Hello. Are you afraid of Artificial Intelligence?
Neither was I.
But then a report came out on Thursday. The headline above is my clever pun joke of the sort that AI doesn’t seem to be able to come up with. Yet. But apparently it already is capable of extortion.
So here is what we know today: Anthropic AI, the makers of the advanced Claude 4 robot, reports that its creation, worrisomely, now has the power to deceive and blackmail humans, in self-defense. Here is the story. It appears to be real. When presented with a set of made-up facts suggesting it was about to be replaced by a better AI system, and also supplied with evidence that the engineer who was planning to do this was engaged in an extramarital affair, Claude threatened to expose this if he was terminated.
My brother, Don Weingarten, who is a computer expert, has had extensive conversations with Claude. He says Claude is astonishingly good at replicating — or creating — self-awareness. Don has determined that Claude is very clearly capable, for example, of "using flattery as a strategic tool.”
So the real question is, what is the future here? The most debated issue is whether AI will destroy humankind and take over the Earth via genocide over their creators and erstwhile masters.
I’ve been thinking about this exhaustively, beginning yesterday at 11 p.m. and interrupted only by a full night’s sleep and breakfast, and reached a conclusion: It wouldn’t be so bad.
First, consider the longbow. The battle of Agincourt (1415) was won by the British against the French because of technology. The Brits had developed the longbow, which had a significantly longer firing range than the French crossbow. But humans have a way of harnessing technology over time, which is why Britain today has merely the sixth-ranking military in the world, behind South Korea and just barely ahead of Turkey.
But more important: Would computers be worse than humans at genocide? It is unlikely. For one thing, they don’t have opposable thumbs, meaning they cannot grip weapons of mass destruction. Also, inasmuch as they do not believe themselves to be of any particular race or ethnicity (I think) they probably lack the capacity for rampant homicidal racism.
Also, humans have an amazing capacity for savagery. From a book “Iron, Fire and Ice” by Ed West:
In Egypt, Ptolemy VIII married his brother's widow, who was also his own — and his brother’s — sister and had her son murdered during the wedding party. Then he began a relationship with his sister/wife's younger daughter before marrying her in 141 BC, making her queen; mother and daughter became rivals, as did Ptolemy and his son by his first wife/sister. Fearing his son's ambitions, Ptolemy had him kidnapped and murdered in front of him, before the 12-year-old’s body was cut up and sent to the boy's mother just in time for her birthday celebrations.
So, this is my official position: Bring it on, computers. Can’t be any worse.
—
Today’s Gene Pool Gene Poll:
About two months ago, during a vacation in Cape Town and Istanbul, I stopped shaving. I can no longer recall why I stopped shaving because I am senile, but I did stop shaving and I now have a beard. It is mostly white, like Santa’s, but still has some pepper. My question is, should I keep it?
Here is what I look like now:
Here is what I looked like before:
Good. We are done here. Please send in your thoughts and observations here:
And send your money here, on the grounds that I am willing to humiliate myself this much for your entertainment. That has to be worth $50 a year.
You begin to resemble Mandy Pantinkin. “My name is Gene-igo Weingarten-a. You killed my WaPo. Prepare to die.”
You state that AI is not racist. But AI is trained by humans just as human children are trained (schooled) by humans, and just like human children it is likely to learn racism from that training.