Computer Vision News - March 2018

23 Xiaodi Hou - TuSimple When do you think the first sale will be? That will be the first half of this year. There are some things I cannot tell you more about, but keep watching, and you will see some interesting things this year. Already in 2018? We will definitely sell which means that we are carrying freight, but there are some regulations that we cannot foresee which don’t allow fully autonomous driving without a human in the vehicle. What we can do the first half of this year is that we can ship some goods regularly for some customers. Regularity is a very important thing because we’re not saying we’ll do it once and that’s all. This is a very big challenge. We will need to do that in 2018. We’ll start doing some trials earlier in the first quarter of the year for the first half of 2018. We’re going to do it regularly by the end of this year. I’m not sure if we can make it by June 2018. Can you tell us more about recruiting talents in this field? This is something that I’ve been asked several times, and I have a good answer for that. Basically, we have found that there is no PhD in autonomous driving so we can’t really hire someone who studied autonomous driving. That’s our biggest challenge. Nonetheless, we found that in academia people are using the same type of tools. One PhD training issue is the ability to solve new problems even from the starting point of defining the new problem and using the right tools to solve it. This is what a computer science and a computer vision PhD is about. What we wanted is to find those PhDs (and sometimes Masters) who can solve those problems efficiently. We are actually defining the new problem instead of just solving an existing problem, since in autonomous driving, the biggest problem is the problem definition. People are always using the wrong definitions and borrowing terms from academia to directly solve these problems, and that does not work. We have to step back and look at the entire problem on the more generic level and try to come up with a new definition of the problem. We define the problem and then we try to solve it with the existing tools that we have mastered during our education. Guest Computer Vision News “We are actually defining the new problem instead of just solving an existing problem, since in autonomous driving, the biggest problem is the problem definition.”