slappy
Member
The automakers are not responding, since AP2 is a bunch of promises at this point. Hardware is generally "easier" and there's an overestimation of what software can do to the point that people throw out statements like "It's just software. The engineers will figure it out." Software is wildly difficult; Microsoft, Google and Apple have significantly more software engineers than Tesla, but I'm not sure I'd want Siri, Cortona, or Google Now on the roads in their current state.
If you've followed AI at all, you'll realize that in the early 90's we were "on the cusp" of things like self-driving, but it never panned out since it's a wildly complex problem. A great use case is at a 4-way stop. The "rules" say the first car to arrive has right of way, but what happens if the first arriver waves you through? How do you differentiate a wave versus someone playing with their hair and not paying attention, or looking at their phone? Humans can rapidly interpret a friendly wave versus someone flipping the bird, and other items that are challenging for a machine. You can't just point an algorithm at this and assume it magically "learns" like a human, no matter how many miles, "neural nets," or teraflops you have.
My house has a half-dozen high definition cameras and nearly 40 sensors, yet it's still not smart enough to manage the HVAC and lighting with any "intelligence" beyond if X happens then turn on sensor Y. While "it's just software" in 40 years of industrial controls and home automation no one has solved this problem, arguably vastly simpler than fully autonomous driving.
My guess (and anyone that suggests they're making anything more than a guess is deluding themselves) is that we'll see fully autonomous highway driving in the next couple of years, perhaps even to the point that your steering wheel recedes into the dash like a scifi movie once you merge onto the highway. We also may see "autonomous corridors" in cities and regional roads similar to HOV lanes, where pedestrian access is manages and there's connected infrastructure or special signage. I'd guess we're 5+ years out from rolling through school parking lots and driving through snowstorms, but if you could move a decent portion of rush hour traffic to autonomous vehicles that zoom through special lanes at high speeds everyone would benefit.
I do personally believe there's a challenge with mass adoption in that Tesla's system is "good enough" to confuse people that it's more capable than it is, demonstrated aptly through ridiculous youtube videos of people sleeping and whatnot that we all lament. Managing a user interface and appropriate education is going to be a challenge that I believe Tesla and others are underestimating. There's also the looming "dark cloud" of liability that's yet to be fully tested. When an autonomous vehicle ploughs though a bus filled with pregnant volunteers from Doctors without Borders who are transporting extremely cute puppies to needy orphans and the public retribution starts flying, it will get ugly fast and that could set back the technology several months or even years.
If you've followed AI at all, you'll realize that in the early 90's we were "on the cusp" of things like self-driving, but it never panned out since it's a wildly complex problem. A great use case is at a 4-way stop. The "rules" say the first car to arrive has right of way, but what happens if the first arriver waves you through? How do you differentiate a wave versus someone playing with their hair and not paying attention, or looking at their phone? Humans can rapidly interpret a friendly wave versus someone flipping the bird, and other items that are challenging for a machine. You can't just point an algorithm at this and assume it magically "learns" like a human, no matter how many miles, "neural nets," or teraflops you have.
My house has a half-dozen high definition cameras and nearly 40 sensors, yet it's still not smart enough to manage the HVAC and lighting with any "intelligence" beyond if X happens then turn on sensor Y. While "it's just software" in 40 years of industrial controls and home automation no one has solved this problem, arguably vastly simpler than fully autonomous driving.
My guess (and anyone that suggests they're making anything more than a guess is deluding themselves) is that we'll see fully autonomous highway driving in the next couple of years, perhaps even to the point that your steering wheel recedes into the dash like a scifi movie once you merge onto the highway. We also may see "autonomous corridors" in cities and regional roads similar to HOV lanes, where pedestrian access is manages and there's connected infrastructure or special signage. I'd guess we're 5+ years out from rolling through school parking lots and driving through snowstorms, but if you could move a decent portion of rush hour traffic to autonomous vehicles that zoom through special lanes at high speeds everyone would benefit.
I do personally believe there's a challenge with mass adoption in that Tesla's system is "good enough" to confuse people that it's more capable than it is, demonstrated aptly through ridiculous youtube videos of people sleeping and whatnot that we all lament. Managing a user interface and appropriate education is going to be a challenge that I believe Tesla and others are underestimating. There's also the looming "dark cloud" of liability that's yet to be fully tested. When an autonomous vehicle ploughs though a bus filled with pregnant volunteers from Doctors without Borders who are transporting extremely cute puppies to needy orphans and the public retribution starts flying, it will get ugly fast and that could set back the technology several months or even years.
