Meta aims to increase security and transparency by making LlaMA open source.
Meta, the social networking company, has announced that it will open source its big language model LlaMA 2 and make it freely available for commercial and research purposes. The move is intended to compete with Meta’s competitor OpenAI’s GPT-4, which is freely available. GPT-4 provides the technology that powers programs like ChatGPT and Microsoft Bing.
Meta made the announcement as part of Microsoft’s Inspire event, promoting the two tech titans’ relationship. The release stressed support for Azure and Windows platforms in particular, and Microsoft provided further information about the AI capabilities integrated into the 360 platform, as well as their pricing. Qualcomm participated in this initiative as well, announcing that it is collaborating with Meta to bring LlaMA to computers, phones, and headphones.
According to Meta’s news release, making LLaMA 2 open source allows enterprises, startups, and researchers to gain access to and use more AI tools. LLaMA 2 includes information from “publicly available online data sources” and is trained on 40% more data than LLaMA 1. It is also believed to outperform Falcon and MPT LLMs.
By making LlaMA open source, Meta hopes to improve security and transparency. The company also revealed that the LlaMA 2 model has undergone internal and external security testing, as well as specifics on the model’s review and enhancement processes. LlaMA 2, an open source project, will be made available through Microsoft’s Azure platform. According to Meta, LlaMA will also be available through AWS, Hugging Face, and other providers.
“Open access to today’s AI models will enable developers and researchers to quickly identify and solve problems as a community,” Meta added, highlighting the need of developing AI models in creative industries. Furthermore, the business noted that the LlaMA 2 is in high demand and believes that this model would appeal to a broad audience.