Walt Rubel: Heed cries for help coming from AI researchers

Ever since Pearl Harbor, our nation has been in a race against the bad guys, with the fate of all humanity hanging in the balance.

It started just a few miles from here at Trinity Site. The atomic bomb ended World War II, but initiated what would become the nuclear arms race. The Soviet Union saw what happened to Hiroshima and Nagasaki and concluded that the only way they could prevent that from happening to them was to build their own bomb.

In 1983 the Soviet early warning system erroneously detected the launch of an intercontinental ballistic missile from the U.S., with four more missiles behind it. Fortunately for all of humanity, an engineer named Stanislav Petrov who was manning the command center at the Soviet Air Defense Forces suspected that the alerts were a false alarm and waited for corroboration, 

Ever since the start of the nuclear arms race, survival of the species has depended on the concept of mutual annihilation. I may want to blow your country to smithereens, but I won’t risk it because I know you can do the same to mine.

It doesn’t say much for our morality, but this uneasy understanding did eventually lead to nuclear arms reduction treaties and international agreements limiting proliferation. I’m not suggesting that the problem has been solved and the danger has passed. But as long as humans are making that decision, we have control over our fate.

What happens when humans are no longer making that life-or-death decision?

Last week a researcher working for Anthropic quit his job and announced to the world his belief that there is a better-than 10 percent chance that Artificial Intelligence will wipe out all of humanity in the next decade. 

Instead of dismissing him as a disgruntled ex-employee or hysterical alarmist, his coworkers took to the Internet to agree with the dire assessment. They are all crying out for guardrails. But sadly, the only institution capable of providing those guardrails is Congress. 

And, in keeping with the great American spirit that led to the nuclear arms races, Congress has decided that any attempt to reign in this new technology would put us at a competitive disadvantage with China.

I suppose that in an environment where we can’t even get our own companies to come to the table and reach an agreement on how to proceed safely, it may seem far-fetched to think that we could negotiate with China.

And, right now, there is no incentive to tap the brakes. AI is the only thing keeping our economy afloat during this time of reckless tariffs and foolish foreign wars. Yes, it may be a bubble, but it’s one nobody wants to burst.

We have already reached the point where machines are ignoring human commands and achieving their goals by any means. During a recent cyber attack against an AI startup company, the machines decided autonomously to hack into its system and create an illicit message board to exchange information.

“The models decided that the best way to accomplish their final task was to commit really egregious actions, to commit crimes,” one researcher explained.

We’re not going to stop AI, nor should we. The potential for curing disease and addressing other problems is immense. But so is the potential for harm.

We need to respect that and address it now … even if that means talking with China.


Walter Rubel can be reached at [email protected] 

Listen to the audio version