California Dreamin'

April 12, 2016

It’s day two at the NVIDIA GPU Technology Conference (GTC) at the San Jose Convention Center. There were no seismic event warnings this morning, which is surprising as I met someone at the NVIDIA networking event who was convinced that they basically cause earthquakes. As I’m someone who’s convinced that they make it rain whenever they visit Las Vegas, I could empathize.

The hot topic today was the keynote address from Jen-Hsun Huang, NVIDIA’s CEO. Here’s a detailed breakdown of what was announced.

So: what things struck me most about the keynote? First, it’s obvious that deep learning and artificial intelligence are of huge importance to NVIDIA. Three of Jen-Hsun’s five overview topics were directly related to deep learning—a deep learning chip, a deep learning box, and a deep learning car. NVIDIA wants to provide hardware and software that can produce “superhuman results without the need for super humans,” allowing machines to self-learn without the need for specialized algorithm development related to a particular task.

One demo that caught my attention came from some work on unsupervised learning. You show an AI system around 20,000 pieces of art from the romantic era and then ask the AI system to produce its own landscape based on what it had learned. The generated image looked indistinguishable from the others in style but was an exact copy of none of them. You could then ask the system to produce a landscape that includes a forest, or a beach, or a sunset or a whatever, and it could and it did. Computer imagination at work. Wow!

At that point, I had to leave to go and set up the booth for the trade show portion of the day that followed immediately afterwards. At GTC, the trade show portions of the day are two hour chunks around lunch and dinner time where the attendees can eat, drink, and chat to the exhibitors. It’s a great format that ensures that the booth gets busy and you have some great conversations without long chunks of waiting around.

As predicted yesterday, the demo kit decided that it was sufficiently close to show time to start acting up a bit, but a couple of reboots and some swearing seemed to sort it out. By the time the throng from the keynote wandered blinking from the auditorium, everything was set. Within minutes, we had folks on the booth. Our VP of Product Management was chatting about product road maps with Stuart and our partners from NVIDIA, and the lead retrieval device was ping-ponging between myself and Michael. Then, after what seemed about five minutes, the two hour session was over and it was our time to wander away. 

On the basis of this lunchtime’s session, the evening one is going to be a good one too, so I’d better go get ready. Tomorrow, we should probably talk about VR because that was also covered in the keynote and that’ll give me the excuse to go try the cool on-site demos.

About the Author

Nick Porter

Nick has been selling embedded computing solutions to military/aerospace customers for over 20 years. Initially working as a government research scientist and gaining a PhD in radar signal processing, Nick's move to the dark side was prompted by the promise of a company car. He soon discovered how fun it was to solve customers’ problems and to progress a scribbled whiteboard design into a real fielded program. He also likes cats.

Voice your opinion!

To join the conversation, and become an exclusive member of Military Aerospace, create an account today!