In a notable move, Meta Platforms, the parent company of Facebook, has collaborated with Microsoft to develop Llama 2, their latest artificial intelligence (AI) language model. Unlike its competitors, including ChatGPT and Google’s Bard, Meta has decided to offer Llama 2 free of charge for both research and commercial use.
Meta has always prioritized the advancement of AI technology and boasts a considerable research team of computer scientists comparable to industry leaders like Google and Microsoft. However, the recent frenzy surrounding the release of ChatGPT and its lucrative “generative AI” capabilities have overshadowed Meta’s efforts in this field.
Meta has set itself apart from other tech giants by adopting a more transparent approach to disclosing the data and code used in building their AI systems. By sharing these resources, Meta believes that external researchers can contribute to identifying and rectifying any biases or toxic elements that may be present in the AI models. This contrasts with the closed-off practices of some of its competitors.
According to Meta CEO Mark Zuckerberg, open sourcing technology is crucial as it drives innovation and facilitates collaboration among developers. Moreover, it enhances safety and security by allowing a larger number of individuals to scrutinize and address any potential issues. As a strong advocate for openness in the tech industry, Zuckerberg stated that unlocking progress requires a more open ecosystem, leading them to open source Llama 2.
Meta has a well-established track record of open-sourcing its AI work, as exemplified by its contribution to the widely acclaimed machine-learning framework PyTorch.
Meta Introduces Llama 2: The Latest Language Model
Meta, formerly known as Facebook, has released its latest language model, Llama 2. This new model was trained using a fresh mix of publicly available data, excluding any data from Meta’s own products or services. Although the specific data sources were not disclosed, Meta made it clear that it excluded websites known for their extensive collection of personal information.
Initially named Large Language Model Meta AI (LLaMA), the first version of the model was introduced in February. Now, with its second version, Meta has dropped the capital letters and simply named it Llama 2.
For accessibility and convenience, Meta allows users to download the new AI models directly or access them through a partnership with Microsoft’s cloud platform Azure. Details of the financial arrangement between Meta and Microsoft have not been disclosed.
While Microsoft is recognized as a “preferred” partner by Meta, the company also plans to make the models available on Amazon Web Services (AWS), a significant cloud rival to Microsoft. Additionally, AI startup Hugging Face and other platforms will feature the Llama 2 model.
It is worth noting that Microsoft has its own investments in AI, including a substantial partnership with OpenAI, the developer of ChatGPT. Neither ChatGPT nor similar offerings from Microsoft or Google are open source.
The announcement of the Microsoft-Meta AI partnership was made at Microsoft’s annual event for business customers, Inspire. Microsoft expressed excitement over Meta’s open approach to democratizing AI and its benefits. It should be noted that Meta is already a customer of Microsoft’s Azure cloud computing platform.
In a separate reveal at Inspire, Microsoft shared that it will be implementing a monthly fee of $30 per user for Microsoft 365 Copilot, its flagship generative AI tool. This fee will be in addition to existing charges for Microsoft services incurred by businesses.
Meta’s introduction of Llama 2 signifies a significant advancement in natural language processing and continues its commitment to expanding the possibilities of AI.