Meta AI has announced the release of Llama 2, the latest update to its large-language artificial intelligence (AI) model, the July 18.
The company said in an additional Twitter comment:
“We believe that an open approach with wide accessibility is the right one for the development of today’s Al models…. Today we are releasing Llama 2… freely available for research and commercial use.
The new version surpasses Llama 1 in many ways. Llama 2 is trained on 40% more data than its predecessor, has longer context lengths, uses models with up to 70 billion parameters, and offers pre-trained models trained on 2 trillion tokens.
The use of public data is another axis of the announcement. Llama 2 was pre-trained on publicly available online data. Its fine-tuned model, Llama-2-chat, uses publicly available instruction datasets and over a million human annotations.
Meta’s announcement provides data showing that Llama 2 “outperforms” competing open-source language models when tested against multiple benchmarks, including coding, reasoning, skill and knowledge tests. In the introduction to an attached document, Meta claims that other publicly released language models cannot fully compete with closed models such as OpenAI’s GPT, Google’s Bard, and Anthropic’s Claude at present.
Meta also released a statement declaring its intent to take a responsible approach to AI. There he says “visibility, scrutiny and trust” and an open approach will be beneficial. The statement was signed by dozens of industry members.
Llama 2 is part of a bigger picture
The announcement represents the final step towards the publication of Llama. Meta released Llama under a non-commercial license in February. Shortly after this limited release, the source code for Llama was released to the public in its entirety.
The latest version, however, allows code from the project to be used in commercial applications, which means the AI model could soon be used in a variety of products.
Meta names several major companies as partners and supporters, including Microsoft, AWS, IBM, and Nvidia. Crypto-friendly venture capital firm Andreessen Horowitz (a16z) is also among those supporters. These companies and others will build with Llama, provide access through cloud services and perform research, according to Meta.
Meta previously announced plans to launch consumer-facing AI apps that are expected to be integrated into its Facebook platform and other products. It is also launching AI-powered speech generation tools and advertising suites.
Although Meta’s AI products are not based on cryptocurrency or blockchain, more general anticipation around AI has produced a robust blockchain AI sector. AI-related cryptocurrency tokens currently have a market capitalization of at least $2.67 billion.