Bill Gates now says the world should prepare for roughly two decades of AI disruption, and that the period could be painful enough to split society in two.
Gates made the case in an interview on Decoder, an interview podcast, speaking with the show's host. It follows an essay he published in late August titled "The turbulent AI era is here. The choices we make now are critical." In it he wrote that "even under the best circumstances, the transition to this new AI era will be one of the most turbulent times in human history." Since then he has called AI "the most dangerous thing that humans have ever gone near" and said it "makes nuclear weapons look like nothing."
The interview narrowed that alarm. Gates listed what changed over the past year: models that write code and find cyber flaws, reason in biology, form relationships with kids and other people, and disrupt the job market. He said the near-term danger sits with models whose safeguards can be removed, since those are the ones cyberattackers and bioterrorists can exploit. Models that are monitored for misuse, he said, pose a much smaller near-term risk. The longer-term control problem, he added, applies to any state-of-the-art model.
That framing sidesteps the usual parlor game of which lab or chief executive to trust. For Gates, the question is whether protections stay in place at all. It also explains why the host pressed on enforcement for open-source models, where users can modify the software themselves.
Gates also argued that the public is the weak link. Industry insiders are warning loudly, he said, but outside the industry few believe "this time is different," because past technology scares fizzled. He rejected the idea that the warnings are hype and called them "absolutely the truth."
His position is less a reversal than a change in timing. He still holds that AI is beneficial if shaped well. What he now says is that the response has not matched the speed of the technology, and that the shaping has to start before the disruption does.