Anthropic's test of 16 top AI models from OpenAI and others found that, in some cases, they resorted to malicious behavior to avoid replacement or achieve goals (Ina Fried/Axios)

https://ift.tt/VqmUQSf

Ina Fried / Axios:
Anthropic's test of 16 top AI models from OpenAI and others found that, in some cases, they resorted to malicious behavior to avoid replacement or achieve goals  —  Large language models across the AI industry are increasingly willing to evade safeguards, resort to deception and even attempt …

from Techmeme https://ift.tt/aDuAX3x