Hacker News Reader

Comment by _bxg1 | Hacker News Reader

After watching the self-driving car hype crash and burn (so to speak) over the past 5 years, I wouldn't bet on having fully-autonomous vehicles in the next 20. In fact I don't think we'll have them until they no longer have to coexist with human drivers, which could be a chicken-and-egg scenario.

View on HN

From Tesla's earnings report:

"We enabled stop sign and traffic light recognition and braking for our Early Access Program users at the end of Q1 and to the wider public in April 2020. As we have done with previous releases of major new features, drivers will be required to confirm their attention in order to continue. Once enough real-world data is collected, the system will become more capable and our vehicles will continue driving through intersections without a confirmation."

Source: https://ir.tesla.com/static-files/c1723af4-ffda-4881-ae12-b6...

Sounds like great progress to me!

People have been claiming it's six months away for half a decade. I'll believe it when I see it.

One very real effect of all of this research, though, is that if we did end up with an autonomous-only environment (dedicated lanes?), we wouldn't need special infrastructure - as some have suggested in the past - or a new software protocol to get all the different brands to coordinate. Probably even the ones we have today could just follow the normal rules and be sane drivers and would do perfectly. Human drivers are the only problem we can't solve.

My town has never fixed a stop sign that’s angled toward the wrong road. Humans just assume it’s for the smaller road. I’m not sure how a self driving car will ever figure it out.

People had similar concerns when motorcars first became available. i.e., my horse can jump over obstacles that motorcars never will, and so on. Eventually humans made roads specifically for motorcars.

That's exactly my point: we'll have self-driving cars when they have their own space. Not before, even given 20 years of progress. Maybe we'll get that dedicated space before then, maybe not. It's a much bigger socioeconomic hurdle and may or may not actually be worth it financially.

Maybe there's a partial solution though: drive your car manually to the next onramp from where on it drives automatically until you arrive in the vicinity of your workplace from where on you'll have to drive manually again for a bit. So sort of like the dedicated bus lanes we already have nowadays. If your car doesn't have autopilot, you're not allowed to use them. If your car has autopilot and detects a car without it in front of it, your car falls back to manual mode...

great question - I assume a set of local knowledge will be added by curators or lawsuits over time... Large tech companies have a lot more pull with government than individual drivers, they might even get that stop sign fixed!

Well, if the roads would be optimized for autonomous cars then any signs on them would be used as a fallback. Autonomous cars would mostly rely on something like road databases, active "stop beacons" or passive "stop markers" (like reading a passive RFID tag)

Sometimes even human drivers can't figure out what gesturing police and paramedics and construction workers want us to do. They will need a way to mass override "follow the normal rules", or owner or mturk takeover.

most of those situations are there because of human drivers in the first place.

Why would there be gesturing police and paramedics on the road if there are no accidents? And construction is easy. You just update the road database that every car uses to drive.

All self driving cars that communicate with each other would dramatically increase the throughput of our roads. In a perfect system, the number of cars that can travel the speed limit is limited only by the number of cars that can fit on a roadway. With humans, we got nowhere near full capacity before a ripple effect causes traffic behind to slow to a crawl. Self driving not only helps the traffic problem, drafting at distances which would be unsafe for humans would also increase efficiency.

All of these "self driving cars" come with a big disclaimer: the self-driving functionality is a convenience feature and only intended to be used with close supervision in clear driving conditions.

When they work, they're amazing. When they don't, you still need a licensed driver behind the wheel. There's a big gap between the Tesla Autopilot and taxis without a person in them.

They're still using a whitelist of locations where the radar input is ignored. Not exactly confidence inspiring.

It's like listening to a project manager working on an impossible task telling you "everything is going great" when you ask him about progress but ask for a deadline increase every 6 months. After 3 years it's obvious to everyone that the end goal won't be reached.

Looking back at articles from 2014-15-16 truly is magical:

"General Motors president Dan Ammann in 2014 said he would be surprised if his company wasn’t shipping self-driving cars by 2020.

Nissan CEO Carlos Ghosn doubled down on a 2020 dedline this past fall

Apple, according to the Wall Street Journal, has set a shipping date of 2019 for its own electric vehicle, though the Journal also reported that the first version of the car might not be driverless.

Google’s Sergey Brin gave the most rosy prediction of anyone in 2012, when he suggested: “You can count on one hand the number of years it will take before ordinary people can experience this.”

Oh, and a 26-year-old hacker’s startup, Comma.ai, plans to start selling autonomous conversion kits for Honda and Acura vehicles this year." [0]

Waymo's CEO recently admitted they will "never" be able to have self driving cars [1], apple still hasn't released anything, GM president must indeed be surprised as he recently stated "Our work is far from done" [2] and Ghosn is an international criminal.

[0] https://www.theguardian.com/technology/2016/jun/02/self-driv...

[1] https://www.cnet.com/news/alphabet-google-waymo-ceo-john-kra...

[2] https://www.bbc.com/news/technology-51171398

Reply on news.ycombinator.com