Meta, the Mark Zuckerberg-owned Facebook parent, has introduced Llama 2, its latest AI language model, to enter the competitive arena alongside OpenAI's ChatGPT and Google's Bard. Llama 2 serves as an upgraded version of its predecessor, Llama, which possessed the capability to generate text and code similar to other chatbot systems. Interestingly, Meta has chosen to drop the capitalised acronym for this second iteration, opting for a simpler name.
Llama 2 comprises a collection of models designed to generate text and code in response to prompts, similar to existing chatbot-like systems. As a large language model (LLM), it outshines its predecessors in terms of power and efficiency, claims Meta. Its training involved an extensive dataset encompassing vast amounts of text and code, enabling it to produce more creative and informative responses than Google's Bard or ChatGPT.
Zuckerberg announced Llama 2 with a Facebook post on Tuesday. "Open source drives innovation because it enables many more developers to build with new technology. It also improves safety and security because when software is open, more people can scrutinise it to identify and fix potential issues. I believe it would unlock more progress if the ecosystem were more open, which is why we're open sourcing Llama 2," he wrote.
In a move that aims to encourage widespread usage, Llama 2 is freely accessible for both research and commercial purposes. Meta's training methodology involves utilising a combination of publicly available data. The company proudly asserts that Llama 2 exhibits significant performance improvements when compared to previous generations of the Llama models.
As a commercial version of its open-source artificial intelligence model, Llama, Llama 2 presents startups and businesses with a cost-effective alternative to the pricy proprietary models offered by OpenAI and Google.
The initial Llama model was already positioned as a formidable contender against OpenAI's ChatGPT and Google's Bard chatbot. However, Llama 2 has received even more comprehensive pre-training, utilising 40 per cent more data than its predecessor. Furthermore, over one million annotations contributed by humans have fine-tuned the quality of its outputs.
To ensure accessibility, Llama 2 can be found in the Azure AI model catalogue, catering to users of the Windows operating system. Additionally, it is available through various providers such as Amazon Web Services (AWS) and Hugging Face, expanding its reach to a broader user base.