Submitted 1 year ago by ooli@lemmy.world to technology@lemmy.world
https://garymarcus.substack.com/p/could-cruise-be-the-theranos-of-ai
At first I was thinking, a bit of human supervision could not be too bad. And then I got to the part where they said 1.5 workers per vehicle. My maths may be off, but to me that sounds like 0.5 more than is necessary to drive a normal vehicle.
Theranos? Maybe, but at that point, I’d compare it to en.wikipedia.org/wiki/Mechanical_Turk
When I worked at Waymo, we had a ratio of about 10 cars to 1 remote human. I dunno if Cruise is being over-protective, if their tech is just that bad that they need more people than cars, or if the number is just incorrect.
Either way, it hardly matters. It's not like these things are commercially available for a long time yet, anyway. In the testing stages - which Cruise 100% is still in - you definitely want a study team of humans capable of intervening for safety reasons.
If the cars are running all day long it might make sense to need another human to pick up later shifts. Still though, that ratio is way too high to be economical.
Chonky TL;DR because I was a little annoyed that there wasn’t one here -
Certainly no commercial product could ever work at a profit if you needed remote operators anything like that often. As Brooks points out, the term “autonomous” barely applies.
Beyond what Brooks pointed out, the story also notes “Those vehicles were supported by a vast operations staff, with 1.5 workers per vehicle”.
Fitting with this general vibe, a source (that in fairness, I don’t know well) just told me that his impression having visited with them not so long ago was that “they’re definitely relying on remote interventions to create an illusion of stronger AI than they really have”.
if Cruise’s vehicles really need an intervention every few miles, and 1.5 external operators for every vehicle, they don’t seem to even be remotely close to what they have been alleging to the public. Shareholders will certainly sue, and if it’s bad as it looks, I doubt that GM will continue the project, which was recently suspended.
As safety expert Missy Cummings said to me this morning, remote operators could well be “the dark secret of ALL self-driving.”
Human lives at are stake.
Cruise CEO Kyle Vogt essentially confirmed that their “driverless” cars need very regular human intervention:
🤣😂
But the average Whitedude Techbro AI company CEO is looking very seriously into the camera and telling me AI is coming for my job next year…
1.5 operators per vehicle!?
Consider that"dumb" cars are only 1 operator per vehicle. This is somehow reverse-AI
You can just watch the different food chains interacting here from legit media to independent authority to bottom feeding headline-shagger.
Unfortunately, the substack article seems to be freely accessible, while the NYT isn’t. I understand the whole “support journalists” angle, but having to sign up to read stuff so they can more easily correlate what I click on and sell usage pattern data rubs me the wrong way.
I’m a paying NYT subscriber so I guess “supporting journalists” (why did you even put this in quotes?) is more important to me than “vague fears” of “personalized advertising” which are probably much the same on every “website” in “the world.”
Headline shagger has me in stitches
Autonomous cars would complete the hellish dependency on cars in many cities.
As a partially sighted person that’s unable to legally drive, an autonomous car is an absolute dream to me and would give me a personal freedom many currently take for granted.
In our car dependent society, I understand that. But a lot of us would rather have better public transit so you wouldn’t have to have a car to have your freedom.
I'm hoping they are sorted by the time I'm 80 and can no longer safely drive myself.
They offer the chance to push the average number of occupants per vehicle below one.
They also offer the chance to push it above one. Ride-sharing will be a lot more attractive with autonomous cars.
It could reduce the need for individual cars by increasing car sharing.
That’s Car Sharing, not autonomous vehicles, no? Car Sharing is a good thing, definitely, but we really need to get rid of cars. Not completely, but to a point where it’s not the default.
But you can do car sharing with any kind of car. In Germany there are cities that run a rent service for their citizens who only need a car occasionally.
Obviously this only works in the context of a robust public transport infrastructure and in cities built for humans rather than cars, so that the need for a car becomes a rare occurrence.
American cities don’t generally fit that description and until they do the type of car they use won’t change a thing, because it’s not addressing the core problem.
No they wouldn’t. Once most cats are robotaxis, there will be drastically less space needed for car parks which will free up huge amounts of space. That can be used for bike lanes, so cycling becomes safer and more convenient. And I don’t expect most rides to be single occupancy. People will opt for shared rides if they are substantially cheaper, which would cut the number of vehicles on the road. Autonomous cars are actually the best chance we have right now to escape the car centric hellscapes of our current cities.
And I don’t expect most rides to be single occupancy. People will opt for shared rides if they are substantially cheaper,
Bus. That’s called a bus. It can also fit more than five people and doesn’t use as much energy to transport each person. You just reinvented a shittier bus
I firmly believe the solution is autonomous shuttles, not cars. Imagine having bus routes that can dynamically change and adapt to demand. Say we replace every bus with 2 smaller shuttles: during normal service the route could have the same capacity, but if there is an extraordinary event (sports event for example) you could divert them from the low-demand areas to the extraordinary-demand zone.
During lower demand times, you can also have more routes at no extra cost. If you’re clever and make an app to call the shuttle (think Uber but through pre-established routes) the demand can be determined in real time to ensure you don’t have empty shuttles.
And because they’re bigger than passenger cars you’re still increasing the ratio of passengers per vehicle, unlike robotaxis which merely replace private cars, with mostly 1- or 2-passenger trips.
Cities that have studied it believe on-demand car service is necessary (but often much more expensive) to reaching 100% transit coverage. But they also said you could reach like 95% with just busses.
Just asking questions title
Yeah the dark “secret” is they have spent $100 billion dollars and these cars still can’t do anything useful and relatively safe.
Which is why it's a lot like Theranos; they raised (and burned through) a ton of money trying to build something that would be really useful but was still decades from technological feasibility.
not decades from feasibility. But a physical impossibility. Some of the stuff they were supposed to detect was literally not present in a detectable quantity in the single drop of blood they scanned.
Well I should hope autonomous driving tech people believe they can make it work, despite the incredible expense and waste.
Pleeeeeeeaaaaase be true please please please be true .
What bullshit clickbait title… being incompetent is not the same as fraud
Killing someone on accident isn’t the same as murder, it’s called manslaughter and it’s just as bad.
Considering a person has been hurt in an accident and allegedly dragged several feet by the AI driver, I’d say this is far from a “bullshit, clickbait title”.
Murder is considered a more serious crime and carries a heavier penalty.
But they explained the similarity and why they think it's fraud
AI needs training data and level 4 autonomous driving is possible. Mercedes and Waymo are doing it. It’s not fraud to try faithfully. Also, the author’s premise itself is wrong, what Theranos promised was physically impossible.
Faking competence to raise money, though, is fraud
Sar please stay on the road
RobotToaster@mander.xyz 1 year ago
Relevant XKCD:
Image
jflorez@sh.itjust.works 1 year ago
Why not post a link to the actual XKCD comic and give the author the views instead of a random site?
ReveredOxygen@sh.itjust.works 1 year ago
The random site is their lemmy instance’s pictrs. Randall doesn’t care about reposting, and this is nicer since you don’t have to leave lemmy
andrew@lemmy.stuart.fun 1 year ago
Aw man I’ve been making this joke thinking I’m clever for years but I read xkcd pretty frequently. I must have inadvertently stolen the joke from Randall.
wombatula@lemm.ee 1 year ago
There’s probably an XKCD about that.