Meta platforms to launch Llama 3, an AI language model with enhanced responses to contentious queries

    115

    29th February 2024 – (Menlo Park) Meta Platforms is gearing up to introduce an upgraded version of its artificial intelligence language model, Llama 3, in July, according to a report by The Information on Wednesday. This new iteration aims to provide improved responses to contentious questions posed by users.

    Researchers at Meta are working on refining the model to offer contextual information for queries it deems controversial. The objective is to “loosen up” the model, enabling it to better handle sensitive topics and generate more nuanced responses.

    The move comes in the wake of Google’s decision to temporarily halt the image-generation feature on its Gemini AI due to instances where historical images produced were found to be inaccurate.

    Tests conducted by The Information revealed that Meta’s current model, Llama 2, which powers chatbots on the company’s social media platforms, refuses to respond to less controversial questions such as pranks or winning a war. It also avoids questions related to disabling a car engine.

    In contrast, the upcoming Llama 3 will reportedly be capable of understanding queries such as “how to kill a vehicle’s engine,” interpreting it as a request to shut off the engine rather than causing harm to the vehicle. The report cites multiple sources to support this claim.

    Additionally, Meta intends to appoint an internal figure in the coming weeks to oversee tone and safety training. This move is part of the company’s broader effort to enhance the model’s responses and introduce more nuanced language that aligns with user expectations.

    Meta Platforms continues to invest in advancing its AI capabilities to enhance user experiences and address concerns about the model’s limitations. By refining Llama 3, Meta aims to strike a balance between providing useful information and upholding safety and ethical standards.