4 ms·
Between the noise of self-driving singularlists, and that of self-driving impossibilists, it's great to see plodding, sequential progress like this.
by unchocked 5y ago
Between the noise of self-driving singularlists, and that of self-driving impossibilists, it's great to see plodding, sequential progress like this.
- oblio 5y agoI think most reasonable people aren't against self-driving, they don't consider it impossible. They're generally against "create a self-driving startup in 2021, IPO by 2024 at the latest". That's just snake oil salesmanship.
- smnrchrds 5y agoAlso against the "move fast and crash into things, and people" attitude some of them have.
- marmshallow 5y ago(Tesla)
- Animats 5y agoYes. I've been saying that for a few years, as Waymo's disconnect rate improved by a factor of 2 every 18 months or so. This is a hard engineering problem. Now that the "fake it til you make it" clowns have dropped out, there's real progress.
- kstrauser 5y agoWhat's the "disconnect rate"?
- klodolph 5y agoI believe it's how often the system requires human intervention... e.g., once per 1,000 miles, or once per 250 hours, or something like that.
- kstrauser 5y agoOh, that would make sense. Thanks!
- exporectomy 5y agoDisengagement rate.
- dylan604 5y agoDid you honestly believe you were clarifying in a less confusing manner than the original question with this response?
- tizzdogg 5y agoThe parent poster is correct though. "Disengagement rate" is the term used in the industry, and what would get you results in google searches. "Disconnect rate" is not what it's called, so it's helpful to correct that.
- ddp26 5y agoThe reported disengage rate has improved a lot, but Waymo only reports disengagements it classifies as "related to safety", based on proprietary analysis and counterfactual simulations it does not share.
- jacquesm 5y agoWhat about the accident rate per miles driven? If that is dropping too I'd see that as confirmation that it really is improving. If that isn't then it's just bookkeeping.
- Animats 5y agoYou can read all California autonomous vehicle accident reports here.[1] Most Waymo reports are "stopped in traffic, was rear-ended". There are two reports this year where the Waymo vehicle was responsible. In both cases, it was backing up while doing a three-point turn in a narrow street and bumped into something at 2MPH. So they need better sensor coverage in back. [1] https://www.dmv.ca.gov/portal/vehicle-industry-services/autonomous-vehicles/autonomous-vehicle-collision-reports/ https://www.dmv.ca.gov/portal/vehicle-industry-services/auto...
- anchpop 5y agonot necessarily, they could be putting the car in more difficult situations as it becomes safer
- ahahahahah 5y agoThis isn't really true. You may be thinking about the CA required disengagement report rules. From a recent waymo safety report: "In this paper, “disengagement” refers to any event in which the AV’s automated driving mode is disengaged. This is broader than the definition used in certain California state regulations, where the term more narrowly refers to certain safety-relevant disengagements."
- dmitrygr 5y ago> Now that the "fake it til you make it" clowns have dropped out I am pretty sure they recently announced "Full self driving v 2.0 for real this time, seriously, guys" and did not drop out
- opinion-is-bad 5y agoFrom personal experience seeing how their data is generated, I don’t trust Waymo reports at all. They straight up lie about the capability and disengage rates to an astonishing degree. I left several years ago, but I see no reason they would have developed honest practices on the meantime.
- sgtnoodle 5y agoI have some personal experience as well, and I didn't pick up on any dishonesty like you describe. There was a consistent emphasis on safety, and multiple levels of thorough testing. When I left, I had generally positive opinions about the project.
- dylan604 5y agoI don't know about the disengage rates, but just the training model of data scares the bejeebus out of me. There was training data released and posted here on HN some time ago. I'm pretty sure the post was talking about how bad the training data was. There were all sorts of crazy things like signage being indicated in the trees or other "to a human" obviously wrong data. However, this was the supposed training data for some ML system. It was at this point that I just shook my head and walked away from interest in the subject.
- opinion-is-bad 5y agoThey definitely care about safety. I don’t doubt their dedication to preventing accidents. My objections come from their hand waving away disengagements from the official results, shared with both regulators and Alphabet leadership. This was done to make their lead look greater than it was and to hide the slow progress with simple, required driving maneuvers like yield signs and unprotected left turns. My position had me handling driver issues directly and what the reports said did not line up with their, or my own, extensive time in the actual car.
- raldi 5y agoWhere can I follow along with the latest data on disconnect rates?
- spectramax 5y agoI should start paying attention to these pesky plots then ;-) https://en.wikipedia.org/wiki/Gartner_hype_cycle#/media/File:Hype-Cycle-General.png https://en.wikipedia.org/wiki/Gartner_hype_cycle#/media/File...