While having a built for purpose machine for training is a huge advantage, for me another major advantage of dojo is owning the machine as capex as opposed to renting it via oppex. Getting FSD working is going to take more than a big machine and a bunch of data. They’re going to need countless iterations tweaking the training algorithms and then subsequent also computationally expensive simulations to test the viability of the trained models on a plethora of situations. If I’m having to shell out millions for each training and testing cycle I’m going to be a whole lot less iterative/agile than if it’s just the cost of electricity. In an unknown space like this I’d way rather take a try out as many things as possible approach than a justify to the board my ongoing 400m oppex spend world.
My bullshit alarm went off when I saw the first Cerebras slide...it is basic knowledge that you can run compute inside specific regions on the major clouds. No one is running half their job in Austin and half their job in Seattle. Maybe the point about the big players using dedicated infra still stands, but this is a very elementary mistake for them to call out. Weird
I'm also very skeptical about the Tesla humanoid bot program. The hardware costs alone make the economics very questionable / impossible today. I say this having had experience making angel investments in the robotics space since 2017. Even if hardware costs could be improved by an order of magnitude, a general purpose humanoid bot is extremely challenging and not even Boston Dynamics is close to producing one. I would ascribe basically zero value to Tesla's humanoid robot program today.
While having a built for purpose machine for training is a huge advantage, for me another major advantage of dojo is owning the machine as capex as opposed to renting it via oppex. Getting FSD working is going to take more than a big machine and a bunch of data. They’re going to need countless iterations tweaking the training algorithms and then subsequent also computationally expensive simulations to test the viability of the trained models on a plethora of situations. If I’m having to shell out millions for each training and testing cycle I’m going to be a whole lot less iterative/agile than if it’s just the cost of electricity. In an unknown space like this I’d way rather take a try out as many things as possible approach than a justify to the board my ongoing 400m oppex spend world.
great take, thank you!
My bullshit alarm went off when I saw the first Cerebras slide...it is basic knowledge that you can run compute inside specific regions on the major clouds. No one is running half their job in Austin and half their job in Seattle. Maybe the point about the big players using dedicated infra still stands, but this is a very elementary mistake for them to call out. Weird
I'm also very skeptical about the Tesla humanoid bot program. The hardware costs alone make the economics very questionable / impossible today. I say this having had experience making angel investments in the robotics space since 2017. Even if hardware costs could be improved by an order of magnitude, a general purpose humanoid bot is extremely challenging and not even Boston Dynamics is close to producing one. I would ascribe basically zero value to Tesla's humanoid robot program today.