Hacker News Reader

Comment by hannasanarion | Hacker News Reader

2014 called, they want their unbridled optimism about the inevitability of self-driving back

> But once autonomous vehicles come online, as they already are starting to

The current status of autonomous vehicles is pretty miserable on city streets. Tesla's new "full self driving" has a reported error rate of one incident requiring manual intervention per 2 miles. Nobody is going to get into a taxi that tries to drive into a wall or onto train tracks a couple of times every hour.

Uber and Lyft have already given up on them, selling off their AI divisions. They have no near-term automation plans.

View on HN

You're looking in the wrong places for the cutting edge of self-driving. While I greatly respect Tesla for energy storage/generation, you should really be looking at companies like Waymo or Cruise.

If you live in San Francisco or the Bay Area, I'm sure you've seen many of their vehicles being driven around.

While OK, cutting edge might be better but big questions still are open.

It is not about making road safer...

Question is, if it will be more cost effective to have self driving car or not. Cutting edge is currently not cost effective as developing and operating such car is costing loads of money.

There are still operation costs that are not obvious like insurance. I don't think there is sane company anywhere that wants to put 100's of cars on the same roads as insurance scammers.

And even so cost effectiveness isn't obvious. The hardware results in quite expensive vehicles that additionally won't be owned and maintained by gig workers for free.

to me, a self flying plane is a lot easier problem to solve but i don't hear about anyone getting into a plane without a pilot (and co-pilot) at the controls. I know we have auto-pilots but there's always a human and backup in case something goes wrong. Flying a plane is a lot easier on a computer than navigating the ground and ground traffic (IMO). I would think we'd see self-flying planes before self-driving cars.

I think the main issue is that in case of a problem you can't just gradually slow down and get out of a plane. A lot of things can go wrong and a broken autopilot == crash.

If the self-driving car breaks down completely it could still slow down until stopped with some emergency lights on. This is the worst case scenario and in many cases wouldn't end up with anyone being hurt.

That is not the worst case scenario.

Dangerous acceleration, braking, steering, or some combination of the three are far, far worse.

I've repeatedly had lane-keeping veer off-road while adapative cruise control accelerated at the same time. That's incredibly dangerous and if I hadn't had my hands on the wheel and my foot resting on the brake pedal I may have driven into a tree/sign/parked car. I don't use lane-keeping anymore.

Slowing down until stopped with some emergency lights on is also plenty dangerous if it happens on a high-speed road.

Are you confident that self driving cars will not work even in 2034 or 2044 ?

I'm not the OP, but personally I just can't see it, not on all roads and at a level that most people would consider "self driving". And as for fully self driving such that there are no manual controls, absolutely no chance.

I think self-driving cars are realistically only going to work in conjunction with other technologies embedded in roads, signs and other vehicles - and even then, it's still a massively challenging problem, with potentially dire consequences when things go l wrong.

Reply on news.ycombinator.com