AI

Nvidia CEO cues up the lighter side of robots and chatbots

Nvidia CEO Jensen Huang has a sense of humor, probably polished with help from his handlers, but demonstrated again in his two-hour GTC2024 keynote in San Jose.

Most of his off-the-cuff/rehearsed quips helped break up an otherwise long list of new products focused heavily on generative AI innovations, including the new Blackwell platform, which is made up of new powerful Blackwell GPUs and related innovations.

 In subtle ways, his little jokes might have served to help Nvidia handle the worries over AI taking away jobs or AI innovations alienating a sometimes tech-weary public or even posing physical dangers, in the vein of Terminator. AI worries have come up in many settings mentioned by AI innovators in recent years, including the founders of OpenAI and Tesla, among others.

When it came time in his keynote to talk about medical applications for AI, Huang revealed a medical video chatbot trained on Hippocratic principles to be used to interact with ordinary patients, offering advice for things such as whether a recommended antibiotic would be safe for a person with a penicillin allergy.

Named Diane, a short video snippet suggested the medical chatbot’s voice could be soothing and smooth enough to pass muster, even if it might leave some human medical practitioners chagrined. But chatbots are everywhere already and Nvidia as an AI leader clearly wants to be able to show its chatbot muscle.

“The future is generative, as in these chatbots, where billions of parameters are used,” he said.

Huang said there is already a chatbot used inside Nvidia to help chip designers build chips, and also the ability to use a chatbot to check for bugs in a new chip design. After listening to the bug chatbot every morning, a designer might get overtaxed and need to use a psychologist chatbot, he quipped.  

“After a chat with the bugs database, you need therapy—another chatbot,” he mused. ( Diane comes to mind.) The audience of 15,000 AI and chip developers clearly appreciated the joke, although maybe it wouldn’t be so funny for a broader audience.   Later, he suggested how a therapist chatbot might coach a patient, “You can do it!”

The most-appreciated light moment of the keynote came near the  keynote’s end when Huang introduced a humanoid robot concept called GROOT, first explaining its tech foundation, and then standing onstage with animations projected onscreen showing how the humanoid concept might look, with similar height to a person. At that point, he introduced two Disney-designed robots, each less than 3-foot tall and named Orange and Green. They both toddled onstage and acted like typical children, with one following directions eagerly and the other shyly holding back.  Both were created using Nvidia’s robotics platform.

While Orange and Green were evidently meant to show robots can be innocent and funny, there was only a quick keynote demonstration of what developers will actually do with GROOT humanoid robots—no fancy dancing, just short videos of robot arms handling objects for industrial uses and one performing some simple dance moves with arms only. GTC2024  is, after all, a pro-tech gathering for the true believers in AI development, so robot ethics were not on the keynote agenda.

RELATED: GTC2024: Jensen unveils AI's next phase: Blackwell GPU, humanoid GROOT, NIM runtime