Still, there are other stages of AI to pay attention to as well. “In the realm of science fiction, the drama of AI is almost always depicted in moving up the three levels from tool to autonomous agent to full sentience,” Nag said. “The drama comes both from how this brings increased danger as AI diverges in its goals from ours, but also in how we understand ourselves by watching deeper reconstructions of what makes us human.”
What exactly do these levels mean and what phase of AI are we at currently? Well, really, we’re at the start of it — in the “tool” zone. The tool level, Nag explained, means an AI that may be capable, but still controlled by humans — like Paul Bettany’s J.A.R.V.I.S. in the Marvel Cinematic Universe. An autonomous agent is an AI that has an extent of freedom but still operates within certain parameters — think Data (Brent Spiner) in “Star Trek: The Next Generation.” Full sentience is essentially what J.A.R.V.I.S. attains after becoming Vision, operating outside of original programming to be as contradictory and surprising as the human experience.
That last level is what would apply to a Skynet-style scenario, but as Nag pointed out, AI poses far more immediate concerns. “In reality, the initial danger of AI won’t come from the last stage of AI (sentience), but from the first (tool),” Nag said, giving an example of a Hong Kong financial worker who was tricked into paying out $25.6 million with an elaborate deepfake conference call hoax. “This is the danger that will define the next few years, not anthropomorphized supervillain AI (yet).”
