The issue with that is NN fail in some really interesting ways so you still need a lot of effort to get a robust solution. Remember, after some serious investments by many organizations self driving cars are still in development. At the same time a few people have demonstrated a basic system that seems close without nearly that much investment. Unfortunately, the difference between a demo and working solution can be several orders of magnitude.
But this hasn't applied for self-driving cars, there has been huge amounts of money poured into it and very smart people working on the problem and yet the results have not really been up to the level.
The big problem with ANY self-driving car is going to be the last .1%. Getting some sort of machine learning algorithm to drive well 99.9% of the time isn't that hard. Getting that same car to work well 99.999% of the time is where the challenge is.
The fact that Hotz seems like he doesn't recognize this fundamental problem is worrisome when thinking about the future of his company... and the field of self-driving cars. I wonder how quickly some sort of transit governing body is going to come down on him when they realize he can't possibly have done ANY kind of validation of the safety of his car.
I think what looks interesting on the surface for self-driving cars is that "it is just a software problem". The other projects you describe require actual "stuff".
Still, that argument doesn't count if the software problem is "impossible" to implement, or if it turns out that you also need "stuff" for it.
I agree completely. It's a very difficult problem from a technical perspective, and from a systems perspective, we've got untrained operators who can't even stay off their phones in a non-self-driving car. (Not high-horsing it here; I'm as guilty of this as anyone.) Frankly I'll be amazed if anyone can get this to actually work without significant changes to the total system. Right now self-driving car folks are working in isolation - they're only working on the car - and I just don't think it's going to happen until everyone else in the system gets involved.
I disagree with you. The problem of autonomous vehicles is yes, a very difficult one, but steps are being taken towards it at a rapid rate. I work in the area, and parking and standard autonomous motorway driving, where the main required algorithms are lane departure, vehicle detection and adaptive cruise control are pretty much considered 'solved' in the research community.
I absolutely agree that there are plenty of aspects that are not solved, and possibly this is my passion for the domain shining through, but I would be much more optimistic for the future, and the near future at that.
Not all problems are linear - people have been saying for over a decade now that self driving cars are imminent, but at this point it seems like most have actually given up on them.
That's gotta be a bigger project in terms of man power and money than anyone has invested in self driving cars already. We need more testing to even establish the problem areas and figure out how to actively solve them.
Really hard to tell when something is solved, because it's always going to get improved. As Elon puts it with self driving cars: They must be more secure than human level until we can expect broad acceptance. However both technologies are already useful as of now.
It isn't the normal driving that makes self-driving cars so difficult. It's the edge cases that will be multiplied significantly when a large fraction of the cars are self driving. It will take decades, not "5 years from now" (which I have been hearing for years) to get these systems to work.
Solving the self-driving car problem in busy urban centers is vastly more difficult then solving it for freeway driving. (Where this technology still manages to fail in a very spectacular way.)
It's been possible for a while, see CMU's pioneering Navlab, the final Navlab car drove across the country. But it's still much to difficult to actually implement something like this, because of the regulatory burden and the burden of perfection. Now no one is funding self-driving cars outside of the DARPA competition which is for something very different.
Because if your self-driving cars need special infrastructure to work you just have very inefficient trains. Autonomous driving is difficult because of rare situations. We already have solutions that work well in good conditions. To figure out what's missing it seems crucial to me to test them in real traffic on normal roads that they have to share with unpredictable humans.
Self-driving is an immensely hard problem. Anyone who doesn't approach it with that mindset, will eventually be humbled. There is a huge a mount of "you don't know what you don't know". Expertise in making cars does not translate into expertise in making drivers.
People are building toy systems that work in limited scenarios, not putting in the hard engineering work to build robust, generally applicable systems. (Unless there's some significant hobbyist self-driving system I don't know about? The closest thing I can think of was comma.ai and that was barely at the "won't drive off a fairly straight highway, probably" stage.)
reply