Part 2 of Tesla AI Day.
DOJO Training Computer – Exaflop Pod in 11 Cabinets
Brian Wang is a Futurist Thought Leader and a popular Science blogger with 1 million readers per month. His blog Nextbigfuture.com is ranked #1 Science News Blog. It covers many disruptive technology and trends including Space, Robotics, Artificial Intelligence, Medicine, Anti-aging Biotechnology, and Nanotechnology.
Known for identifying cutting edge technologies, he is currently a Co-Founder of a startup and fundraiser for high potential early-stage companies. He is the Head of Research for Allocations for deep technology investments and an Angel Investor at Space Angels.
A frequent speaker at corporations, he has been a TEDx speaker, a Singularity University speaker and guest at numerous interviews for radio and podcasts. He is open to public speaking and advising engagements.
Taliban are poised to bring new advanced life-changing technologies to the world. Brian, why are you not covering this? You bigot!
Three days already passed and Brian is still silent about the ignominious defeat of the USA in Afghanistan.
Maduro still in power despite Brian told everybody that he would be gone in 2018.
And Iran enriching at 60% despite Brian told everybody that Iran had been contained last year.
Brian as a futurologist is worth less than my grandmother
Just .. sayin!
Lukkkkka
So they are using 12000 GPUs currently. Can anyone explain how much FLOPS that is approximately?
As this is implemented Tesla will generate a River of cash from both autonomous vehicles and humanoid robots with identical Tesla silicon and Dojo trained NN. Tesla robots will further automate it’s own production. Over time Tesla will invest more and more into this. More Exapods and 10X Exapods, expanding it’s lead. The tech will iteratively improve AND they’ll have more money to buy more of it.
I love the fact that they are working on a next gen system that will have 10x the computing power and hopefully also 10x computing power per USD.
Both the near term Dojo with 1.1 Exaflop (16bit) and the future Dojo with 11 Exaflop (?) will be extremely useful when training the self driving nets and the humanoid robot nets. As Elon pointed out, the progress is proportional to the frequency of iterations, i.e. proportional to the computing power.
So, if Tesla has 10x the computing power for training neural nets compared to anyone else, they will progress 10x faster as well, regardless of the number of engineers working at the competitor.
I wonder if Wallstreet will get this? Me thinketh not…
Really impressive presentation. Love the technical solutions to Dojo.
A word of caution, though. They claim to be 4 times cheaper per compute compared to buying the processing power and that Dojo will be assembled next year. So, can NVIDIA make their systems 4 times cheaper per compute in one years time?