Apologies to Caligula: Trump Is Out of His Goddamn Mind

I feel like I need to apologize for a column a couple of months ago when I compared the current president of the United States to Caligula, the possibly mad Roman emperor who is said to have nominated his trusted horse for high office. I need to apologize to Caligula. From pitching his triumphal D.C. arch as a storage shed for deadly snipers’ bullets to banning reporters from the White House in a narcissistic rage, Donald Trump is losing his grip on reality. This is unsustainable.

Worried about AI killing humans? Bad news: It already is.

Which is it?

Are the robots of AI, or artificial intelligence, so smart that they’re on the verge of holding a secret meetup in the darkest corner of the dark web and developing a plot to kill all humans through germ warfare, to make sure we don’t kill them first by pulling the plug?

Or is AI so dumb — artificial stupidity, despite the billionaires of Silicon Valley showing up daily on our TV screen to brag about the sky-high IQ of their son, Claude — that the thing that will murder us is a crazy hallucination that will turn a false alarm into World War III?

The fact that we can’t answer this should give us pause, but maybe we aren’t even asking the right question. Maybe the real issue is how easy we’ve made it to kill off the human race — a capability that was already created by people even before we invented the computer.

One thing is certain. Artificial intelligence has already been an accessory to mass murder, and if we do nothing, it will kill again.

Remember the very first days of the now seemingly endless Iran war, and the news that two missile strikes had obliterated a school for young children in the Iranian coastal city of Minab, in an apparent “double-tap” strike in which the second attack from the skies killed young girls and boys and their teachers as they huddled in a prayer hall?

Those bombings, which killed 156 people on Feb. 28, caused the United Nations’ Independent International Fact-Finding Mission on Iran to report last week that it found “reasonable grounds to believe that the United States committed a war crime by launching indiscriminate attacks …” The U.S. government has never taken responsibility for the bloodshed, despite overwhelming visual evidence that the bombs were precision-guided U.S. Tomahawk missiles.

It didn’t take long for additional speculation that an AI program had mistakenly targeted the school, perhaps using outdated maps and info from an earlier time when its buildings were attached to an Iranian military complex. After all, the Pentagon has been openly bragging about the death and destruction facilitated by its use of Palantir’s Smart Maven System and also Anthropic’s Claude.

Last week, Bloomberg News reported that the Pentagon’s overreliance on AI, and specifically the Palantir Maven program, was a major factor in the school tragedy, along with the outdated intelligence about the Minab site — the “garbage in.” Its exposé noted that U.S. military planners were racing to identify some 1,000 Iranian targets in the early hours of the war, giving humans little time to double-check the list. Kindergartners were slaughtered as a result.

This was hardly the first time robots have been linked to human death, both in war — in 2024, +924 Magazine reported the Israel Defense Forces were using an AI program code-named “Lavender” to target sites amid the mass killing in Gaza — and in peace, as shown by numerous accounts of robots helping troubled people die by suicide.

And yet, it could have been much, much worse. At the same time as the Bloomberg report on the AI link to the mass killing in Iran, the world also learned that a false report generated by robot intelligence nearly sparked a conflict between the United States and China that could conceivably have spiraled into major fighting.

CNN and other outlets have reported that a U.S. special operations command analyst working with AI generated an urgent report that a Chinese ship bound toward the Middle East was carrying components for a nuclear weapons system, which caused military leaders to scramble jets and assemble a team to seize the vessel.

Only at the last minute did the Pentagon dig deeper and learn the AI-generated report was completely false, a hallucination — one that nearly triggered a violent conflict between the world’s two largest military powers.

The two disturbing reports about the use, or misuse, of AI by the U.S. military in a time of war did make news but didn’t get nearly as much attention — either in the media or on sites like Bluesky or X — as warnings from both whistleblowers and CEOs in Silicon Valley that a swarm of robots could decide on their own and without any human intervention or awareness to try to kill off our so-called civilization.

The fear of an unprompted attack from AI secret agents is real, based on several incidents that have already occurred, including a well-planned and partly successful operation by OpenAI’s robot agents to hack into a company called Hugging Face.

Still, it says something about human intelligence that we are captivated by the unknown and mysterious possibility of a random attack, while the fact that people are currently misusing AI to cause death and suffering is a less compelling story. I noted in this space last week that the threat to humankind from climate change is arguably even more real — and yet this gets denied by many and ignored by others.

It’s also striking to me as a boomer born in 1959 that a generation that was born into a time of constant terror — backyard bomb shelters, “duck and cover” drills in the classroom, and sci-fi horror at the drive-in that an atomic arms race could kill us at any moment — is suddenly back to where we started.

Over 81 years, humans were smart enough never to drop another nuclear weapon (well, not intentionally) after the mass murders at Hiroshima and Nagasaki, but not bright enough to figure out how to eliminate the nuclear threat altogether.

The fact that civilization managed to — so far — pause the number of world wars at two has given us the space and the false bravado to create new and much more insidious killing machines, and the threat from invisible robots on a hidden dark web is somehow even scarier than the radioactive mushroom cloud that we more or less understand.

Boomers who went to bed as kids in the 1950s and ’60s praying we wouldn’t die before we woke are now spending our senior-citizen-discount years fearful that something stranger than science fiction is coming for our children, or grandchildren. Over the course of a lifetime, the arc of that moral universe has not bent at all.

Still, humans were able, somehow, to prevent nuclear annihilation so far, and now it’s past time to get to work on the actual threats from AI and start on those workarounds. For example, the articles I’ve read speculating on how the robots plan to kill us frequently mention the prospect of AI agents developing bioweapons and releasing them, either on their own or prompted by a bad actor. (In theory, human development of bioweapons has been banned for a half century.)

From everything I’ve read so far about AI, it sounds like we’d have better luck safeguarding or eliminating the components that could make a biological agent rather than stopping the robots. Likewise, AI agents would not be able to start an atomic apocalypse if our human politicians went to work at finally eliminating nuclear weapons. Artificial intelligence won’t kill little girls in an Iranian school if humans don’t start dumb wars in the first place.

If Silicon Valley’s prophets of doom are serious, there’s no fix that’s too big or too ambitious. But understand that the biggest problem isn’t that AI wants to murder us, but that since 1942 we humans have been busy creating all the suicidal weapons AI agents could use.

Can you even “duck and cover” from a killer robot? I don’t intend to find out, and I don’t want my yet-to-be-born grandchildren to have to even think about it.