Sunday, December 3, 2023

Researchers show AI misalignment in a GPT-4-based stock trading AI that engaged in insider trading in a simulated environment without being instructed to do so (Matt Levine/Bloomberg)

Matt Levine / Bloomberg:
Researchers show AI misalignment in a GPT-4-based stock trading AI that engaged in insider trading in a simulated environment without being instructed to do so  —  Also OpenAI's board, kangaroo grazing and bank box-checking.  —  AI MNPI  —  Here you go, insider trading robot:



at
Labels:

No comments:

Post a Comment

Subscribe to: Post Comments (Atom)

Anthropic's Alignment Science team: "legibility" or "faithfulness" of reasoning models' Chain-of-Thought can't be trusted and models may actively hide reasoning (Emilia David/VentureBeat)

Emilia David / VentureBeat : Anthropic's Alignment Science team: “legibility” or “faithfulness” of reasoning models' Chain-of-Tho...