Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

S.F. says incidents by Cruise, Waymo driverless taxis are ‘skyrocketing.’

sfchronicle.com

Post reply on HN
Print view View original

71–80 of 1001 posts

Re: S.F. says incidents by Cruise, Waymo driverless taxis are ‘skyrocketing.’

#71

Post by tiltowait »

cozzyd wrote: post #70
scarmig wrote: post #58

Earlier quoted context omitted.

About a month ago a human driver ran into me while I was riding a bike and sped away (luckily I was unharmed, aside from a scrape or two). How exactly would stricter driving tests or revocation of licenses have prevented that? If a Waymo had done that, on the other hand, it'd have stopped and I'd be getting a nice big check from Google.

traffic enforcement cameras (e.g. automatic tickets for speeding, running red lights, hitting cyclists etc.) would probably help, but for some reason motorists are against that. I guess the nice thing about robocars is that they can be effectively regulated not to drive at unsafe speeds (though probably people would be mad about that too...).

Red light cameras don't seem to increase safety: https://www.scientificamerican.com/article/red-light-cameras...

Motorists hate them because they're a money grab from local municipalities.

Re: S.F. says incidents by Cruise, Waymo driverless taxis are ‘skyrocketing.’

#72

Post by reso »

yazaddaruvala wrote: post #67
reso wrote: post #63

Earlier quoted context omitted.

You need AGI for good self-driving because driving requires predicting the actions of other human drivers at an extremely high level. This is second-nature for humans so we barely notice that we are doing it, but it is extraordinarily difficult for non-humans.

This is overcome with super human sensing and reaction time and better visual angles. i.e. radar based deceleration to avoid accidents already helps many humans avoid collision. It'll help the robot too. The rest of driving is relatively simple, methodological, and slow.

That is a conjecture which does not represent the current state of the technology.

Re: S.F. says incidents by Cruise, Waymo driverless taxis are ‘skyrocketing.’

#73

Post by reso »

valine wrote: post #68
reso wrote: post #63

Earlier quoted context omitted.

You need AGI for good self-driving because driving requires predicting the actions of other human drivers at an extremely high level. This is second-nature for humans so we barely notice that we are doing it, but it is extraordinarily difficult for non-humans.

No you don’t. I can feed images of crazy driving scenarios into LLaVA and get reasonable responses. That’s a general purpose LLM with $500 worth of fine tuning running locally on my PC. You should look into what can be done with the current state of the art LLMs. Your intuition for what’s possible is out of date. If I can do that with open source LLaMA variants, I can only imagine what’s possible if you have an actua…

That's a nice conjecture, we will see in the coming years if it plays out.

Re: S.F. says incidents by Cruise, Waymo driverless taxis are ‘skyrocketing.’

#74

Post by valine »

yazaddaruvala wrote: post #65
valine wrote: post #47

Earlier quoted context omitted.

Why does it require AGI? That's a bold, unsubstantiated claim. Let me make my own bold claim: Simple common sense reasoning is very close to being solved with our latest LLMs. You would need AGI if you wanted to replace the team developing your self-driving car. A decent LLM, with vision input that's been fine-tuned on driving scenarios, will be more than sufficient for Level 5 driving.

I roughly agree. I think the only other piece, that is critical to mention, would be remote humans for support in extremely awkward situations to help get the car get back on track as RLHF. This requires the car can always come to a safe stop, which I think the LLM-based driver should be very capable of doing.

Automatic emergency braking would be a good first step, it would certainly solve the case in article where the car drives though downed power lines.

I think the logical next step is to have the LLM output the driving path similar to how GPT4 outputs SVGs. Feed in everything you have, raw images, depth maps, VRU positions, nav cues, and ask the LLM output a path.

Re: S.F. says incidents by Cruise, Waymo driverless taxis are ‘skyrocketing.’

#75

Post by jarjoura »

Laws are written and designed for humans who constantly try to push it to the limit and get away with things. These cars are all over the place and I've been caught stuck on a street in Bernal Heights because it couldn't quite fit and just froze in place. I had to back out of the street because there wasn't any room for me to go around it.

Yes it may be following laws to the letter, but not only was I grumpy about having to carefully back out of a street, there wasn't any way to communicate with the car. It doesn't have any feedback mechanism for me, so it just looks like a confused dead robot in the middle of the street.

I think Waymo and Cruise are both happily working on hard problems in this space with limitless VC funding (or daddy GM money) thrown their way. So I support their efforts, because at the very least, they'll enrich our understanding of self-driving cars even if it were to fail in the marketplace long term.

I'm curious how profitable it will be in practice once the VC money dries up. Paying half a million dollar salaries to employees who need to keep these things up to date adds up. That, on top of hardware costs for precise sensors that will fail fast in harsh weather conditions. On top of normal wear and tear of people in and out of the cars all day for cars that won't need down time.

Lastly, I'll end my thoughts with this... why are we obsessed with the idea of self driving cars anyway?! What human problem does this solve? I am 100% on board with building autonomous cruise-control for my own car, so I'm not against the tech. I just find the taxi angle to be weird. All I see so far in my mind is a product that just takes away an entire class of job opportunities with no actual gain.

Re: S.F. says incidents by Cruise, Waymo driverless taxis are ‘skyrocketing.’

#76

Post by bluGill »

MontyCarloHall wrote: post #30

>Cruise and Waymo say city officials have mischaracterized their safety track records. Their driverless taxis, the companies say, have lower collision rates than human drivers and public transit. This is comparing the mean driverless incident rate to the mean human driver incident rate. This is disingenuous, since a small minority of human drivers cause the vast majority of incidents, thereby severely inflating the m…

I care about the overall mean though. Do you know what we call the bottom 10% drivers? Drivers! So long as overall they are better, then get humans off the road, sure the top 10% are worse off, but overall I'm better off as I will never have to deal with that bottom 10%.

Re: S.F. says incidents by Cruise, Waymo driverless taxis are ‘skyrocketing.’

#77

Post by AlotOfReading »

carapace wrote: post #59
peterisza wrote: post #5

Earlier quoted context omitted.

I think they are training and not testing. I might be wrong, but if I am right, the use of public roads is a necessity. This is how AI works, I guess.

What I think they should do (what I would do if I were in charge at a AI-car co.): There's a whole ersatz city out in the desert somewhere the whole purpose of which is to allow in situ modelling of new "smart" hardware. (This is a real thing, but I forgot what it's called and I'm too lazy to look it up, I apologize for my barbarism.) The people that "live" there are actually paid employees. That's where you test you…

As far as I'm aware, all L4 AV companies have done years of closed course testing. You can (not) see these vehicles being tested in places like GoMentum station in Concord, Altamont Raceway in Livermore, and TRC in Merced, not to mention other courses in Vegas, Seattle, Tahoe, AZ, Florida, Michigan, and China.

Of course, companies also started with small, slow vehicles like the Waymo firefly and the Nuro R vehicles. Voyage (acquired by Cruise) was doing their testing in a low speed access controlled retirement community, with essentially golf carts.

Re: S.F. says incidents by Cruise, Waymo driverless taxis are ‘skyrocketing.’

#78

Post by gonehome »

reso wrote: post #63
valine wrote: post #47

Earlier quoted context omitted.

Why does it require AGI? That's a bold, unsubstantiated claim. Let me make my own bold claim: Simple common sense reasoning is very close to being solved with our latest LLMs. You would need AGI if you wanted to replace the team developing your self-driving car. A decent LLM, with vision input that's been fine-tuned on driving scenarios, will be more than sufficient for Level 5 driving.

You need AGI for good self-driving because driving requires predicting the actions of other human drivers at an extremely high level. This is second-nature for humans so we barely notice that we are doing it, but it is extraordinarily difficult for non-humans.

While this might be true for the truly general case (though I’d bet it’s not), when you have a very constrained operating area it’s a lot less true.

Waymo in Phoenix and current cruise cars in SF seem like good counter examples.

The bar is also a lot lower - human drivers are pretty bad.

Re: S.F. says incidents by Cruise, Waymo driverless taxis are ‘skyrocketing.’

#79

Post by xboxnolifes »

cozzyd wrote: post #69
yazaddaruvala wrote: post #62

Earlier quoted context omitted.

> According to the NHTSA, 19% of motor vehicle fatalities involved drivers with invalid licenses. Furthermore, drivers with invalid licenses comprise 13% of all drivers in fatal crashes. https://www.carinsurance.com/Articles/driving-without-licens... I'm all for more rigorous driving standards. However, the USA is car centric and people need jobs. To really do what you're suggesting, the USA would need to regulate ca…

Yes, car-centricity is a big problem, I agree (as someone who doesn't own a car or regularly drive...). Are robo-cars going to prevent people from driving with invalid licenses?

> Are robo-cars going to prevent people from driving with invalid licenses

Yes. Because they wouldn't be driving.

Re: S.F. says incidents by Cruise, Waymo driverless taxis are ‘skyrocketing.’

#80

Post by slashdev »

bshipp wrote: post #34
slashdev wrote: post #18

Earlier quoted context omitted.

Uber’s driverless vehicle killed someone jaywalking at night in another state. So you can’t say they’ve never killed anyone. Plus Teslas have killed their own occupants plenty of times.

I'm debating about renewing my truck license for this reason. If it's taking this long for automobiles to be approved and accepted, it'll be 30 years before trucking is automated. The two decades during which human oversight of automated systems will be mandatory would be long enough for me to finish off my career getting paid to drive while I sit in a cab writing code, periodically checking over the status of my lea…

I think there will be a strong demand for truck drivers long after most people would expect.

We’re far from self driving vehicles still, and trucks will be the last to be automated, and once they are, I expect they’ll require human supervision just like you predict.

Post reply on HN