A new large language model (LLM) called Meta had released LLaMA to assist academics working in artificial intelligence. Meta claims that LLaMA might surpass GPT-3, a current model based on ChatGPT. This is Meta’s third LLM following the shutdown of Galactica and BlenderBot 3 due to inaccurate findings.
CEO Mark Zuckerberg stated that Meta had trained a new huge language model that will be offered to academics. The LLaMA model is designed to assist researchers and engineers in investigating potential uses for AI, such as answering queries and summarizing materials.
The announcement of Meta’s new model, created by its Fundamental AI Research (FAIR) team, comes as well-funded startups and significant tech firms compete to highlight developments in AI methods and incorporate the technology into consumer goods. Applications like OpenAI’s ChatGPT, Microsoft’s Bing AI, and Google’s unannounced Bard are supported by large language models.
According to Zuckerberg’s post, LLM technology may one day be used to perform scientific studies or solve mathematical issues.
“LLMs have shown much potential in creating language, holding conversations, summarizing written material, and more difficult tasks like solving arithmetic theorems or predicting protein structures,” said Facebook CEO Mark Zuckerberg.
Meta states that its LLM differs from rival models in many respects. Initially, it says many sizes will be available, ranging from 7 billion to 65 billion parameters. In recent years, larger models have succeeded in increasing the technology’s potential, but they are more expensive to run, a stage that researchers call “inference.” For instance, Chat-GPT 3 by OpenAI includes 175 billion parameters.
Additionally, Meta stated that it is accepting submissions from researchers and will make its models available to the research community. The underlying models for OpenAI’s ChatGPT and Google’s LaMDA are proprietary.
Meta is dedicated to this open research approach, and Zuckerberg stated that “we will make our new model available to the AI research community.