4 ms·

And i though self driving car is a hard technological problem. Silly me. It's a design problem.

by minthd 11y ago

pjkundert 11y ago
Yup. We don't even have "cruise control" that is usable in the wintertime -- but we just placidly believe every yahoo who claims to have solved the problem of completely autonomous driving! Have we ever heard of Unit Testing? It is unlikely that my 10-line function is 100% correct. But, I just know that any random company's Autonomous Vehicle is certainly able to handle black ice in a white-out. What could possibly go wrong?
- comex 11y ago
  Unlike a privately owned autonomous vehicle, a taxi has very little need for environment flexibility. The article says they're planning to start in a "sunny city like Las Vegas", where there's unlikely to be much black ice... and even if they operate solely in such sunny cities for a decade, there's still enough market for them to grow huge. Of course, anyone who can make a vehicle that can handle white-outs will have an advantage, but I haven't heard even Google claiming to have a handle on that.
  - minthd 11y ago
    Let's be more realistic here. In any weather condition - this thing will live and die on reliability. To get to the needed reliability(say 10x better than humans, i.e one accident per 2.5 million miles you need to drive at least 25 million miles, but considering the variety of situations this needs to be tested against , maybe more than a billion miles). To solve this google has built a simulator simulating all the roads of california(4 million miles) with a variety of situations and is lobbying to get this approved as a way for testing. And that's just for testing. You still have to solve problem! Another data point - hyundai will be investing $70 billion in the next few years and wants to hire 7000 new engineers to work on their self driving car project. This is not startup land.

4 ms·

And i though self driving car is a hard technological problem. Silly me. It's a design problem.

by minthd 11y ago

And i though self driving car is a hard technological problem. Silly me. It's a design problem.

pjkundert 11y ago
Yup. We don't even have "cruise control" that is usable in the wintertime -- but we just placidly believe every yahoo who claims to have solved the problem of completely autonomous driving! Have we ever heard of Unit Testing? It is unlikely that my 10-line function is 100% correct. But, I just know that any random company's Autonomous Vehicle is certainly able to handle black ice in a white-out. What could possibly go wrong?
- comex 11y ago
  Unlike a privately owned autonomous vehicle, a taxi has very little need for environment flexibility. The article says they're planning to start in a "sunny city like Las Vegas", where there's unlikely to be much black ice... and even if they operate solely in such sunny cities for a decade, there's still enough market for them to grow huge. Of course, anyone who can make a vehicle that can handle white-outs will have an advantage, but I haven't heard even Google claiming to have a handle on that.
  - minthd 11y ago
    Let's be more realistic here. In any weather condition - this thing will live and die on reliability. To get to the needed reliability(say 10x better than humans, i.e one accident per 2.5 million miles you need to drive at least 25 million miles, but considering the variety of situations this needs to be tested against , maybe more than a billion miles). To solve this google has built a simulator simulating all the roads of california(4 million miles) with a variety of situations and is lobbying to get this approved as a way for testing. And that's just for testing. You still have to solve problem! Another data point - hyundai will be investing $70 billion in the next few years and wants to hire 7000 new engineers to work on their self driving car project. This is not startup land.