Launch of advanced large-scale language model, Meta official announcement to deepen the AI ​​war

Zuckerberg said on social media that LLaMA developed by Facebook AI Research is the "currently highest level" large-scale language model and its goal is to help researchers Advance their work in artificial intelligence (AI).

"Large Language Models" (LLM) can digest large amounts of text data and infer relationships between words in the text. With the advancement of computing power and the continuous expansion of input data sets and parameter spaces, the capabilities of LLM have also increased accordingly.

Currently, LLM has been proven to efficiently perform a variety of tasks, including text generation, question answering, written material summarization, etc. Zuckerberg said that LLM also has great development prospects in more complex aspects such as automatically proving mathematical theorems and predicting protein structures.

It is worth mentioning that the recent popular ChatGPT is a chatbot built using LLM. ChatGPT is supported by GPT3.5, which is an LLM trained based on the OpenAI 175B parameter basic model. 175B is the content it has learned and accumulated from the training data.

Meta stated in the official website introduction that LLaMA has four basic models: 7B, 13B, 33B and 65B, which are better than GPT3.5’s predecessor GPT3-175B in most benchmark tests, while LLaMA-65B can Compete against the best in the industry, the Chinchilla-70B and PaLM-540B.

The above is the detailed content of Launch of advanced large-scale language model, Meta official announcement to deepen the AI ​​war. For more information, please follow other related articles on the PHP Chinese website!