Tesla’s NEW Supercomputer ‘DOJO’ Takes The World By STORM!

Tesla Reveals New Secrets Behind Their Supercomputer DOJO!

Did we miss anything?

#LLM #Largelanguagemodel #ChatGPT #AI #ArtificialIntelligence #MachineLearning #DeepLearning #NeuralNetworks #Robotics #DataScience #IntelligentSystems #Automation #TechInnovation

[andrerodriguez]

Now Nividia is on the same track. The future will be amazing.

[azamat_bezhanov]

how many gates in gaafet transistor with backside power delivery

[lengould]

I worked in the computer industry thru the transition from large centralized compute resources to small task-sized compute resources.

It strikes me how similar the discussions are now. Seens no-one notices that AI is being designed with ALL intelligence in one huge central "training" machine, with only a very small dumb inference engine, incapable of learning, in the application point (crt tty, FSD car, humanoid robot, etc). This is not how human intelligence works.

Are we sure this path has any future? For how long? Is it that we simply don't know how to implement a proper learning Distributed AI (DAI) processor at each point of use, with high-capacity inter-DAI communication to enable requesting a good algorithm for the present problem, and have another unit provide the required trained routine. ?

[NoBobPro]

Cyberpunk fact just really put everything in perspective.

[tmarq]

I wouldn't get overly excited about Dojo since Musk has a tendency of overexgerating his projects and then see those project underperform. So far, his works haven't blown me away. I will say that he is a furuist but one whose risk levels are a bit too high for human comfort.

[johnsonjjohnson]

You are not keeping up. Dojo is a complete flop.
Musk fired the head module designer for non-performance.

He designed the Dojo module with little to no memory.
The idea was that it would make the system faster
Turns out AI is a memory pig to work efficiently!!!
So at the very least the entire Dojo project will need to be redesigned or scrapped altogether
or
Maybe repurpose the chips for a huge data center after adding memory

The good news is that FSD AND Optimus are NOT dependent on Dojo

Our original supercomputer is still the one we lean on for all FSD and Optimus iterations
The 10, 000 H100 Invidia chips went to our old supercomputer -- NOT Dojo

[visualcontrast]

If 100 exaflops can be achieved... and one FLOPS = one calculation (per second):

One way to look at it is in terms of human calculation. If one operation was equivalent to one calculation that a human could do in a second, and you had every person on Earth (let's say roughly 8 billion people) calculating non-stop, it would still take them about 4, 000 years to do what a 100 exaflop computer can do in a single second.

[ehjoepr]

What if create small rockets and start them, and physically pet chips for they grow in kind, justful and careful way :)

[BalancingEternity]

I just hope my AI owner will feed me real meat and let me play outside in the climate controlled backyard whenever I want.

[jan_phd]

People forget Musk's other project, NeuralLink... the remote control of humanity.