Meta releases beefed-up AI models

Meta unveiled its enhanced AI assistant, powered by the latest iterations of its open-source Llama large language model. The upgraded Meta AI boasts greater intelligence and speed, leveraging the advancements of the publicly available Llama 3 model, as Meta highlighted in a recent blog post.

In a video shared on Instagram, Meta’s co-founder and CEO, Mark Zuckerberg, expressed confidence in Meta AI’s newfound capabilities, claiming it to be the most intelligent AI assistant accessible to everyone.

One of the key advantages of using an open-source model like Llama 3 is the flexibility it offers to developers outside of Meta. They can customize the model to suit their needs, and Meta can then incorporate these enhancements into future updates.

Meta emphasized its excitement about the potential of generative AI technology, not just for users of Meta products but also for the broader tech ecosystem. However, the company stressed the importance of responsible development and release practices to mitigate risks associated with AI technologies.

Meta has been steadily improving Meta AI since its initial launch, although its cautious approach may have impacted its consumer reach compared to competitors. Nonetheless, Meta plans to integrate AI features across its suite of apps, leveraging its vast user base for testing and refinement.

One example cited by Meta is how it refines AI responses to sensitive topics, ensuring a balanced and informative approach rather than a biased viewpoint. The company has fine-tuned Llama 3 to better understand the context of prompts and provide appropriate responses.

To enhance transparency, Meta will label content generated by AI as “Made with AI” starting in May, including videos, audio, and images. Additionally, Meta plans to expand Llama 3’s language capabilities beyond English in the near future.