Meta Platforms, earlier known as Facebook, recently made a significant announcement. The tech giant revealed the preliminary versions of its language model, Llama 3, and an image generator that modifies images as users type prompts. This ambitious move is an attempt to secure a stronger position in the highly competitive AI market.
Meta’s efforts to develop and integrate AI models into its virtual assistant, Meta AI, is a strategic move to outperform its rivals, including the likes of Google and the French startup, Mistral AI. The updated Meta AI assistant is set to take centre stage within Meta’s suite of applications, including Facebook, Instagram, WhatsApp, and Messenger, as well as on a new standalone website.
“The latest developments position Meta AI as the most intelligent AI assistant freely available to users,” said Mark Zuckerberg, Meta’s CEO.
Upon landing on the standalone website, users are encouraged to explore the capabilities of the assistant. Whether it’s creating a vacation packing list, playing 90s music trivia, helping with homework, or painting images of the New York City skyline, the assistant is programmed to handle a wide array of tasks.
The introduction of generative AI products to its vast user base is a part of Meta’s strategy to challenge OpenAI’s dominance in the field. This bold approach has necessitated a costly revamp of computing infrastructure and the fusion of previously separate research and product teams.
The release of Llama models for use by developers building AI apps has raised safety concerns among critics. They worry about the potential misuse of the model by unscrupulous actors.
Despite these concerns, Meta has made significant advancements with Llama 3. The model has been equipped with new computer coding capabilities and has been trained using images as well as text. Although it currently outputs only text, more advanced features like the ability to craft longer multi-step plans and generate both text and images will be incorporated in future releases.
The aim of developing these models is to simplify and enhance the user experience. Whether it’s interacting with businesses, drafting a document, or planning a trip, Meta wants to ensure that its AI assistant can handle a range of tasks with ease.
“Our goal is to make your life easier,” said Chris Cox, Meta’s Chief Product Officer.
In addition to these advancements, Meta is expanding its AI assistant beyond the US to markets including Australia, Canada, Singapore, Nigeria, and Pakistan. Meta is still exploring the best way to roll out its services in Europe, where privacy rules are more stringent and incoming regulations may impose additional requirements.
The competition between Meta and OpenAI is evident in Mark Zuckerberg’s recent announcement video, where he claims Meta AI to be the most intelligent AI assistant available for free use. It is worth noting that Llama 3 has scored favourably against other free models on performance benchmarks used to assess model quality. The largest version of Llama 3, with 400 billion parameters, is currently being trained.
