Tesla Is No Longer Training Compute Constrained So AI Will Accelerate

Get ready for major progress. Tesla FSD and Teslabot will accelerate. Tesla is no long AI compute constrained.

Now we will see what unrestricted progress looks like.

7 thoughts on “Tesla Is No Longer Training Compute Constrained So AI Will Accelerate”

  1. Well, if it was just Elon saying it, then I would agree, but a few other smart people are saying it now.

  2. Unless Tesla has some breakthrough in learning algorithms, the existing scaling laws mean it doesn’t matter how much compute you have — you are compute constrained by the number of subatomic particles in the universe and the speed of light, and that’s true even if you have quantum supremacy.

  3. Like jonathan says, noone believes those claims anymore. I don’t care at all about it (I cared a lot during 2016-2020). Deliver car without steering wheel which humans without drivers license can own and use, then I will start paying attention.

    We are tired of “FSD this year” talking.

  4. The problem is we don’t believe anymore people who said full self-driving at the end of the year for like 8 years in a row. We will believe it when the regulators say it is level 5.

    • Jonathan … Yes agreed I am a heavy investor in Tesla and do actually believe that they will solve FSD this year from seeing the progress HOWEVER on this subject Elon has zero credibility he DID say year after year `this year` so his word around this is useless, showing his messages means nothing because of his past on this, I will believe when it is indeed classified as such by regulators

    • FSD will never happen until we get to Level III of Insurance Liability: That any accidents – which will ALWAYS happen, even when FSD cars are much better than human drivers – will be no fault of both driver AND car and insurance companies will have enough confidence in FSD that they will pay reasonable claims and be able to “risk-project” that enough to set a premium.
      The problem now is that FSD makes errors when humans could easily avoid them, and does well when humans would have trouble, and there’s little predicting when each eventuality will occur, so the driver has to be hyper-alert. That is actually more exhausting – including hovering ones’ hands above the steering wheel instead of resting on it – than just driving. Neither drivers nor Musk trust the cars well enough to drop that requirement, let alone regulators. Critically, Musk does not want to assume liability for a purely car-driven ride. He knows even a single accident could start a roll of lawsuits that could bankrupt Tesla. Insurance for Teslas is more expensive, not less, even adjusted for the cost of the car.

Comments are closed.