3 ms·

> The smarter problem would be changing the controls/UX for cars to be more like a hoverboard or balance/attention and gesture based than twisting a wheel aroun

> The smarter problem would be changing the controls/UX for cars to be more like a hoverboard or balance/attention and gesture based than twisting a wheel around, abstracting that to remote telemetry, and then work on using ML to replacing that remote driver. Go ahead and do it then. Show them how it is done. :) > Self driving cars is the perfect example of sprawling corporate groupthink, where someone has a poorly formed idea and everyone aligns around managing a response to it instead of asking whether it's the right problem to solve. Or maybe people thought about the problem for more than hot second and come to a different conclusion than you do. Your comment makes it clear that you are thinking about the joy of driving. You are talking about motorbikes, and UX, and being one with the car. I'm happy for you for that, and I hope you will find a lot of joy doing that. That on the other hand is not the problem self-driving cars are out to change. Every single thing in a modern world have traveled a ton. Transportation of raw materials, transportation of parts, transportation of the finished product. Then on top of that people can't make up their mind if they want to be in A or in B and seems to change location all the time. This is economically significant. Driving for joy, is a cool hobby, but is a tiny sliver of the whole economy. People and companies working on self driving cars believe that the time has come when we have enough compute that we can automate the act of driving and be at least as safe as vehicle driven by a human. If they are right, and these autonomous systems can be cheaper than human drivers then those who can make such machines can capture a portion of all the wages of every activity served by a commercial driver today. > mediated service where driving is reduced to a transaction. Do you think CEO of Walmart cares about the user experience of driving when they hire a small army to drive supply trucks to supply their stores? It is already reduced to a transaction. Uber is extremely successful. Do you think people care about a car becoming an extension of their body when they hail one? Currently people who drive because they enjoy it, and people who drive because they have to are all jumbled into one mass. If we have commercially viable self driving systems you will see who is driving for the joy of it, and who is driving to get stuff or themselves from A to B. I suspect we will see that it will become just an other hobby. Like going skiing or horse riding. Many will enjoy it, there will be probably a whole cottage industry catering to those who would like to "become one with the machine" and etc.

motohagiography 3y ago
> Go ahead and do it then. Show them how it is done. :) One of my weekend projects is a POC demo for how something like this works. Maybe I'll launch it here, or maybe I'll use it to establish a new autonomous region with an air defense perimeter, tbh, I haven't quite decided, but your insight is valuable. I'm saying what people think is a "self-driving car" is effectively a fully autonomous robot that has to navigate among humans, and presenting it as just solving a car operation problem is misleading. The problem they are solving is creating a virtual android to replace taxi, bus and truck drivers, which is surely a great and noble pursuit, but as a consumer product, it's the wrong problem. Some of the smartest people I have ever met have given it a lot of thought and disagree with me, but when you look at what every single other generalized robot product has succeeded at, it's in scaling human ability and in very limited tasks. Self-driving cars do neither of these things. It's not a car problem and using cars is a misleading constraint. Your maximum upside is some nuanced variation of a train or a streetcar on dedicated roads or paths/rails that is totally dependant on new public infrastructure to support it. Maybe you get self-driving golf carts and ATVs for limited contexts, just like other robots, but a general purpose one to replace drivers without dedicated lanes, air routes, or infrastructure is implausible. I'd focus on solving autonomous navigation in more narrow contexts like dedicated lanes and rails, flying, trails, and anything that doesn't have to deal with human obstacles. Otherwise, the failure mode of self driving cars is just a hunter/killer robot with limited commercial applications.
- krisoft 3y ago
  > One of my weekend projects is a POC demo for how something like this works. Nice! Please share it if you can. Sounds very exciting. > Maybe I'll launch it here, or maybe I'll use it to establish a new autonomous region with an air defense perimeter I say keep your options open and do both. ;)

motohagiography 3y ago

> Go ahead and do it then. Show them how it is done. :) One of my weekend projects is a POC demo for how something like this works. Maybe I'll launch it here, or maybe I'll use it to establish a new autonomous region with an air defense perimeter, tbh, I haven't quite decided, but your insight is valuable. I'm saying what people think is a "self-driving car" is effectively a fully autonomous robot that has to navigate among humans, and presenting it as just solving a car operation problem is misleading. The problem they are solving is creating a virtual android to replace taxi, bus and truck drivers, which is surely a great and noble pursuit, but as a consumer product, it's the wrong problem. Some of the smartest people I have ever met have given it a lot of thought and disagree with me, but when you look at what every single other generalized robot product has succeeded at, it's in scaling human ability and in very limited tasks. Self-driving cars do neither of these things. It's not a car problem and using cars is a misleading constraint. Your maximum upside is some nuanced variation of a train or a streetcar on dedicated roads or paths/rails that is totally dependant on new public infrastructure to support it. Maybe you get self-driving golf carts and ATVs for limited contexts, just like other robots, but a general purpose one to replace drivers without dedicated lanes, air routes, or infrastructure is implausible. I'd focus on solving autonomous navigation in more narrow contexts like dedicated lanes and rails, flying, trails, and anything that doesn't have to deal with human obstacles. Otherwise, the failure mode of self driving cars is just a hunter/killer robot with limited commercial applications.

krisoft 3y ago
> One of my weekend projects is a POC demo for how something like this works. Nice! Please share it if you can. Sounds very exciting. > Maybe I'll launch it here, or maybe I'll use it to establish a new autonomous region with an air defense perimeter I say keep your options open and do both. ;)