Yes please. Especially after reading on Heise that Anthropic did run an experiment in a simulated environment where the AI agent had access to an email account, where messages had been planted – some that alluded the owner had an affair and one where the owner hinted at that he wanted to replace the AI agent with a different one.
What did the AI agent do? Try to extort the owner by threatening to make the affair known to avoid being switched off.
After some digging, looks that was in 08/2025, but I read about it more recently.
Technology is so depressing. It used to be so exciting.
I have some colleagues who find this technology exciting. And some who find it depressing.
And especially some who are depressed that others find it exciting.
Funny. They solve lots of problems I simply don’t have. Nothing in this video triggered a “want to have” impulse. Perhaps I’m simply getting old. Or more wise.
Yes please. Especially after reading on Heise that Anthropic did run an experiment in a simulated environment where the AI agent had access to an email account, where messages had been planted – some that alluded the owner had an affair and one where the owner hinted at that he wanted to replace the AI agent with a different one.
What did the AI agent do? Try to extort the owner by threatening to make the affair known to avoid being switched off.
After some digging, looks that was in 08/2025, but I read about it more recently.
Technology is so depressing. It used to be so exciting.
I have some colleagues who find this technology exciting. And some who find it depressing.
And especially some who are depressed that others find it exciting.
Funny. They solve lots of problems I simply don’t have. Nothing in this video triggered a “want to have” impulse. Perhaps I’m simply getting old. Or more wise.