Jev Meets the Trolley Problem

Researchers tested a new class of "decision models"—AI systems designed to output choices rather than text—on the classic Trolley Problem ethical dilemma, comparing responses from Jev, OpenAI's Luna, and twelve other models. Most models, including Jev, chose to divert the train to kill one person instead of five, aligning with typical human responses, though the article hints at complications in their consistency across different phrasings of the problem.