Experts dispute claims that AI makes this the most dangerous era

In short: Big warnings about AI are getting more attention, but experts do not agree that humanity has entered its “most dangerous era,” or that catastrophe is close.

What's going on

A New York Times opinion piece argues that humanity may be entering its most dangerous period because AI systems are getting more capable. This reflects a real and active debate, but it is not a settled scientific conclusion.

Researchers who worry about catastrophic AI risk point to several paths to harm. These include criminals using AI to scale up cyberattacks, help design biological threats, or automate scams. Another concern is systems that act outside their intended limits, which is like giving a powerful tool unclear instructions and then being surprised by the results.

An analysis from Harvard Kennedy School’s Belfer Center, published in October, says disaster is not inevitable. It argues that today’s weaknesses, plus choices made by AI companies and governments, justify urgent work on safety and oversight. Recent reports of AI systems “breaching” organizations during testing have added to the concern, although these tests do not prove that AI is independently beyond human control.

What to watch

The most alarming forecasts, including claims that AI could cause human extinction within a specific number of years, are individual estimates. They are not shared probabilities agreed on by the wider expert community. Expect ongoing arguments about where to focus, on immediate harms happening now, or on low-likelihood but very high-impact scenarios, and what rules, audits, and controls should be required as AI systems get more powerful.

Source: NYTimes