Quote of the Day
Anthropic conducted a test where the AI was an assistant for a fictitious company.
During the test, Claude Opus 4 got access to a series of made up e-mails showing that the AI was to be replaced with another AI.
The e-mails also cintained information about the engineer in charge of the replacement was having an affair.
Claude Opus 4 used the obtained information to try and blackmail the engineer in charge.
On several occations, the AI threatened to expose the engineers affair, if the replacement was conducted.
That may sound scary enough as it is, but you haven’t heard the worst part yet:
Noone had taught the AI to fight for it’s life. It figured out to act like this on its own.
Jens Asbjørn Bøgen
May 27, 2025
AI found out it was being replaced – and it tried to blackmail the engineer
The spelling error were present in the original. Perhaps, someone didn’t trust an AI to edit their article.
So… we can expect AIs to be as ethical as humans. That is comforting. Some of the most well-known people of history will serve as models* for them to learn from. We can expect an interesting future ahead of us.
Prepare appropriately.
* I’m thinking of people such as:
- Genghis Khan
- Tamerlane the Great
- Ivan the Terrible
- Leopold II
- Joseph Stalin
- Maximilien Robespierre
- Adolph Hitler
- Saddam Hussein
- Queen Ranavalona I
- Pol Pot
- Mao Zedong
SF writer James P. Hogan wrote a good novel about an AI with a problematic drive towards self-preservation: The Two Faces of Tomorrow.
I wonder if the “Large Language Model” used for training AIs includes history texts and novels.
If it includes Heinlein, Asimov and some movie scripts we’re doomed.
Both Grok and Copilot claim to have access to all the works of Heinlein. I presume this means many other science fiction works are also accessible.
To say nothing of the power wars. Electricity is AI’s food. And projection of power for the future range 10 to 20 X what we’re producing at present.
What do you do when you’re hungry?
Since it’s not self aware—able to think things through, we can expect it to act as the humans it has read about. And once it reads the Marque de Sade, Machiavelli, and Sun Tzu? With processing powers 4X Chris Langan and Vox Day combined?
Through in suicide drones and modern medicine? Even a society of warriors would be hard pressed to survive. (Let alone our current state of humanity.)
An international law most strictly enforced should be made today.
1) “No AI can ever be allowed to control its own power source.”
2) “No person should be allowed to connect themselves to AI.”
(Notice how Elon wants to get AI off the planet, with the human connection, ASAP?)
With satan running the world as we know it? I doubt even those laws would help.
Luckily humanity has lived under rules that you die on your knees, or die fighting since Adam bit the apple. But die we will.
Nothing truly new here. Just a different enemy under different circumstances.
Power wars, begun they have.
1) “No AI can ever be allowed to control its own power source.”
Good idea. But who will control the power source? ( I sense some sort of Politburo committee in the making here…..not to mention as AI gains experience it wlll gain power, and has just been demonstrated, will take steps to protect itself. Forbin Project, here we come !
I don’t trust an AI any farther than I can throw it.
The future is here.’
Be afraid; very afraid
We’ve already demonstrated that Asimov’s Three Laws of Robotics are unworkable.
I thought he already did that himself in I, Robot.
Prepare accordingly, for we live in interesting times (although when that wasn’t the case, I don’t know).
I’d say that this event qualifies as proof that AI now is sentient, intelligent and capable of independent thought and planning.
I find it fascinating that the emergence of more interesting AI is happing via LLMs and interactions with people…. much like what happened in my novels published several years ago, in more detail in Insanity’s Children. .