Unless there is a change in how autonomous driving software is trained, it seems Tesla is destined to dominate that market, right? Tesla has the most training data consisting of video input + corresponding driver behavior. And they have the most compute power. Which twist of fate could happen, so that Tesla falls behind in the future? I could imagine that there is an invention which lets AI learn about the world by j…
Tesla FSD head to head against Mercedes Driver Assist on the same road [video]
21–30 of 90 posts
Re: Tesla FSD head to head against Mercedes Driver Assist on the same road [video]
#22Unless there is a change in how autonomous driving software is trained, it seems Tesla is destined to dominate that market, right? Tesla has the most training data consisting of video input + corresponding driver behavior. And they have the most compute power. Which twist of fate could happen, so that Tesla falls behind in the future? I could imagine that there is an invention which lets AI learn about the world by j…
I couldn't actually find the amount of data that Tesla uses for training, but I'm curious: how much is it, and is it larger than what comma.ai has?
https://www.autoevolution.com/news/elon-musk-explains-the-la...
they will be using 85K H100s (up from 35K now) for training by the end of 2024:
https://www.shacknews.com/article/139619/tesla-ai-training-n...
Re: Tesla FSD head to head against Mercedes Driver Assist on the same road [video]
#23Unless there is a change in how autonomous driving software is trained, it seems Tesla is destined to dominate that market, right? Tesla has the most training data consisting of video input + corresponding driver behavior. And they have the most compute power. Which twist of fate could happen, so that Tesla falls behind in the future? I could imagine that there is an invention which lets AI learn about the world by j…
I couldn't actually find the amount of data that Tesla uses for training, but I'm curious: how much is it, and is it larger than what comma.ai has?
Re: Tesla FSD head to head against Mercedes Driver Assist on the same road [video]
#24Unless there is a change in how autonomous driving software is trained, it seems Tesla is destined to dominate that market, right? Tesla has the most training data consisting of video input + corresponding driver behavior. And they have the most compute power. Which twist of fate could happen, so that Tesla falls behind in the future? I could imagine that there is an invention which lets AI learn about the world by j…
That visual sensors only seem like they make no sense when they can get obstructed by weather
Re: Tesla FSD head to head against Mercedes Driver Assist on the same road [video]
#25Unless there is a change in how autonomous driving software is trained, it seems Tesla is destined to dominate that market, right? Tesla has the most training data consisting of video input + corresponding driver behavior. And they have the most compute power. Which twist of fate could happen, so that Tesla falls behind in the future? I could imagine that there is an invention which lets AI learn about the world by j…
I couldn't actually find the amount of data that Tesla uses for training, but I'm curious: how much is it, and is it larger than what comma.ai has?
https://www.statista.com/statistics/502208/tesla-quarterly-v...
And I would expect that those generate more data than they can crunch, so that the bottleneck is the number of GPUs they have.
They recently reported that they have the equivalent of 40k H100 GPUs:
Re: Tesla FSD head to head against Mercedes Driver Assist on the same road [video]
#26Earlier quoted context omitted.
I couldn't actually find the amount of data that Tesla uses for training, but I'm curious: how much is it, and is it larger than what comma.ai has?
they are collecting 1B miles of data every 2-3 months https://www.autoevolution.com/news/elon-musk-explains-the-la... they will be using 85K H100s (up from 35K now) for training by the end of 2024: https://www.shacknews.com/article/139619/tesla-ai-training-n...
> This vastly expanded the FSD testing pool, with the Tesla fleet adding 1 billion FSD miles every 2-3 months
It doesn't mean they collect the video feed and all input from every second of those miles. Just that the FSD drives those miles.
Re: Tesla FSD head to head against Mercedes Driver Assist on the same road [video]
#27And yet speed limit detection and phantom breaking in my Tesla has only gotten worse the last year, or do FSD improvements not trickle down to non-FSD?
Re: Tesla FSD head to head against Mercedes Driver Assist on the same road [video]
#28The Consumer Reports comparison at the beginning of the video is Mercedes "Driver Assistance" vs Tesla "Autopilot". These are technologies for lane assistance, speed limit assistance, etc. "Advanced cruise control" is a better term for these. This video does not compare these.
Mercedes self driving system is called "Drive Pilot". Tesla's is called "FSD" (Full Self Driving)
This video compares FSD to Drive Assist, these are two different types of technologies.
Re: Tesla FSD head to head against Mercedes Driver Assist on the same road [video]
#29Earlier quoted context omitted.
they are collecting 1B miles of data every 2-3 months https://www.autoevolution.com/news/elon-musk-explains-the-la... they will be using 85K H100s (up from 35K now) for training by the end of 2024: https://www.shacknews.com/article/139619/tesla-ai-training-n...
That's not what the article says. > This vastly expanded the FSD testing pool, with the Tesla fleet adding 1 billion FSD miles every 2-3 months It doesn't mean they collect the video feed and all input from every second of those miles. Just that the FSD drives those miles.
Re: Tesla FSD head to head against Mercedes Driver Assist on the same road [video]
#30Earlier quoted context omitted.
I couldn't actually find the amount of data that Tesla uses for training, but I'm curious: how much is it, and is it larger than what comma.ai has?
They can use every Tesla vehicle for training. They definitely have more data than anyone.
There's also a matter of deduplication where they really don't need to use ~250 days a year of the same commute. They probably have more data than others, but I'm asking for a proper confirmation of this common knowledge.