Mark Zuckerberg announces the big Meta LLaMA language model

Mark Zuckerberg, Co-Founder and CEO of Meta Platforms, in July 2021.

Kevin Deitch | Getty Images News | Getty Images

Meta has trained and will release a new big language model for researchers, CEO Mark Zuckerberg announce Friday.

model, It’s called LLAMAaims to help scientists and engineers explore applications of artificial intelligence such as answering questions and summarizing documents.

Meta’s release of its new model, developed by the Fundamental Artificial Intelligence Research (FAIR) team, comes as big tech companies and well-capitalized startups alike race to promote advances in AI technologies and integrate the technology into commercial products.

Large language paradigms support applications such as OpenAI’s ChatGPT, Microsoft’s Bing AI, and Google’s unreleased Bard.

In his post, Zuckerberg said that LLM technology could eventually solve math problems or conduct scientific research.

“LLM has shown a lot of promise for text generation, conversations, abstraction of written material, and more complex tasks like solving mathematical theorems or predicting protein structures,” Zuckerberg wrote Friday.

Below is an example of the output of the system, from: meta sheet:

This is the model’s rendering of a fictional scene from “Seinfeld”:

Meta says its LLM is distinguished in several ways from competitive models.

First, it says it will come in multiple sizes, from 7 billion parameters to 65 billion parameters. Larger models in recent years have expanded the technology’s capability, but it costs more to run, a phase researchers call “inference.”

OpenAI’s Chat-GPT 3 contains 175 billion parameters, for example.

Meta also said it will make its models available to the research public and will take submissions from researchers. The base models for Google’s LaMDA and OpenAI’s ChatGPT are not public.

“Meta adheres to this open model of research and we will make our new model available to the AI ​​research community,” Zuckerberg wrote.