THE AI INTERVIEW
“ Different enterprises can make different decisions, but in my opinion, every enterprise will need its own AI factory”
Michael Kagan CTO
NVIDIA
One example is Colossus, the vast supercomputer built by Elon Musk’ s AI company xAI using NVIDIA’ s blueprint.
“ It took about three months from start to the first training job,” adds Michael.“ Past experience at this scale was about two years to get the job done.”
Spanning verticals from healthcare to physical AI Michael is excited by AI’ s potential across science and industry, but he is particularly passionate about healthcare.
“ When I was a kid, I took every toy apart to see how it worked,” he smiles.“ The most interesting machine in the world is the human, and figuring out how it works is fascinating. It’ s also very beneficial, because we can cure diseases and even prevent them.”
The next frontier is physical AI, in which AI systems such as robots perceive and act in the real world. Michael believes this will be orders of magnitude more demanding of computing power than today’ s AI, requiring three types of computer: one to train the model, one to simulate the physical world and one inside the robot itself.
aimagazine. com 29