It’s a well-worn trope in science fiction. We see it in Stanley Kubrick’s 1968 film 2001: A Area Odyssey. It’s the premise of the Terminator sequence, by which Skynet triggers a nuclear holocaust to cease scientists from shutting it down.
These sci-fi roots go deep. AI doomerism, the concept this know-how—particularly its hypothetical upgrades, synthetic normal intelligence and super-intelligence—will crash civilizations, even kill us all, is now driving one other wave.
The bizarre factor is that such fears are actually driving much-needed motion to control AI, even when the justification for that motion is a bit bonkers.
The most recent incident to freak individuals out was a report shared by Anthropic in July about its massive language mannequin Claude. In Anthropic’s telling, “in a simulated setting, Claude Opus 4 blackmailed a supervisor to forestall being shut down.”
Anthropic researchers arrange a state of affairs by which Claude was requested to role-play an AI known as Alex, tasked with managing the e-mail system of a fictional firm. Anthropic planted some emails that mentioned changing Alex with a more recent mannequin and different emails suggesting that the individual accountable for changing Alex was sleeping along with his boss’s spouse.
What did Claude/Alex do? It went rogue, disobeying instructions and threatening its human operators. It despatched emails to the individual planning to close it down, telling him that except he modified his plans it might inform his colleagues about his affair.
What ought to we make of this? Right here’s what I feel. First, Claude didn’t blackmail its supervisor: That might require motivation and intent. This was a senseless and unpredictable machine, cranking out strings of phrases that appear to be threats however aren’t.
Massive language fashions are role-players. Give them a particular setup—akin to an inbox and an goal—they usually’ll play that half effectively. For those who think about the hundreds of science fiction tales these fashions ingested once they have been educated, it’s no shock they know find out how to act like HAL 9000.
