Even After $100 Billion, Self-Driving Cars Are Going Nowhere

Seems like nothing more than corporate hype.
edit:
Tesla Again Paints A Crash Data Story That Misleads Many Readers
Brad Templeton

I have done analysis of these reports before, generally concluding that what they actually show is that users of Autopilot have a roughly similar number of crashes to those not using it. The Impact report did not cite a number for non-Autopilot use with ADAS active safety, but the quarterly safety report for Q4/2022 reported about 0.71 airbag events per million miles, a bit worse than the annual summary.

I might point out that this degree of crash safety may actually be a psychologically viable level to enable a certain advanced capability of self-driving automotive technology to make it in the marketplace and be adopted by most new car buyers. 
 
It seems to me that the public in general might buy a system that statistically has a similar safety level to the present, as long as it has the key practical user advantage of enabling the user to not have to remain attentively focused on and alert to the immediate traffic situation and the performance of the AI system. That is, such a system would appear to be psychologically and economically viable only if this technology actually allows the car owner to focus his attention elsewhere on his smartphone, music, a movie, conversation, or whatever, along with being statistically no more unsafe. This capability would seem to be very valuable and one of the main attractions of a practical self-driving technology.

I don't know the details about the Tesla Autopilot system, but I'm sure it doesn't have such an advanced capability.

If the technology falls short of this level, and/or the crash safety is significantly inferior to present, it doesn't seem to me to be good enough to adopt. If I can't just relax and leave the driving to it, just it having some labor-saving features aren't enough for me to be willing to put out major more dollars for a car, along with much higher complexity, ways it can fail, and potential for expensive repairs.

Based on the performance of the industry and technology so far, I don't think I will hold my breath. Even this suggested level of performance might only come after another 30-40 years of development, or more likely never.  
 

‘No! You stay!’ Cops, firefighters bewildered as driverless cars behave badly

Joe Eskenazi

...fire officials who’ve worked with the autonomous vehicle companies had nice things to say about their representatives. But not about their vehicles.

“They’ve made every effort to work with us in public safety measures and be a good partner,” said one. “But they do not have a good product.”

Mission Local has obtained some 15 Fire Department incident reports documenting dangerous and/or nuisance situations in which Waymo or Cruise vehicles interfered with fire vehicles or emergency scenes. The vast majority of these reported incidents occurred in recent months, and a majority took place in April (driverless cars were only in December given the green light by the state to traverse San Francisco 24/7).

These incidents are either happening more regularly or being documented more regularly — or both. Within the marginalia of reports written last week, fire department officials complain that driverless car incursions are now a “daily occurrence.” This does not appear to be an overstatement: The notes on an April 26 report state, “This is an increasing problem. I believe there are many more incidents that are not being reported.” A subsequent note states “Number 3 today!”

A firefighter then spoke via a radio in the car to a Cruise employee. “The individual apologized for the ‘inconvenience,’ and said a team was working on moving the car,” reads the report. “Even after talking to him, the car did not immediately get moved from the scene. I informed him that this time it was an ‘inconvenience,’ but if someone needed to be rescued by the Truck it could have been a life and death situation.”

Every firefighter I spoke with wanted them off city streets to work out the kinks, posthaste. But that’s not the call of a local fire department — or, it turns out, any local official. Autonomous vehicles are regulated at a statewide level. And, in addition to tech wizards, all of these companies have government wranglers on the payroll.

Get a Tesla if you want to learn about AI trying to kill you, says Apple cofounder Steve Wozniak

Fred Lambert

When talking about Tesla CEO Elon Musk’s statements promising self-driving capabilities through future software updates during a new CNN interview, Wozniak implied that Tesla vehicles on FSD are trying to “kill” people:

I actually believed those things, and it’s not even close to reality. And boy, if you want a study of AI gone wrong and taking a lot of claims and trying to kill you every chance it can, get a Tesla.

It looks like Woz is not scared of Tesla’s “hardcore litigation” team, which has been a bit trigger-happy when it comes to suing people for defamation lately.

One unapologetic capitalist cheap shotting another unapologetic capitalist.  Doesn't tell me much.

One unapologetic capitalist cheap shotting another unapologetic capitalist.  Doesn't tell me much.

I'd say it's a big sign of dwindling investor confidence in AI as it exists to provide driverless cars.

People are going to pay attention to what Wozniak says/thinks about tech.

Police tried to unstick a cluster of confused Waymos in Arizona, but to no avail

Sam Smith

The jam occurred along Roosevelt Street in Downtown Phoenix last month and saw 12 Waymos get “confused” and ended up clogging the road. According to one TikTok user, police attempted to intervene, but as there were no human drivers behind the wheel, the cops weren’t exactly able to tell the befuddled Waymos where to go.

Waymo blamed the jam on a software glitch, reports the AZ Mirror. A spokesperson said, “As safety is central to our mission, our autonomous driving technology prioritizes the safest driving path with the information it has at any given moment. Sometimes, that means our vehicle will pull over or come to a stop if it’s assessed to be the safest course of action in that instance, as happened here. We identified the software that contributed to this situation and made appropriate updates across our fleet within 24 hours.”

However, the incident is the latest in a long line of run-ins with authorities for driverless cars...

Police tried to unstick a cluster of confused Waymos in Arizona, but to no avail

Sam Smith

Some more info:

Waymo says software glitch caused a driverless traffic jam in Phoenix last month

Jerod MacDonald-Evoy

“My sense is that there are probably ways to disrupt the system if you were creative enough, but certainly not on the level of a conventional DDOS, simply because of the current scale of autonomous vehicle deployment at the moment,” said Andrew Maynard, a professor of advanced technology transitions at Arizona State University’s School for Future of Innovation in Society. 

Maynard said that a DDOS attack, or distributed denial of service attack, would be unlikely given the current way Waymo operates and how the company identifies threats and glitches within its system. 

“I suspect here that autonomous vehicle systems are developing at a measured enough rate that, as they grow, they will be pretty resilient to DDOS-type attacks,” Maynard said. “This doesn’t make them immune, but it does suggest that very disruptive attacks will be challenging (and therefore not worth it to most perpetrators) and quickly addressed.” 

...people could still try to create havoc and game the system. 

In 2020, an artist used 99 phones to create a phantom traffic jam in Google Maps, something that theoretically could create problems for self-driving systems like Waymo, which rely on mapping programs like Google Maps. 

The Self-Driving Vehicle Oversight Committee that Ducey created has only met once, in 2016.

I'd say it's a big sign of dwindling investor confidence in AI as it exists to provide driverless cars.

People are going to pay attention to what Wozniak says/thinks about tech.

I don't think so Sci.

Time will tell but I fully expect capital to keep flowing into all forms of automation, including autonomous driving.  Wozniak doesn't seem plugged into the flow of capital anyway and comes off more like a consumer with an axe to grind.  (A fair perspective to be sure as Tesla has epically overpromised.)

I don't think so Sci.

Time will tell but I fully expect capital to keep flowing into all forms of automation, including autonomous driving.  Wozniak doesn't seem plugged into the flow of capital anyway and comes off more like a consumer with an axe to grind.  (A fair perspective to be sure as Tesla has epically overpromised.)

Automation in general sure, but the current iteration of driverless cars are not looking good with all these problems. I think we're one or two fatalities away from them getting pulled off the streets, though ideally they'd be taken off before then.

That said, I do think driverless cars are possible just you need a different approach that allows one to properly model the problem rather than just accumulate data and hope it works out. Someday I'm sure we'll have them, just when is hard to say.

Tesla’s “Full Self-Driving” sees pedestrian, chooses not to slow down

Jonathan M. Gitlin

The version 11.4 update in April was supposed to improve how the cars behaved, but there's now more evidence that the FSD Beta still leads to Teslas breaking traffic laws. Section 7 of California's Driver's Handbook, which deals with laws and rules of the road, says that pedestrians are considered vulnerable road users and that "pedestrians have the right-of-way in marked or unmarked crosswalks. If there is a limit line before the crosswalk, stop at the limit line and allow pedestrians to cross the street."

This is not the first time Tesla's software has been programmed to break traffic laws, either.

Despite that bold stance in public, Tesla has been far more circumspect when dealing with authorities—in 2020, it told the California Department of Motor Vehicles that it did not expect FSD to become significantly more capable and that it would never pass beyond so-called SAE level 2, which requires an alert human in the driver's seat who remains liable for the car's actions.

Or, as author Ed Niedermeyer more concisely put it, "Full Self-Driving" is not, and never will be, actually self-driving."