Meta said on Friday that it is releasing a new big language model to researchers, the underlying software for a new artificial intelligence system, heating up an AI arms race as big tech companies incorporate the technology into their own products. I am rushing to integrate and impress investors.
The public battle to dominate the AI technology space began late last year with the launch of Microsoft-backed OpenAI’s ChatGPT, and has seen tech heavyweights from Alphabet to China’s Baidu push their offerings. Encouraged to increase.
META’s LLAMA, short for Large Language Model MetaAI, will be available under a non-commercial license to researchers and institutions in government, civil society and academia, it said in a blog. .
Large language models use large amounts of text to summarize information and produce content. They can answer questions, for example, with sentences that read as if they were written by humans.
The model, which Meta said requires “much less” computing power than previous offerings, is trained on 20 languages with a focus on the Latin and Cyrillic alphabets.
“Meta’s announcement today appears to be a step toward testing their creative AI capabilities to implement them in future products,” said Gil Loria, senior software analyst at DA Davidson.
“Generative is a new AIAI application that Meta has little experience with, but is clearly important to the future of their business.”
AI has emerged as a bright spot for investment in the tech industry, whose sluggish growth has led to widespread layoffs and cutbacks in experimental bets.
Meta said LLAMA can outperform competitors that examine more parameters, or variables, that the algorithm takes into account.
In particular, he said that a version of LLaMA with 13 billion parameters could outperform GPT-3, the most recent predecessor of the model on which ChatGPT is built.
He described his 65-billion-parameter LLaMA model as “competing” with Google’s Chinchilla70B and PaLM-540B, which is larger than the model Google used to demonstrate its bard chat-powered search. Is.
A Meta spokesperson attributed the performance to a large amount of “cleaner” data and “architectural improvements” to the model that increased training stability.
Meta released the large-language model OPT-175B in May last year, also aimed at researchers, which formed the basis of a new iteration of its chatbot Blenderbot.
He later introduced a model called Galactica, which could write scientific essays and solve math problems, but quickly scrapped the demo after it produced authentic-sounding wrong answers.
