The University of California, Berkeley recently launched the new Center for Human-Compatible Artificial Intelligence, whose goal is to make sure future artificial intelligence (AI) systems will help humanity.
The new center, which received a $5.5 million grant from the Open Philanthropy Project, is expected to find ways that will guarantee AI systems, despite whatever technological advancements they achieve, remain under human control.
Headed by UC Berkeley professor Stuart Russell, the team plans to explore an approach called “inverse reinforcement learning,” which will see that researchers design AI systems capable of learning human values by watching humans and imitating their everyday actions, however mundane they may be. According to Russell, this beats having designers “specify” the AI’s values, which could probably lead to a disaster.
Technology experts like Stephen Hawking and Elon Musk have been tirelessly warning of the possibility that advanced AI systems will overthrow humanity in the future. Musk, in particular, imagines a future where humans are crushed by a sentient computer overlord—a horrific version of Skynet waiting to happen.
But Russell, an advocate for incorporating human values with AI system designs, believes otherwise. As the story states in Berkeley News:
"Russell is quick to dismiss the imaginary threat from the sentient, evil robots of science fiction. The issue, he said, is that machines as we currently design them in fields like AI, robotics, control theory and operations research take the objectives that we humans give them very literally. Told to clean the bath, a domestic robot might, like the Cat in the Hat, use mother’s white dress, not understanding that the value of a clean dress is greater than the value of a clean bath."
The professor, however, admitted that it's not an easy problem to undertake since "humans are inconsistent, irrational, and weak-willed," and human values vary from one location to another.
“Instead of pure intelligence, we need to build intelligence that is provably aligned with human values,” Russell wrote, noting that the output could be educational not only for the robots in the future, but for humans as well.