Meta Platforms is making a big push in generative AI with the release of early versions of its new large language model, Llama 3, and an image generator that updates pictures in real-time.
These tools are being integrated into Meta AI, the company’s virtual assistant, which they claim is the most advanced free option available. Meta AI will be prominently featured within Facebook, Instagram, WhatsApp, Messenger, and a new standalone website, putting it in direct competition with OpenAI’s popular ChatGPT.
Catching Up and Looking Ahead
This announcement reflects Meta’s efforts to catch up to OpenAI in the generative AI market. They’ve been working on revamping their computing infrastructure, merging research and product teams, and pushing out these AI products to their vast user base.
Llama 3 is a more powerful language model than its predecessor, trained with both text and images. While the initial version will only generate text, Meta promises future iterations that can handle complex tasks like creating multi-step plans and producing both text and images (“multimodality”).
Meta AI’s Expanding Capabilities
Chris Cox, Meta’s Chief Product Officer, envisions Meta AI as a tool that simplifies users’ lives. It could help with tasks like interacting with businesses, writing content, and trip planning. He also highlighted how Llama 3’s training with images will enhance the upcoming update to Ray-Ban Meta smart glasses. Meta AI will be able to identify objects users see and answer questions about them.
Collaboration and Expansion
Meta announced a new partnership with Google to include real-time search results within Meta AI responses, adding to their existing agreement with Microsoft’s Bing. The assistant is also expanding globally, with availability in over a dozen new markets, including Nigeria and Pakistan. However, European rollout is on hold as Meta navigates stricter privacy regulations and the upcoming AI Act.
Competition and Safety Concerns
The race to develop powerful generative AI models has raised concerns about data usage and potential misuse. Meta’s strategy of releasing models like Llama 3 for free could disrupt competitor plans to monetize their technology but also raises safety questions if used for malicious purposes.
Focus on Quality and User Safety
Meta emphasizes that they’ve addressed issues of context-awareness in Llama 3 by training with “high quality data.” They haven’t revealed specific datasets but claim to have used seven times more data than with Llama 2. Additionally, they’ve incorporated “synthetic” data to improve areas like coding and reasoning.
While Meta maintains they haven’t significantly changed their data sourcing approach, they are clearly focused on improving the quality and safety of their AI models. This push comes after both Meta and Google faced criticism for past AI tools generating inaccurate or harmful content.
The race in generative AI is heating up, and Meta’s latest advancements show their determination to be a major player in this rapidly evolving field.
