In this Q/A session with Josh Hartung, founder and CEO of portfolio company PolySync, Josh shares his perspectives and outlook on the future of autonomous vehicles. PolySync is the operating middleware system for autonomous vehicles, connecting algorithms, sensors and actuators into plug-and-play apps.
What does the future of autonomous vehicle look like to you?
We think sensors and AI — these are the two big sticking points for the industry, and we're working through those; we’ve largely determined that it's miles, it's data. We have to get some refinement in these systems to make the choice in the trolley problem — to know whether we should hit the baby or the grandma. There’s just a period of refinement to handle all those edge cases and it's going to be a long one. And it requires a lot of miles before full autonomy can arrive.
What are the biggest hurdles to autonomous vehicles becoming mainstream?
Our concern is safety.
The question that we're focused on is once you've decided to take action, how do you guarantee that that will happen. How do you guarantee that when a computer goes up in smoke — when one of these high performance Ferraris of GPU processors goes up in smoke when you're taking a hard left hand turn — that everything still works? That’s when you need to make a critical decision — one that happens in a timely enough fashion to keep you safe. So our focus is on that kind of safety — the safety of knowing that your computer system will without a doubt work.
How do you see the industry as being ripe for disruption today?
We've seen rapid progress as the result of the hardware and software that has been refined to amazing levels in other industries. You know — web, mobile, IOT — all these things have put a cost pressure on super high performance computing resources. The level of developers — the rapidity of developers from Agile have given us things just so quickly. It really is amazing. We really think there's a ton of disruption in going back to basics and looking at: how do we get the same level of convenience and performance —speed of development —, but in a system that has people's lives at stake.
And there's big questions that come up when you ask that question. Is Agile development the right model? Does it make sense. Agile development was developed to get prototypes of customers fast so we could get feedback and refine our app or whatever — is that an appropriate tool when safety is critical? I'm not sure.
There's a bunch of these roadblocks that will be coming up that we have to have answers for and we're just starting to see the industry look at those. And I think that the answers are going to require some hard truths because there is no shortcut to safety. There's just a lot of rigor. I think maybe in terms of disruption, it's pretty clear that the companies that are first to market here are going to have a very significant lead, but I think that challenge of building a safe system is going to be the thing that prevents them from getting there at scale. So for the companies that figure out safety, I think are going to have a massive advantage and are going to get to be the disruptive players.
What is the current issue in the industry that you are helping to address and how are you doing so?
The thing that we're very focused on here at PolySync is what does the safety of the system look like. How do you build a safe autonomous vehicle from a computing standpoint. Much of the code and computing systems that have enabled the rapid progress of autonomous vehicles — this is sort of a catch 22 — much of the systems are effectively unsafe. They are not reliable. You don't want your fate in the hands of the Linux operating system or Python or something like that. And so, this is back to basics, what we have to do to define what a safe system is, and I think that there will be some hard lessons in that. We have different expectations of that. We have expectations of the type of rapidity and accessibility that these sophisticated web APIs have enabled that I think are unrealistic in the autonomous vehicle space. There’s just a level of rigor for safety critical systems that is extremely high and just has to be.