ChatGPT is one of the most popular (if not the most popular) generative AI chatbots around, and its underlying GPT language model powers many other services. Meta, the company formerly known as Facebook, has now partnered with Microsoft to create an alternative: Llama 2.
Llama was initially released in February 2023 as a large language model, but unlike the GPT model that powers ChatGPT and Bing Chat, it wasn’t intended as a general-purpose chatbot. Instead, it was provided under a noncommercial license and intended for research purposes, such as analyzing datasets, for groups like universities and NGOs. At least one version of the model could also run without a powerful server attached — after it was trained, the LLaMA-13B model could run on a single data center-grade NVIDIA Tesla V100 GPU. However, Llama model was leaked online a week after it was introduced. More recently, it has been the subject of a lawsuit by Sarah Silverman and two other authors, who alleged that Llama and OpenAI’s ChatGPT uses their copyrighted books as training data without their permission.
Meta has now introduced Llama 2, which is avaialble free of charge for research and commercial use, and is also open-source. OpenAI’s GPT-3.5 and GPT-4 models are not free and not open-source. Similar to the original version, it’s designed to be trained on custom datasets, such as research databases or software documentation. Meta is also partnering with Microsoft to make Llama 2 easier to deploy on servers.
Meta said in its announcement, “We’re now ready to open source the next version of Llama 2 and are making it available free of charge for research and commercial use. We’re including model weights and starting code for the pretrained model and conversational fine-tuned versions too.” The documention explains, “Llama 2 pretrained models are trained on 2 trillion tokens, and have double the context length than Llama 1. Its fine-tuned models have been trained on over 1 million human annotations. Llama 2 was pretrained on publicly available online data sources. The fine-tuned model, Llama-2-chat, leverages publicly available instruction datasets and over 1 million human annotations.”
There’s no Llama 2 chatbot right now, but you might end up using an AI application or tool in the near future that uses Llama.