According to recent announcements on news media, Meta Platforms is set to launch the latest version of its large language model, Llama 3, in July.
Meta logo – 3D artistic version. Image credit: Mariia Shalabaieva via Unsplash, free license
The aim is to enhance the model’s ability to respond more effectively to contentious questions from users.
Meta researchers are reportedly working to make the model more flexible, allowing it to provide context even in response to queries it deems controversial. This development is part of Meta’s broader effort to improve the utility of its new Large Language Model (LLM).
Meanwhile, the move comes as Google temporarily halted the image-generation feature on its Gemini AI due to inaccuracies in historical images produced by the system.
Tests conducted by The Information revealed that Meta’s current language model, Llama 2, refuses to answer certain less controversial questions. For instance, it avoids responding to queries related to pranks, warfare strategies, or car engine shutdown techniques.
However, with the upcoming Llama 3, Meta aims to address this limitation. The newer version is expected to comprehend questions more effectively, allowing it to understand nuances like the difference between shutting off a vehicle’s engine and causing harm to it.
Written by Alius Noreika