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?
John Roese: We’re aging them out. We’re producing new ones but we’re also creating better tools so that a typical program is using dev ops model and modern computer frameworks and code generation. We aren’t producing them fast enough. If you go into China, it’s a slightly different story. They are accelerating in building their baseline, but eventually, they will plateau.
If you look at places like India, it’s the same thing. They’re growing their base and they have a great talent pool. Eventually it will plateau. Just like the US, you have a fixed number of people. We can’t rely on just minting more computer scientists to do the work. We have to make sure they have an environment to keep their skills fresh.
Most importantly, we have to build underlying technologies that allow a single computer scientists to be able to do more work in the same amount of time. That’s where things like modern cloud frameworks like Kubernetes give us an ability to do more. Imagine if you had to build your own AI framework to do AI work. You would never be able to get anything done.
Sramana Mitra: Just to follow-through on your point about modernizing these curriculum, are those curriculum tying into the kinds of augmentation framework that you’re taking about. Are they aware? Are we factoring them into the curriculum? Are they training people in those frameworks?
John Roese: Not to be snarky, but there’s one easy litmus test if they are or aren’t. If the computer science curriculum for AI and ML is about building Alexa functions and tasks, that’s not a sufficient curriculum. If it’s about working with TensorFlow and Caffe, that’s more sufficient.
A good example of where this is happening is actually in China. They have a thing called AI 2.0 where they made huge investments in their universities. If you look at the curriculum, it’s not about feel-good AI. It’s about learning the hardcore underlying technologies. I was talking to a gentleman who’s the CIO of a western bank out of Boston and who’s based in China.
The reason he’s in China is that the bank’s transformation and future is entirely based on being to exploit data in new ways using machine intelligence. The curriculum and investment in China to create and cultivate the talent pool at these low-levels is actually quite serious. It’s not happening universally in the world.
We have to be careful to not think that having a smart speaker talk to us using some high-level API’s is sufficient answer versus the skill set we need to build an AI-enabled business. We do have to be careful because there is a spectrum right now. It’s easy to tell the ones that are good and are the right formula because they operate at the low level. The other side is people who are using somebody else’s technology.
Real value is created when you build something original and when you work at the low level. You want to create a business and monetize it, create something foundational. That means we need to operate at the foundational levels of these AI frameworks which are quite good given the open source initiatives and the collaborative ecosystem that we have today, but they’re not universally and consistently adopted across the educational sector or the industry.