Cruise Robotaxis are failing in groups in SF

Topic: 
Tags: 

Several instances have taken place where multiple Cruise robotaxis have all frozen as a group, sometimes blocking intersections.

I discuss reasons for that, and why it's not that big a deal, in this new article on the Forbes site.

Cruise Robotaxis are failing in groups in SF

Comments

With regularity?

Ouch.

I agree, if they don't fix it, that's an issue. The great thing about robots is, if they have bugs, you fix them, and they are unlikely to repeat the pattern. Humans, not so much.

These were systematic failures and not random failures. These systemic failures need systemic solutions.

Nobody said otherwise.

You are implying Cruise did not test for extended communications outage. Since when does planning for obvious basics like an extended outage morph into only the super competent and super-funded will do it very well.

We don't know what exactly went wrong, but it seems likely they failed to plan or test for this sort of situation.

Or maybe it was even a conscious risk decision. After all, Cruise decided to deploy before implementing pick-up/drop-off, and that was a deliberate "one thing at a time" decision.

General Motors Co has lost nearly $5 billion since 2018 trying to build a robotaxi business in San Francisco.

GM lost $500 million on Cruise during this past second quarter.

GM didn't "lose" $500M on Cruise, and more than a VC fund which makes a $500M investment would be said to have lost the money right after making the investment. They might lose it in time, of course. Cruise is not a business, it's an investment.

The loss phrasing was a Reuters reporter, and content virtually the same from websites such as Autonews, Streetinsider, USNews, Insideevs, Teslarati, etc..

Still doesn't make it valid, though.

John Berry is Aniccia on Twit.
Read the last 3 months and it will suprise you.

hats off to john
well worth the time to review posts

Andrew Ng on The Batch
Letters - Article

The Trouble With Reinforcement Learning

Published Aug 03, 2022

Deeplearning dot ai

www dot autoevolution dot com/news/two-tesla-evs-embarrass-themselves-in-a-self-parking-test-there-s-a-simple-reason-for-it-196451 dot html

Aurora's burn rate is ~$2 million per day.

“We don’t expect that a conventional fundraising opportunity of sufficient scale to add a year of runway to the company will present itself over the next six months”

Cruise burn rate is $5 mil per day vs $2 mil for Aurora

Conti

must read - PhilKoopman calls bullshit on Cruise

Pretty intense comments from PhilKoopman on Linkedin

www dot linkedin dot com/feed/update/urn:li:activity:6971082783884505088/

How is it that they determined the other vehicle would turn right (because that lane is either right turn OR taxi/bus) if it was going 40 mph? Sounds like lack of common sense checks in their system and a system integration issue.

- If it had been a Prius taxi instead of a Prius rideshare in that oncoming lane, it would have been justified in continuing straight as far as I can tell. Does it actually know the difference? That is quite a perception challenge. And a rideshare driver incorrectly thinking they qualified for a taxi lane is a completely foreseeable confusion that you should not be pinning your own passengers lives on not being made.

- This is a behavior defect that should have been stressed in simulation. There is nothing weird about an unprotected left into a car going faster than it should on an empty street. So I'm not buying the implicit argument that this was the one singular exceptional bug. The real question is what they're doing about the other bugs that are so likely to be out there. Now that they have taken out the safety driver, does each bug get fixed only after a crash?

- So very much blaming. Blames the other vehicle for changing lanes to try to avoid hitting the Cruise vehicle that slamming to a stop directly in front of it. Blames the other vehicle for 40 in a 25 zone (not great, but not unexpected either). The writing is obviously blame-casting rather than clinical analysis, significantly detracting from credibility.

I have never seen a recall report this full of spin and blame. I

Add new comment