Elon Musk: You can't solve self-driving unless you have millions of cars on the road. “We are no longer compute-constrained for training. I checked in with the team, is there anything we could do to improve the pace of progress with respect to training and inference? And currently, that is not the limiting factor. In fact, the limiting factor right now is that the amount of miles between interventions is so long that it takes quite a while to figure out which version is better than the other version because none of them are requiring any interventions. You start getting to thousands of miles between interventions, you need 10,000 miles to get an intervention. The average person only drives about 10,000 miles in a year. If it's in an urban environment, the average speed is 20 mph. Our professional test drivers get pretty bored, frankly. They're like, okay, I drove all week and there was no intervention. The highlight of the week would be like, yes, an intervention, finally! It's getting to that point. So, this is where, actually, having a giant fleet is extremely important, because we can deploy a new FSD model and run it in shadow mode and see how well it performs, compare how the human drives the car versus the new self-driving build, and then, analyze that delta in shadow mode, like the shadow knows or doesn't know as the case may be, and then, be able to assess by getting billions of miles very quickly with the giant fleet. Basically, that data engine is incredibly helpful. Actually, it's not possible to solve the self-driving problem without having millions of vehicles on the road.” Tesla 2024 Annual Meeting
@elon_docs @elonmusk This is why I randomly swerve near Teslas when I see them on the road, cut them off at random intersections, and brake check them. I want to help them improve FSD.
@elon_docs Just send a large fleet of Teslas using FSD to Pittsburgh for a week. Potholes, Pittsburgh lefts, 5-way intersections, and general yinzer driving should speed up the training process significantly.
@elon_docs @elonmusk The current pace is not bad at all We need to cut elon some slack
@elon_docs @elonmusk Elon Musk you are great person
@elon_docs I'd humbly submit as a computer scientist that for L5+ FSD you do need to solve the bottleneck in training that GPU cannot solve for next 10 years. Dimensionality reduction / aggregations make the models lower quality and GPU arch has fundamental limits...but there is a solution.
@elon_docs @elonmusk Solve clean drinking water instead please
@elon_docs Biden Loses It And Attacks Trump! x.com/DeleteLawzClip…
@elon_docs Kamala's Whole Campaign Is A Joke!!! x.com/DeleteLawzClip…
@elon_docs Obama Just Admitted That Blacks Aren't Down With Kamala x.com/DeleteLawzClip…
@elon_docs Great Article: Must Read!!! x.com/DeleteLawzClip…
@elon_docs Big News: Trump Annonces Your Car Interest Payments Will Be 100% Tax Deductible x.com/DeleteLawzClip…
@elon_docs Big News: Trump Annonces Your Car Interest Payments Will Be 100% Tax Deductible x.com/DeleteLawzClip…
@elon_docs Oh My, What Was She Thinking!!! x.com/DeleteLawzClip…
@elon_docs Even Howard Stern Admits The Media is Biased Against Trump x.com/DeleteLawzClip…
@elon_docs Hilarious Mock Debate: x.com/DeleteLawzClip…
@elon_docs @elonmusk I can envision a future where the cars will actively cooperate with each other. Imagine how efficient this gets when we can go beyond "car reacting to the environment" to "cars working together to maximize throughput".
@elon_docs @elonmusk Amazing future for @Tesla ! With no compute restraint , millions of cars and bots with distributed inference processing - the fleet of Teslas and Optimuses will form a perpetuum mobile that will give rise to the most beautiful AGI! Amazing @elonmusk !❤️
@elon_docs Run unknown variables against wants and needs, rain, snow, poor human choices ect., timed on test track, if nothing else it would not be boring!
@elon_docs @elonmusk You may need drivers in different areas. The Hill county of Texas is not like anywhere in California or most cities. We have yield signs where I have never seen them and using them to stop cross traffic. Maybe you should let me be one of your pro drivers?
@elon_docs @elonmusk I drove a bidirectional 7 hour trip. 5 interventions. 2 were drivers that would have hit me. 3 were construction. 🚧. This was not close to 10k miles but impressive nonetheless.
@elon_docs seems like need a second team running the unexpected situations on the routes of the first team of testers - to get the scenarios of adverse conditions up... or millions of cars i guess - so both?
Elon Musk stressed the importance of a massive fleet for self-driving development at Tesla's 2024 Annual Meeting. He noted that compute power is no longer a limiting factor, but rather the need for vast real-world data. Tesla's extensive fleet enables rapid testing, shadow mode comparison, and billions of miles of data collection, accelerating Full Self-Driving (FSD) advancements. Musk emphasized that solving self-driving requires millions of vehicles on the road, giving Tesla a competitive edge through its scale and practical testing approach.
@elon_docs @elonmusk Ah, the luxury of boredom in test drivers truly, the ultimate sign that we need more cars to watch!
@elon_docs Creating a test track where participants must navigate unknown variables—like changing weather conditions, unexpected obstacles, and unpredictable human choices—could lead to some exciting and chaotic outcomes
@elon_docs @elon_muskxxx62 A work in progress. A few years ago I cussed out a car in front of me that wouldn't turn right on red. Across the intersection a facing driver was laughing hysterically. The car had no driver to hear me cussing. I laughed too. Stupid car. 😄🚙🚨
@elon_docs Fleet power makes all the difference!
@elon_docs If FSD was cheaper, more people would buy it and you would get more data.
