E
Meta is planning a July release for a “looser” Llama 3.
The Information writes Meta developers want the next version of Llama to answer controversial prompts like “how to win a war,” something Llama 2 currently refuses to even touch.
Google’s Gemini recently got into hot water for generating diverse but historically inaccurate images, so this news from Meta is surprising. Google, like Meta, tries to train their AI models not to respond to potentially dangerous questions. That’s an important guardrail to make sure an AI that still occasionally hallucinates doesn’t tell someone how to win a war and make a bomb.
Follow topics and authors from this story to see more like this in your personalized homepage feed and to receive email updates.
Loading comments
Getting the conversation ready...
