self-driving cars turned out to be harder than people expected
As someone with an electrical engineering degree, a fair amount of experience programming computers, and who used to work in engineering at a GM facility where we designed and built automotive subsystems, I predicted that self-driving cars would be very hard, and am not surprised at the slow progress. I'm confident that they are coming though, and the sooner the better, as my daughter, who is blind, would benefit greatly, as will many other people with disabilities.
Some good info in this article, but I think the author is still overly optimistic. I'm with the Volkswagen CEO. I don't think fully autonomous vehicles will ever happen unless we completely re-engineer and rebuild every single bit of road infrastructure in the entire country in order to facilitate them, which I certainly don't think will happen.
Writing software to operate a vehicle and avoid objects is almost trivial (which is what made the fatal Uber crash so disturbing). The difficult part is trying to come up with a system that will allow a car to safely navigate among pedestrians, other drivers and random infrastructure anomalies without having any actual understanding of what it means to be driving a car. Computers cannot think. It's absolutely not a given that "more data" is an answer to anything.
As someone with an electrical engineering degree, a fair amount of experience programming computers, and who used to work in engineering at a GM facility where we designed and built automotive subsystems, I predicted that self-driving cars would be very hard, and am not surprised at the slow progress. I'm confident that they are coming though, and the sooner the better, as my daughter, who is blind, would benefit greatly, as will many other people with disabilities.
Some good info in this article, but I think the author is still overly optimistic. I'm with the Volkswagen CEO. I don't think fully autonomous vehicles will ever happen unless we completely re-engineer and rebuild every single bit of road infrastructure in the entire country in order to facilitate them, which I certainly don't think will happen.
Writing software to operate a vehicle and avoid objects is almost trivial (which is what made the fatal Uber crash so disturbing). The difficult part is trying to come up with a system that will allow a car to safely navigate among pedestrians, other drivers and random infrastructure anomalies without having any actual understanding of what it means to be driving a car. Computers cannot think. It's absolutely not a given that "more data" is an answer to anything.
Great recent short video from Tom Scott that highlights the underlying problem: The Sentences Computers Can't Understand, But Humans Can