 Alright guys, welcome back. It sounds like the rooms weren't pretty well, so I'm glad you guys have some good discussion. So now let's open up the floor for some questions. So we have our first question. It's 90% of the road accidents are caused that happen are due to human error. So what is the autonomous vehicle car manufacturer's responsibility in this case, especially if it's a human error that causes the autonomous vehicle to crash? That's a great question. I think it is the manufacturer's responsibility, and that's why I give talks almost every week about the importance of adding driver-facing cameras to cars. So the car needs to know what the driver is doing, because most of the error is caused by human beings. And so you want to use computer vision and machine learning methods to detect everything the driver is doing in terms of where they're looking. Are they looking at their smartphone? Are they completely distracted? Are they drowsy? asleep? Are they emotions or frustration? Come to the play of things. And then just body pose estimation. That's just a big problem in computer vision. Detecting the different hands, the shoulders, the elbows. Detecting if their hands are on the wheel or not. Detecting if they're in the seat at all or not. With some of these autonomous vehicles, you're able to leave the standard central aligning position. And so the car needs to know that. And that for that, we need the cameras. Without that, right now, the only way the car knows anything about the human occupant is touch sensors on the steering wheel, and sometimes touch sensors, pressure sensors on the seats. That's all the car knows about you. We believe that the car should know more about your preference as your style, the patterns of drowsiness, glass allocation, smartphone use, all these things need a camera. Yeah, absolutely. So then that kind of comes up with a follow up question. What happens to another driver on the road? Just switches lines and hits your car. Then where's the responsibility fall? How is the autonomous vehicle really supposed to react in these situations? So that's a great question. It touches always these discussions with ethical questions, because ultimately, when as a society, we're a little bit uncomfortable with an AI system, making a decision that may that's risky. So as human beings, when we get on the road and just in our daily existence, take risks all the time, especially during driving, emerging lanes and changing lanes, there is some probability that that was on the fatality for the driver. There's always some it's not zero. And but with the moment we start discussing when an AI system makes that decision, and that calculates a fatality, making a decision that includes the possibility of fatality would get uncomfortable. And liability starts coming into play. For me, from an engineering perspective, from a perspective of wanting to save lives, all that matters is that systems, these AI systems have a significant decrease in the number of fatalities and injuries. So if we can decrease the number from 38,000 to 10,000, that's amazing. And the fact that sometimes the AI system results in a fatality or injury is tragic and horrible. But if the final result is that in some, the number of fatalities is lower, that's amazing. Yeah, so it's always interesting to me because a lot of people will say there's an AI that got an accident. It's terrible. But it turns out these guys have driven significantly more miles than any human has and not gotten an accident. And so in terms of ability to actually react with road, it's actually much better. But humans can't think of it like that. They see that the computer made a mistake. They have a lot less ability to accept computer mistakes. Yeah, and that's something we'll struggle, I think, of this entire century, folks will say now could be part of the conversation for decades to come as our relationships to robots. When they enter the home, you have Amazon Alexa, which you work with closely. When these robots, natural language systems and then humanoid robots enter the home, to which degree do we trust them, to which degree when they result in an accident or injury or death in the case of cars, how do we emotionally and philosophically deal with that idea? It's an open question. It's totally new and exciting. That's awesome. All right, so we have another question. And it is, what is the required infrastructure and ecosystem that should accompany autonomous vehicles and driving? And so then the kind of follow up on this is can developing countries benefit from this? Or is this mostly something that will kind of sit and develop countries for a while? So there is so many amazing solutions on the infrastructure side that will help autonomous vehicles. The big challenge of connected vehicles, ability for vehicles to communicate with infrastructure is would help make the autonomous driving task so much easier will result in so many safe lives. But the problem is, there's bureaucracies involved. There's a lot of red tape. The thing is, if you think of the traffic light, the our current traffic lights in America and the developing countries and everywhere in the world are really dumb. From the perspective of AI, they have very little artificial intelligence inside them. And that's just unacceptable. The the ability for traffic lights to make intelligent decisions that help people is incredible by requires bureaucracy requires a lot of investment infrastructure. So I tend to sort of focus on the systems of how to build these AI systems because policy, this is the thing with cars is because government is so deeply involved in as they should be because human life is a stake. A lot of the stuff, a lot of the decisions you make with infrastructure investment and vehicle requirements, in terms of AI, have to go through a long process of getting permission of getting getting everybody on board. And the society was still struggling with the role of AI. And so that all come into play in the infrastructure while being an awesome way to solve the self driving car problem is seems like a 10 2030 year out kind of progress as opposed to what we want is impact today. Absolutely. So kind of a nice follow up for that. So a lot of us actually have long commutes to work, especially in Boston, I know other cities around the world. A lot of the students will have a long commute to their work as well. Can you see artificial intelligence improving this in the next 100 years? Yes, absolutely. The commute to work the traffic. There's two things where I can have a giant role in helping driving is traffic and highway travel long distance highway travel. They want to take a road trip to Florida to visit your friends, you know, 90%, 95% of those miles are driven by interstate. And for most of us, that's not fun driving. AI systems. This is what we focused on are really good at that. Also traffic where there's vehicles all around you. Detecting vehicles is very easy with computer vision with radar and LiDAR. So traffic situation of just the car taking control, saying sit back and listen to some good music, some Led Zeppelin or whatever you're into. That's much that's technologically available. And I think in the next few years, more and more AI will make your traffic make your morning commute much easier. That's awesome. I'm looking forward to that. Alright, so this is a good question. So let's say someone has an emergency. Is there a way that they can override their AI system? Or is that something that we should have programmed in so they can quickly get to a hospital or somewhere? Or should the AI balance that out and make it a safer trip for them? Because their chance of living might be higher and drive to those types of things? How do we how do we think about this ethically? And then what is the current kind of thought process in that right now? The current thought process is AI only helps in a completely secondary way to help in various aspects of your life in driving. It's an assistive technology. It kind of it kind of monitors what the your performance on the driving task and just suggests improvements. There are there are either warnings or their suggestions about how to control the vehicle. You human always overrides. If you're in an emergency situation, you need to get somewhere, you're being always overrides with the control of the vehicle. It's very unlikely that in the foreseeable future, human being will not be allowed to override. And the other approach from a technical perspective of how we build AI systems that control vehicles is that they're always in a very boring and sometimes frustrating way, cautious and safe. So current our current AI approaches are cautious, careful, very conservative. They do not want to make mistakes. As human beings are always in a hurry, always late, we're always want something now and everything is frustrating to us. So for now, AI is the calm voice that kind of takes care of the boring bits. And then human beings are the ones that take control and they're just annoyed with traffic and they want to go on the opposite lane to pass the vehicles in front of them. So for now, humans always always come first and always take control. So what impact can you see for this in the trucking industry in the longer halls for these trucks? What do we see with that? Like what will happen will be the process. So this is probably on the technological side in terms of supply chain, I think would be the biggest near term impact is the trucking, automating the transport on interstate highways, transport of goods. Because the control the lateral and longitudinal control of the vehicle on the interstate is much, much easier than in urban sort of last mile delivery. So the long distance travel is as much easier. And then there's also the ability to, for multiple vehicles to follow each other on the highway, to form these long lines of vehicle where the perhaps the first vehicles controlled by a human, but all the ones after are controlled autonomously. That's another possibility. So I think in the neck comes a train of yes, okay, exactly, but a an autonomously assembled one. And that's much more responsive to the place where those goods need to be delivered and so on. So a train is very fixed to railroad tracks with our highway system in the United States, at least, that you can be you can think of how you can optimize delivery of goods in a much more flexible way when those trains self assemble self assemble and arbitrary way. Absolutely. So one thing that I can see with this is that you'll have a trucker almost as a port person. And he'll actually take the truck through the city, he'll get it to the highway, he might get out and then the truck will find one of these almost highway groupings. Yes. And then at the other end, it'll pull over and then maybe a trucker will get back in and drive it through the city again. Absolutely. Is that kind of a thought process? That's the thought process. And technologically speaking, we're the closest to that. And I think the technology is already there. Okay. The next challenge is policy and our public acceptance of the idea of a unmanned truck, a truck without a human driver, sharing the highway with human drivers. Absolutely. Hey, so I think we're going to wrap this up now. So it's real pleasure, Lex. And we look forward to learning the more learning the new things that you're doing and hearing more about in the future. Great. Yes, thank you so much.