Meta has announced the release of a new “state-of-the-art” Artificial Intelligence model dubbed “LLaMA” that is designed to help in carrying out varying tasks, including predicting protein structures.
The CEO of the social media conglomerate, Mark Zuckerberg, said on Friday that LLaMA will be made available to researchers to help with their work in “important, fast-changing field.”
“Today we’re releasing a new state-of-the-art AI large language model called LLaMA designed to help researchers advance their work,” he said, adding that LLaMA “have shown a lot of promise in generating text, having conversations, summarizing written material, and more complicated tasks like solving math theorems or predicting protein structures.”
Tekedia Mini-MBA edition 14 (June 3 – Sept 2, 2024) begins registrations; get massive discounts with early registration here.
Tekedia AI in Business Masterclass opens registrations here.
Join Tekedia Capital Syndicate and invest in Africa’s finest startups here.
The release came in the wake of ChatGPT’s sweeping reign – OpenAI’s chatbot has triggered a global AI model language race.
Meta said LLaMA will assist researchers who don’t have access to large amounts of infrastructure to advance their study in the subfield of AI.
“Even with all the recent advancements in large language models, full research access to them remains limited because of the resources that are required to train and run such large models,” Meta said in a blog post on Friday. “This restricted access has limited researchers’ ability to understand how and why these large language models work, hindering progress on efforts to improve their robustness and mitigate known issues, such as bias, toxicity, and the potential for generating misinformation.”
While ChatGPT has garnered a lot of interest in a short while, with Microsoft incorporating it into its Bing search engine and also investing billions of dollars in OpenAI, the startup still beams concern about trust. The company’s cofounder and CEO Sam Altman said the tool cannot be relied on now as it is still prone to mistakes and misinformation.
LLaMA works by taking a sequence of words, which are easier to retrain and fine-tune for specific potential product use cases, as an input and predicts a next word to recursively generate text. Meta said they chose text from the 20 languages with the most speakers, “to train our model,” focusing on those with Latin and Cyrillic alphabets.
The social media platform said in a blog post that access to the model will be granted on a case-by-case basis to academic researchers; those affiliated with organizations in government, civil society, and academia; and industry research laboratories around the world.
However, it admitted that more research is still needed to be done to address the risks of bias, toxic comments, and hallucinations in large language models.
Both Microsoft and Google – which recently tested Bard – AI chatbot designed in response to OpenAi’s ChatGPT, are witnessing the AI-powered tools making mistakes. Meta said LLaMA is open to researchers who could help make it better by addressing its mistakes.
“By sharing the code for LLaMA, other researchers can more easily test new approaches to limiting or eliminating these problems in large language models. We also provide in the paper a set of evaluations on benchmarks evaluating model biases and toxicity to show the model’s limitations and to support further research in this crucial area,” the company said.
But Meta is not releasing the tool to the public yet. The company said in order to “maintain integrity and prevent misuse,” the AI model will be released “under a noncommercial license focused on research use cases.”