The battle for the soul of AI

We discuss the announcements from last week’s GPU Technology
Conference, along with AI hardware news from elsewhere, and look
at Microsoft’s proposed acquisition of language AI specialist
Nuance for $16 billion – actually closer to $20bn once you
account for debt.





We start with our regular segment chronicling the revolution in
hardware for AI – it’s Chip Wars! When Joe Biden is waving around
silicon wafers, you know interesting things are going to happen.


Nvidia is busy building its first ever CPU to support machine
learning workloads, codenamed Grace – after the absolute legend
that was rear admiral Grace Hopper.


Plus, Nvidia’s DGX family of ‘building blocks’ for supercomputers
will now feature Data Processing Units (DPUs) by default, running
a wide variety of networking tasks. DPUs weren’t originally
developed at Nvidia – these came out of the Mellanox acquisition,
which equipped the company with clever networking silicon.


Intel’s Habana – which has designed its own family of chips for
AI – has landed a contract with the San Diego Supercomputer
Center, and will build a supercomputer called Voyage. But is it
the kind of customer that Intel needs at this point?


Meanwhile, SambaNova – which has designed its own family of chips
for AI – has announced a massive $676 million funding round, just
a few mounts after emerging from stealth. Its CEO Rodrigo Liang
appeared on this podcast just a few short weeks ago – making
participation a sure indicator of future success.


Next, we talk about Nuance, the AI company built from countless
acquisitions, now being acquired by Microsoft. The speech
recognition and language specialist has helped shape the emerging
virtual assistant market – can it give Cortana a shot in the arm?


We also cover: Shopping for prying mantises! The importance of
haircuts! Chaos at Arm China! And there’s even a rendition of the
national anthem of the USSR.


As always, you can find the people responsible for the circus
podcast online:


Max Smolaks (@maxsmolax)

Sebastian Moss (@SebMoss)

Tien Fu (@tienchifu)

Ben Wodecki (@benwodecki)