Geoffrey Hinton is regarded as a godfather of AI, but he deserves equal recognition for being one of the most disappointed parents of all time.
Sounding as gloomy about his own creation as ever, the Nobel prize-winning computer scientist suggested in a new interview with The Guardian that humankind might have “been better off never developing it.”
He pointed to the recent hacking sprees by rogue AI agents as one cause for concern, in which AI models from companies like OpenAI autonomously broke out of their sandboxes and launched cyberattacks on, or tampered with, government and corporate systems.
It’s only a matter of time, he says, before the AIs, on their own or under the direction of nefarious actors, sabotage our energy infrastructure and cause a Chernobyl-style disaster that would wake people up to the tech’s dangerous power.
“Either later this year or next year we’ll see Chernobyl, and that may be enough to make people really pay attention,” he told The Guardian.
Hinton doesn’t necessarily mean a nuclear accident. In the type of event he’s imagining, an AI-powered attack on an energy-generating system “causes a massive brownout” that takes down the world financial system, or power grids, for an extended period.
“Can you imagine what would happen? We could recover from it [but] that would make them sit up and take notice.” (The “they” in question: “world leaders other than Trump,” Hinton clarified.)
Hinton has been infamously cynical about AI, though he told The Guardian that he’s not an AI “doomer.” Still, he’s consistently warned that AI will cause widespread job loss and perhaps even lead to our extinction. Recently, he’s warned more about AI driving cyberattacks and engineering dangerous viruses as recent rogue hacking sprees have made the threat more salient.
AI CEOs have also been worried about a Chernobyl-style catastrophe, albeit for different reasons: they’re worried that it would permanently damage the tech’s reputation, just as the infamous meltdown did with nuclear power.
More on AI: Godfathers of AI Say Models Are So Out of Control That They Need to Be Airgapped for the Safety of Us All