Meta Releases New AI-based Big Language Model
2023.02.25 01:35
Meta Releases New AI-based Big Language Model
By Tiffany Smith
Budrigannews.com – (NASDAQ) Meta Platforms Inc. said on Friday that it would be making available to researchers a new large language model, which is the core software of a new artificial intelligence system. This is escalating the AI arms race as Big Tech companies rush to incorporate the technology into their products and impress investors.
The launch of OpenAI’s ChatGPT, backed by Microsoft, marked the beginning of the public competition for dominance in the AI technology market, which prompted tech titans from Alphabet (NASDAQ:) Inc. to Baidu Inc. in China (NASDAQ:) to promote their own products.
According to a blog post, Meta’s LLaMA, which stands for Large Language Model Meta AI, will be made available to researchers and organizations associated with academia, government, and civil society under a non-commercial license.
To generate content and summarize information, large language models mine vast amounts of text. They can, for instance, respond to questions with sentences that appear to have been written by humans.
The model is trained on 20 languages, focusing on those with the Latin and Cyrillic alphabets, and Meta claimed that it requires “far less” computing power than previous offerings.
Gil Luria, senior software analyst at D.A. Davidson, stated, “Meta’s announcement today appears to be a step in testing their generative AI capabilities so they can implement them into their products in the future.”
“Generative AI is a new AI application that Meta does not have much experience with, but it is clearly important for their business’s future.”
Despite widespread layoffs and a reduction in experimental bets due to the tech industry’s slowing growth, AI has emerged as a bright spot for investments.
According to Meta, LLaMA may perform better than rivals that take into account more parameters, or variables.
More specifically, it claimed that a version of LLaMA with 13 billion parameters could perform better than GPT-3, a model that came before ChatGPT.
It claimed that its 65-billion-parameter LLaMA model was “competitive” with Google’s Chinchilla70B and PaLM-540B models, which are even larger than the model Google used to demonstrate its Bard chat-powered search.
A spokesperson for Meta credited the performance to a greater quantity of “cleaner” data in addition to “architectural improvements” in the model that improved training stability.
A new version of Meta’s chatbot, BlenderBot, was based on the large language model OPT-175B, which was made available to researchers in May of last year.
Later, it introduced a model called Galactica that could write scientific articles and solve math problems. However, the demo was quickly taken down after it generated false responses that sounded authoritative.
