Breaking News
Deporting Illegals Is Legal - Military In America's Streets Is Not!
Turn Your Homesteading into a Farm (Making Money on the Homestead) | PANTRY CHAT
"History Comes In Patterns" Neil Howe: Civil War, Market Crashes, and The Fourth Turning |
How Matt Gaetz Escaped Greenberg's Honeypot and Exposed the Swamp's Smear Campaign
Top Tech News
Forget Houston. This Space Balloon Will Launch You to the Edge of the Cosmos From a Floating...
SpaceX and NASA show off how Starship will help astronauts land on the moon (images)
How aged cells in one organ can cause a cascade of organ failure
World's most advanced hypergravity facility is now open for business
New Low-Carbon Concrete Outperforms Today's Highway Material While Cutting Costs in Minnesota
Spinning fusion fuel for efficiency and Burn Tritium Ten Times More Efficiently
Rocket plane makes first civil supersonic flight since Concorde
Muscle-powered mechanism desalinates up to 8 liters of seawater per hour
Student-built rocket breaks space altitude record as it hits hypersonic speeds
Researchers discover revolutionary material that could shatter limits of traditional solar panels
Tesla Dojo, Supercomputers and AI
There does not seem to be a limit for neural nets to utilize more resources to get better and faster results.
Tesla is motivated to develop bigger, faster computers that are precisely suited to their needs.
The Google TPU architecture has not evolved as much over the last 5 years. The Google TPU chip is designed for the problems that Google runs. They are not optimized for training AI.
Tesla has rethought the problem of AI training and designed the Dojo AI supercomputer to optimally solve their problems.
If Tesla commercializes the AI supercomputer that will help to get to lower costs and greater power with more economies of scale.
One of the reasons that TSMC overtook Intel was that TSMC was making most of the ARM chips for cellphones. TSMC having more volume let them learn faster and drive down costs and accelerate technology.
99% of what neural network nodes do are 8 by 8 matrix multiply and 1% that is more like a general computer. Tesla created a superscalar GPU to optimize for this compute load.