Meta, the parent company of Facebook, is releasing a commercial version of its open-source AI model called Llama 2. This move provides businesses and start-ups with a free and powerful alternative to expensive proprietary models offered by companies like OpenAI and Google. Previously available only to select academics, Llama 2 will be distributed through Microsoft’s Azure cloud service and will run on the Windows operating system. It will also be available via direct download and other providers like Amazon Web Services and Hugging Face.
The new Llama model has been trained on 40% more data than its predecessor and has undergone fine-tuning with over 1 million annotations by humans. This makes it competitive with models powering OpenAI’s ChatGPT and Google’s Bard chatbot. Meta believes that making Llama widely available for businesses may disrupt the dominance of closed-source AI models and offer more choice for developers. It also aligns with Meta’s strategy of building a flourishing open-source ecosystem to draw in developers and improve infrastructure costs.
While the release of Llama into the public domain may raise concerns about potential misuse by unscrupulous actors, Meta believes that public releases actually reduce safety risks by harnessing crowd wisdom to identify and fix problems. The company has implemented an “acceptable use” policy for commercial Llama to prevent certain misuse cases, including violence, terrorism, child exploitation, and criminal activities. Overall, Meta’s move to release Llama 2 aims to foster AI innovation and provide businesses with a powerful AI model for free.