Meta, the company behind Facebook and Instagram, is making a significant entry into the AI landscape with its latest AI offensive. Under the leadership of Mark Zuckerberg, Meta recently launched a ChatGPT-like service and unveiled its latest language models from the LLaMA series.
Unlike other AI firms, Meta is pursuing a unique strategy by releasing its top-tier LLaMA models as largely usable software. This open-source initiative allows users, associations, and groups to adapt and modify the models, leading to a variety of projects such as Alpaca, LeoLM, Hermes, and Vicuna. With the release of the third version of LLaMA, Meta has also introduced its own ChatGPT competitor.
The Meta AI Assistant, introduced last September, is now being integrated into all major Meta products. From Instagram and Facebook to WhatsApp and the Facebook Messenger, the chatbot is set to be accessible. Additionally, Meta has launched Meta.ai, a dedicated website that leverages search results from Google and Bing, coupled with its own image generator. While the service is not yet available in the EU, Meta plans to gradually roll it out in various countries throughout the year.
Mark Zuckerberg emphasizes the potential of the Meta AI Assistant, aiming to make it a key product. The new LLaMA 3 model demonstrates superiority in individual tests, particularly in areas such as mathematics and code generation. The release of two smaller versions of the LLaMA-3 model for the open-source community underscores Meta’s commitment to openness and collaboration.
Initial users praise the models for their remarkable expertise and ability to tackle complex tasks. Furthermore, LLaMA 3 shows significantly fewer errors than its predecessors and rarely provides inappropriate responses. With Meta, a new player in the AI market emerges, winning over audiences with its openness and accessibility.
