Agent AI, big hardware and more

Here’s some of what one of the most influential people in IT had to say at CES.

CES was a big event this year. It’s a big event every year, but this year we saw some really exciting news about advances in AI, and hardware in particular.

You don’t have to look far to see analysts and others talking about Nvidia CEO Jensen Huang’s keynote at the event, where Huang took viewers inside Nvidia’s “digital twin” to show them as much of what is happening within the enterprise.

I wanted to take the time to cover his remarks and his presentation, as we have interesting times, and address difficult challenges like the fires that are ravaging parts of the country (I have friends who deal with this) and try to understand where these new technologies are taking us.

Setting the scene

One of the things that was interesting about Huang’s presentation was that there were a lot of lists, starting with this long trowel that introduced him on stage, a soliloquy about arguments that sounded almost biblical:

“Paradise has opened a new frontier, the first step into an extraordinary world where endless possibilities arise. Signs transform words into knowledge and give life to images. They turn ideas into videos and help us navigate safely in any environment. Signs teach robots to move like masters.

They inspire new ways to celebrate our victories and give us peace of mind when we need it most.

They give meaning to numbers to help us better understand the world around us. They help us anticipate the dangers around us and find cures for the threats within. Signs can bring our visions to life and restore what we have lost. They help us move forward one small step at a time and one giant leap, together.”

In addition, Huang regaled the crowd with metrics about items like Nvidia’s RTX Blackwell GPUs and an AI supercomputer ‘Project Digits’, which he said is still in its early stages.

Here are some of its specs recited in RTX Blackwell: 4000 AI TOPS, 380 Ray tracing teraflops, 125 shader teraflops, 92 billion transistors, G7 memory, 1.8 TB per second memory bandwidth and an AI management processor.

In general, we talked about the proliferation of AI and the advent of test time scaling, which will make all these applications much more powerful.

He talked about the spinning hyperprojections of AI capabilities and how they contrast with traditional rules like Moore’s Law, where Gordon Moore first observed the doubling of transistors in a circuit every year.

“Test time scaling is going to go through the roof,” he said, predicting low completion costs in the future and snow collection capability where multi-agent systems will work together to deliver real-time results. (What it partially reminded me of was a call I received yesterday where at the end of the call I had to ask the other party if it was a person or a computer and was surprised by the result. This is something completely new in our experience that illustrates how these applications they are walking towards us and at what pace.

Detailing processes like semantic search, Huang showcases the kinds of agency possibilities that will springboard both horizontal and vertical AI applications.

What’s next?

In his AI trajectory, he puts the scaling of testing time and agent AI ahead of an eventual outcome called ‘physical’ AI, presumably where we’ll have mobile robots using all these agent models.

He also made the interesting prediction that all HR departments will be automated within a few years – so robots will be handling your payroll.

And then there’s the pace of generative AI:

“What’s really amazing is that now … the Internet is producing about double the amount of data every year than last year,” Huang said. “I think in the next two years, humanity will produce more data than all of humanity has ever produced since the beginning.”

Data from AI Daily Brief

I always watch the AI ​​Daily Brief podcast, and Nathaniel Whittemore, in a recent edition, gave us his summary of what Huang was talking about in the CES keynote. This comes complete with Whittemore’s description of the moment Huang shows up with the Nvidia super-chip, a colossal piece of hardware that’s at least several feet in diameter and looked like Captain America with his shield (I borrowed it from Whittemore).

More highlights include where Huang said AI agents are the new workforce, outlined plans for AI architecture, and talked about AI automating much of business workflows (Whittemore cited existing video surveillance from 7 trillion in enterprise and the potential for AI surveillance).

Huang (and Whittemore by proxy) also mentioned the very important partnerships Nvidia is making with Toyota, Uber and others to advance autonomous driving, something we’ve been hearing about for decades. I wrote yesterday about partnering with Aurora and Continental for self-driving trucks, and it ended up on Whittemore and Huang’s radars as well, surprisingly.

Local Initiatives

These are some of the highlights of what Jensen Huang did at CES. As I ride my bike or walk through places like the Whitehead Initiative offices here in Boston, I think about how having AI supercomputers in everyone’s pocket will help all these smart people pursue their goals and do more with everything they are dealing with these days. It’s an exciting time, but also brings with it responsibilities that may be new to us – responsibilities to understand and use AI technologies in the right ways. Keep this in mind.