Meta has developed several versions of LLaMA that require more or less resources. (Photo: 123RF)
Meta on Friday unveiled its own version of artificial intelligence (AI) capable of text generation, like ChatGPT, opening it up to researchers for now so they can respond to the risks of these emerging technologies.
The aim of this new language model, called LLaMA, is “to help researchers progress their work” on this subject, particularly because it does not require very large infrastructures to be studied, the company specifies in a presentation text.
The launch of start-up OpenAI’s conversational robot ChatGPT in November actually shook up the world of AI by allowing the general public to see the ability of new “language models” to generate text on a given topic or to provide an explanation in a matter of seconds to give to a complex subject.
But they also harbor risks, be they factual errors, bias or privacy.
A test version of Microsoft’s Bing search engine, developed in partnership with OpenAI, was quick to return conflicting answers, with the computer program notably expressing threats or a desire to steal the nuclear codes.
“Additional research is needed to address the risks of bias, toxic comments and hallucinations,” says Meta, the parent company of Facebook and Instagram.
But it requires significant resources, especially in terms of computational power, to train and run these language models.
This “limits researchers’ ability to understand how and why these large language models work, and hampers efforts to improve their robustness and mitigate known issues such as bias, toxicity, and the ability to generate erroneous information,” notes the company firmly.
Because of this, Meta has developed several versions of LLaMA that require more or less resources.
As OpenAI and Microsoft limit access to the technologies that power their AI, Mark Zuckerberg’s company decided to share how it built LLaMA so researchers can “more easily test newer approaches to limit or eliminate the problems.”