>>