Ex-Google CEO warns there's a time to consider "unplugging" AI systems
 Former CEO & Chairman of Google Eric Schmidt speaks during a book talk for Fareed Zakaria's new book, "Age of Revolutions: Progress and Backlash from 1600 to the Present" at Columbia University's School of International and Public Affairs on April 3. Photo: Shahar Azran/Getty Images
figcaption>span]:font-sans" style="margin-block: 0px; margin-inline: 0px;">Former Google CEO Eric Schmidt warned that when a computer system reaches a point where it can self-improve, "we seriously need to think about unplugging it."
Why it matters: The multi-faceted artificial intelligence race is far from the finish line — but in just a few short years, the boundaries of the field have been pushed exponentially, sparking both awe and concern.
- Regulations are in a state of flux, with discussions on Capitol Hill sputtering as this chapter of Congress nears its close.
- But companies are still charging ahead.
- "I've never seen innovation at this scale," Schmidt said on ABC's "This Week." While he celebrated "remarkable human achievement," he warned of the unforeseen dangers of rampant development.
Driving the news: "We're soon going to be able to have computers running on their own, deciding what they want to do," he said.
- When the system begins to "self-improve," he said, it may be time to consider pulling the plug. Asked if a powerful system could counter efforts to shut it down, Schmidt told ABC's George Stephanopoulos, "In theory, we better have somebody with the hand on the plug."
- He continued, "the power of this intelligence ... means that each and every person is going to have the equivalent of a polymath in their pocket."
- "We just don't know what it means to give that kind of power to every individual," Schmidt added.
Threat level: A chorus of industry experts have warned of the risks of rushing the development of generative AI.
- Schmidt told Axios last year that computers making their own decisions may be only two to four years away, according to some experts.
- There's a consensus among fans and foes of generative AI that the most powerful models could operate with the intelligence of a Ph.D. student as soon as next year, Axios' Mike Allen and Jim VandeHei report.
- Within a year or two, Schmidt told ABC, systems may be able to do their own research.
Zoom out: He said that he used to think U.S. technology was a few years ahead of China — but in the last six months, Beijing has caught up "in a way that is remarkable."
- But it is "crucial that America wins this race, globally, and in particular, ahead of China."
If he were given complete control over the world's handling of AI, Schmidt pointed to two steps he'd take:
- "The most important thing I'd do is make sure that the West wins," he said, emphasizing the need for funding, hardware and manpower.
- Secondly, he said he'd identify the "worst possible cases" and build a second system to monitor the first.
- "Humans will not be able to police AI," Schmidt said. "But AI systems should be able to police AI."
|
