Live data from Hacker News

Comma openpilot – Open source driver-assistance

comma.ai

201–210 of 220 posts

Re: Comma openpilot – Open source driver-assistance

#201
post #196

Earlier quoted context omitted.

What you call "platforming" I often call "listening to what someone says/thinks". Not every interview needs challenging questions, or to be a battle/debate, and sometimes it's not appropriate (above George Hotz being an example, difference in qualifications being another). But, I enjoy trying to understand someone, quirks and all, especially the human aspect , flaws and all. It's interesting seeing the differences in…

> Not every interview No, not every interview. But if an interviewee presents fiction/hatred as fact the interviewer should have the ability to call that out or at least caution the reader with a "I don't know about that". A specific example that comes to mind is Eric Weinstein's appearance on the podcast and letting him talk about his "long mouse telomere experiment flaws" without questions which at that point had b…

> letting him talk about his "long mouse telomere experiment flaws" without questions

This requires that the interviewer is as knowledgable as the interviewee (the qualification problem I mentioned). Unless the questions and answers are known ahead of time, it won't be possible to know everything an interviewee will say. Assuming this is the case, how should he have handled that response? Should he not interview people outside of his own expertise? I think one way would be "is there any disagreement?" but then you're left with the same problem.

I think Lex Fridman not knowing much about the history/current state of rat telomere research is entirely reasonable. I think a requirement of knowing the entire context of a person is not reasonable. I also don't think it's reasonable to believe everything you hear in an interview, from either human. "Charitable interoperation, but verify" is a good way to take in information.

Re: Comma openpilot – Open source driver-assistance

#202
post #38

Earlier quoted context omitted.

What a coincidence, LTT tested it a few months ago in a Corolla. Here’s the video: https://www.youtube.com/watch?v=xdmxM-v4KQg

15 mins in. It's terrible. How is this allowed on the roads?

FWIW my experience has been very good, and LTT said recently that they are working on a long term review, so I expect that to be a more accurate review than first impressions when you aren't familiar with how it works.

Re: Comma openpilot – Open source driver-assistance

#203
post #132

Earlier quoted context omitted.

Speaking of hacker spirit... Comma has a lot of restrictions on what software you can load / what changes you can make to the software.

No it doesn't? You can run stock, or any fork simply by providing the URL of the version you want to run. Where exactly is the restriction?

https://github.com/commaai/openpilot/blob/master/docs/SAFETY... - they will ban your device from the training set if you run an unsafe fork.

Re: Comma openpilot – Open source driver-assistance

#204
post #33

There's a few Lex Fridman podcasts with George Hotz, the founder. Highly recommend them: #31 https://www.youtube.com/watch?v=iwcYp-XT7UI #132 https://www.youtube.com/watch?v=_L3gNaAVjQ4 #387 https://www.youtube.com/watch?v=dNrTrx42DGQ

These are really just ramblings, nothing of value is being said.

Re: Comma openpilot – Open source driver-assistance

#206

Can someone with technical knowledge explain the key differences between the assisted driving technologies used by Waymo, Tesla, and comma.ai?

Waymo basically prioritizes accuracy of surroundings. This is why they started to operate only in certain areas after mappings have been completed. Then within those surroundings they do a mix of things including ML reinforcement learning for behavior as well as optimizations in terms of route planning.

Tesla on the other hand basically went on Musk infinite wisdom and said "humans only need eyes to drive, so we are going to do cameras only". Their methodology is essentially end to end - i.e take a sequence of frames, and train a model to predict where the car is going to go based on human driving data.

Comma Ai basically followed Tesla, except they did so with a much more hacked together approach. Their final product is good, but there is a reason Waymo is winning.

Geohot, holding the title for being #1 simp for Musk, of course subscribed to that philosophy.

Re: Comma openpilot – Open source driver-assistance

#207

This is awesome. I hope this technology continues to advance and decrease in price - but it's already a great value at this one.

The thing is, none of this is really better than any of the simpler systems in terms of use.

If you are required to pay attention most of the time, it defeats the purpose of self driving. Most humans can perform very complex tasks while keeping a car in lane, and adaptive cruise control is nothing new.

Also, being in my coworkers/friends Teslas that are using Autopilot, it almost seems like there is more cognitive load in using it in crowded areas, because you have to be very attentive to catch a disengagement.

Re: Comma openpilot – Open source driver-assistance

#208
post #142

How does it steer the wheel) Is there like an "API standard" for sending steering direction to cars via some physical port?

It requires a car with lane centering capabilities -i.e a module that commands the steering angle through canbus. Thats why not all cars are supported.

Re: Comma openpilot – Open source driver-assistance

#209
post #132

Earlier quoted context omitted.

No it doesn't? You can run stock, or any fork simply by providing the URL of the version you want to run. Where exactly is the restriction?

https://github.com/commaai/openpilot/blob/master/docs/SAFETY... - they will ban your device from the training set if you run an unsafe fork.

That seems very reasonable, doesn't it? Otherwise the fork could pollute the training data through bad modifications, and tracing that would be a pain.

Re: Comma openpilot – Open source driver-assistance

#210

Earlier quoted context omitted.

New Toyotas are encrypted and will likely not be supported.

2025 bZ4X is just the 2022 with some tweaks. Definitely does not have encrypted CAN bus. Maybe the 2026 bZ has it encrypted.

2023+ bZ4x has an enhanced version of TSK (security/encryption) and is not supported as a result.
Post reply on HN