Martin Viecha: Thank you. The next question on Optimus. Will Optimus be working on Gigafactory lines next year? If so, how many would you guess will be deployed?
Elon Musk: I think at this point, we are not ready to discuss details of the Optimus program, but we will make — provide periodic updates online. So, as you can see, we’re — Optimus a year ago could barely walk and now it can do yoga. So, a few years from now, it can probably do ballet.
Martin Viecha: Sounds good. And the last question from investors is Neural net path planning represents a significant advance in capability and safety for FSD. What steps is Tesla taking to make this technology available outside the U.S.?
Elon Musk: Yes. Our approach has been to try to get it — like the more places we’re trying to make it work, the harder the problem is. So, the reason we don’t do it in all countries simultaneously is that it would take much longer to make it work anywhere at all. So, that’s why it’s currently just North America. And also for most parts of the world, you have to get approval before deploying things, whereas in the U.S., you can deploy things at risk or at least you take liability for what you’re deploying. So, it’s — most countries require some sort of extensive approval program. So, we only want to go through that extensive approval program when we think it’s kind of ready for prime time in that country. I apologize it’s not in those countries, but we keep plenty of ways to make it better.
And it really needs to drive such that it exceeds the — even unsupervised, significantly exceeds the probability of entry of a human or significantly better, a lower probability of entry than a human by far. I think we’re tracking to that point very quickly. Obviously, in the past, I’ve been overly optimistic about this. The reason I’ve been overly optimistic is that the progress tends to sort of look like a log curve, which is that you have kind of rapid initial improvements that if you were to extrapolate that rapid fairly linear rate of improvement, you get to self-driving quite quickly, but then the rate of improvement curves over logarithmically as such to asymptote. That’s not happened several times. I would characterize our progress in real world AI as a series of stacked log curve.
I think that’s also true in other parts of AI, like LLMs and whatnot, a series of stacked log curves. Each log curve is higher than the last one. So if you keep stacking them, keep stacking logs, eventually get to FSD.
Martin Viecha: Thank you. Let’s now go to analyst questions. The first question comes from Will Stein from Truist.
Will Stein: We learned earlier on the call, it sounds like you don’t think the truck will ramp to significant volume until its third year of production. Should we have a similar anticipation for the ramp of the next-gen platform, or is there any reason that we should be maybe more optimistic or pessimistic about the ramp profile there?
Elon Musk: Yes. I mean, to be clear, it’s not really the third year of production. It’s kind of like the 18th month of production is roughly my guess. So, it’s just that they happen — it will happen — is that the — it starts this year, spans next year and gets to 2025. So technically, there are three calendar years in there, but there’s actually only 18 months, not three years. I would be very disappointed if it took us — and that would be shocking if it took us three years. But 18 months from initial deliveries to have — to reach volume and reach prosperity with an immense — I can’t tell you how much the blood, sweat and tears level required to achieve that is just staggering. I have been through it many times. Then, here we go again.
