Autonomous Car Progress

[Supcom]

Waymo and Cruise also have firewalls to protect from hacking.

Funniest thing I've read today.

 

[diplomat33]

Funniest thing I've read today.

Why is it funny? AV companies like Waymo take cybersecurity very seriously. They use multiple layers of security.

Here is what Waymo says about cybersecurity in their safety report:

We complete a comprehensive review of all potential security access points to our autonomous driving system from both the interior and exterior of the physical vehicle, and take steps to limit the number and function of those access points.

This begins by collaborating with our OEM partners at the onset to identify and mitigate vulnerabilities of the base vehicle. In addition, our software and vehicle design processes incorporate cybersecurity risk assessments, allowing us to implement defenses and protections according to the risk posed by each known vulnerability. New software releases go through an extensive peer review and verification process. Our hazard analysis, threat modeling and risk assessment processes have been designed to identify and mitigate risks that might affect safety, including those related to cybersecurity.

We use layers of security to protect our autonomous driving system, especially safety-critical functions like steering and braking, against unauthorized communications, including vehicle control commands. We also consider the security of our wireless communications. Our vehicles do not rely on a constant connection to operate safely. While on the road, all communications (e.g. redundant cellular connections) between the vehicles and Waymo are encrypted, including those between Waymo’s operations support staff and our riders. Our vehicles can communicate with our operations team to gather more information about road conditions, while our vehicles maintain responsibility for the driving task at all times.

These protections help prevent anyone with limited physical access to our autonomously driven vehicles, whether passengers or malicious actors nearby, from impairing or altering their security. We have diverse mechanisms for noticing anomalous behavior and internal processes for analyzing those occurrences. Should we become aware of an indication that someone has attempted to impair our vehicle’s security, Waymo will trigger its company-wide incident response procedure, which involves impact assessment, containment, recovery, and remediation.

@Dewg you might find this section from Waymo on cybersecurity also interesting since it pertains directly to your question.

 

[Dewg]

Thanks Diplo! I really appreciate your insight on this thread - I learn much reading your comments.

 

[eli_]

If Waymo was really serious about security they would open source the thing and let the world see how it works and validate it and have bug bounties and so forth. As it stands it's just a totally opaque black box and we only have "trust us we're Waymo" as proof that it's secure.

 

[Doggydogworld]

Waymo and Cruise can do full autonomous driving with no cell signal or internet connection because everything needed for autonomous driving is on board the car. So they would not be affected by signal jammers.

The self-proclaimed Cruise employee who posted on Reddit the other day said Cruise vehicles do need a cell connection. If they lose connect they go into some kind of safe mode. I recall when Kyle Vogt explained about the remote monitor program he gave an example of a 20 second "teleguidance event" every five minutes. If they really get involved that often it's probably a good idea to require a cell connection.

 

[diplomat33]

If Waymo was really serious about security they would open source the thing and let the world see how it works and validate it and have bug bounties and so forth. As it stands it's just a totally opaque black box and we only have "trust us we're Waymo" as proof that it's secure.

There is no way any company would just open source their entire autonomous driving software. That would be equivalent to just giving all their autonomous driving to the competition. And why would the competition help Waymo validate their software? LOL.

 

[diplomat33]

The self-proclaimed Cruise employee who posted on Reddit the other day said Cruise vehicles do need a cell connection. If they lose connect they go into some kind of safe mode. I recall when Kyle Vogt explained about the remote monitor program he gave an example of a 20 second "teleguidance event" every five minutes. If they really get involved that often it's probably a good idea to require a cell connection.

I stand corrected about Cruise. But Waymo says that their vehicles do not need a constant cell connection:

Our vehicles do not rely on a constant connection to operate safely.

 

[Supcom]

Why is it funny? AV companies like Waymo take cybersecurity very seriously. They use multiple layers of security.

I'm sure that they all do. But saying, "they have a firewall', is a joke. Every hacked organization had a firewall.

 

[stopcrazypp]

Waymo and Cruise can do full autonomous driving with no cell signal or internet connection because everything needed for autonomous driving is on board the car. So they would not be affected by signal jammers. Waymo and Cruise also have firewalls to protect from hacking.

Also, Waymo and Cruise don't have any remote control functions. So there is no remote control functions to hack.

Besides from the point about the cell connection, I'm not sure you are correct on the remote control function either for Cruise. I think you are making a general assumption just because that is how Waymo operates, that it applies to all the autonomous vehicle companies, which is not necessarily true.

When Cruise launched the operation of their vehicles without a safety driver (but with safety operator in passenger seat who can press a stop button), they declined to answer when asked about remote control:
"Asked whether remote operators are able to take control of the vehicle when needed, Ammann declined to answer."
Cruise is now testing fully driverless cars in San Francisco

Unless you have something more new where Cruise says conclusively that the cars have no ability to be remote controlled (as opposed to Cruise choosing not to use at the moment even thought they have the ability), I don't think you can say Cruise does not have remote control functions. Waymo instead has said clearly there is no way to remote control the car for their system.

 

[Daniel in SD]

If Waymo was really serious about security they would open source the thing and let the world see how it works and validate it and have bug bounties and so forth. As it stands it's just a totally opaque black box and we only have "trust us we're Waymo" as proof that it's secure.

We're also trusting Tesla that their security measures prevent someone from installing software that causes millions of cars to wake up and aim for the nearest pedestrian. It's scary stuff...

 

[Supcom]

We're also trusting Tesla that their security measures prevent someone from installing software that causes millions of cars to wake up and aim for the nearest pedestrian. It's scary stuff...

The same goes for anyone building cars that have ADAS features and any type wireless data interface (cellular, wifi, bluetooth).

In fact, I would be more concerned about those that have no over the air update capability, since those are the ones that never get updated for security flaws.

 

[Doggydogworld]

We're also trusting Tesla that their security measures prevent someone from installing software that causes millions of cars to wake up and aim for the nearest pedestrian.

You don't need 3rd party software for that!

The whole video is worth watching. 16:15 is terrifying.

 

[diplomat33]

Aurora is planning to launch driverless trucking service with about 20 autonomous trucks by the end of next year in Texas:

Aurora has designed and configured the hardware and software it will use to launch a service toward the end of next year in which roughly 20 trucks will ply highways without a human on board. “We’re now in the phase where we are doing the final refinements and the validation system-wide,” Sterling Anderson, the company’s co-founder and chief product officer, said in an interview just south of Dallas.

Why Texas:

Aurora is starting in the Lone Star state for a few reasons. Texas is the US’s largest truck freight market and has long, sometimes very boring, stretches of freeway. Its interstate highway network boasts almost a third more miles than second-ranked California. Texas also has some quirks that are helping teach Aurora’s system how to deal with unexpected scenarios. One is the incessant building and repairing of roads, resulting in 3,100 construction sites statewide, including 40 or so on Aurora’s route between Fort Worth and El Paso, Anderson said. There’s also the Texas U-turn, the horseshoe-shaped turnabouts at underpasses below major highways in cities and rural areas alike.

Autonomous driving progress:

It took about a month for Aurora’s sensors and software to master the Texas U-turn, which allows vehicles to reverse course on a highway without hitting a stoplight. The maneuver requires the autonomous truck to yield to traffic coming at it from multiple sides and part of the methodical learning Aurora’s computer does with each test run.

So far, it’s working. Human operators who sit with hands poised to grab the wheel aren’t having to preemptively disengage the self-driving system as often for situations it’s not yet been trained to handle. The ability to navigate through constructions sites has improved dramatically, Anderson said. Aurora, whose other co-founder Chris Urmson used to lead Google’s self-driving program, declined to offer detailed metrics. Unlike California, Texas doesn’t require companies to publicly report the number of times their human test drivers disengage the autonomous-driving systems they’re testing on roadways.

Robot Trucks on Texas Highways Herald Era of Driverless Big Rigs

The startup founded by the former heads of Google’s Waymo and Tesla’s Autopilot is refining its driverless system ahead of a launch next year.

www.bloomberg.com

 

[Bitdepth]

You don't need 3rd party software for that!

The whole video is worth watching. 16:15 is terrifying.

I watched the video, and it shows just how bad the system is at understanding road rules, bad prediction for agents on the road, planning and bad driving policy. Lots of room for improvement and nowhere close to a safe L4 ADS.

 

[AlanSubie4Life]

16:15 is terrifying.

What surprises me is that this surprises anyone who has used FSDb (b) for more than about two minutes.

It’s literally exactly what I would expect based on my experience. Probably would have expected it after watching a couple videos, too, hard to say, but any personal experience with the system is very clarifying.

It just mystifies me that he was disappointed with Tesla at the end of the video. How could he possibly be disappointed? Hadn’t he used FSD previously?

 

[diplomat33]

It just mystifies me that he was disappointed with Tesla at the end of the video. How could he possibly be disappointed? Hadn’t he used FSD previously?

Well, it might depend on what narrative he is listening to. According to the TSLA bulls on Twitter, FSD beta is ready to be a robotaxi.

https://twitter.com/x/status/1544717953597308928

 

[Dewg]

The truth always lies somewhere in the middle. There will always be people who are polar opposites on anything - It's the best thing ever and 100% flawless, and OMG it sucks and can't do anything right. The truth is that it's somewhere in between - it's probably doing somethings right and somethings wrong. It will never be 100% flawless, but it we watch the bell curve of complaints, we'll see it shift over time towards the flawless side.

 

[Doggydogworld]

It just mystifies me that he was disappointed with Tesla at the end of the video. How could he possibly be disappointed? Hadn’t he used FSD previously?

It does great on his normal commute. Usually zero or one disengagements, very few murder attempts.

Or maybe he just watched too many Whole Mars videos....

 

[Daniel in SD]

The truth always lies somewhere in the middle. There will always be people who are polar opposites on anything - It's the best thing ever and 100% flawless, and OMG it sucks and can't do anything right. The truth is that it's somewhere in between - it's probably doing somethings right and somethings wrong. It will never be 100% flawless, but it we watch the bell curve of complaints, we'll see it shift over time towards the flawless side.

The truth is that driverless vehicles need to be near 100% flawless. Human drivers achieve 1 death per 100 million miles of driving in the US.

 

[AlanSubie4Life]

It does great on his normal commute. Usually zero or one disengagements, very few murder attempts.

Hmm. Guess it comes down to perception and how a person extrapolates expectations. Does acceptably (if I am the sole occupant) on my commute as well, except on residential streets. But it is also pretty clear to me that it will murder me if I don’t pay attention. It only took a single unexpected jerk or turn of the wheel to make that clear.

Does not seem ready for autonomous driving. Progress is slow.