There’s lots of discussion about human-centered AI, which augments the human experience rather than replaces it. But aren’t humanoid robots a displacement technology?
It’s a touchy subject for many. The fact is we’ve been building a pyramid scheme throughout mankind’s existence. Every generation banks on the next generation paying for the people that exist today. And now we’re at a potential crisis, where the next generation in many places is actually going to be smaller. We just don’t have enough labor. Robots are the only hope we have of producing the services we need to sustain our society. How are we going to do all of that stuff if we don’t have advanced manufacturing, advanced transportation, that are using agents that are superhuman in many of those capabilities?
Your work involves creating a digital twin of the real world, called Omniverse, in which AI robots can learn how to act. How does that work?
It’s just like a video game. Omniverse is a kind of video game but different in that they’re not fantasy worlds. They’re constrained to the laws of physics of the real world as accurately as possible, so we can run a bunch of simulations and test things. The AI, the robot, thinks it’s inside the real world and can go try things much faster, because we can just throw more computing at it. In one hour of the real world, it drives millions of hours in the virtual world, and if it makes mistakes, it’s not going to harm anyone.
Were you tempted to call it the Matrix?
It’s maybe like the Matrix in that it’s indistinguishable from the real world. That’s the key. Because at some point you want to transfer the robot brain into the actual robot to operate in the real world. And if what it’s trained on is a cartoony, non-realistic version, then it’s not going to operate well.
What do you see as the first uses for AI robotics?
It’s always a difficult thing to predict. There’s $100 trillion of stuff that could be helped by robots with general capabilities. I’m particularly focusing on the industrial world, manufacturing, warehousing, supply chains, and logistics. Construction eventually. There aren’t enough people to work in our factories. Some robots will be better working in areas where humans can’t go, like operating in a nuclear reactor, or dangerous mining.
Some people fear this technology could have military applications.
I don’t know if it’s a worry of mine. I feel like we’ve already created many technologies that are really frightening but somehow we found a way, despite all the geopolitical conflicts, to come to some understanding of what’s OK and what isn’t. All technology can be used for good or bad, but I think we can find some way to agree.
Nvidia’s CEO Jensen Huang has been outspoken about the harmful effect of U.S. export controls. Your most advanced chips can’t be sold in China, and it feels like new restrictions are unveiled constantly. Are you concerned about the trajectory of global trade?
We comply with all laws in the countries where we work. We’re a U.S.-based company and, of course, we not only abide by our rules but respect the fact that our democratically elected government has to be stewards of our country’s security. But every person on earth deserves to benefit from these new capabilities, not just the United States. And so we hope that whatever regulations and export controls they put into place also take that into account.
