In an industry that has recently been dominated by internet competitors Google and Microsoft, Meta offered a research tool for creating chatbots and other products based on artificial intelligence.
Large language models, or LLaMA, are Meta’s newest offering in this field. According to CEO Mark Zuckerberg, these models “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.”
According to a spokeswoman for Meta, the social media sites Facebook and Instagram do not currently use LLaMA. The business intends to give researchers studying AI access to the technology.
According to Zuckerberg, “Meta is committed to this open style of research.”
Massive AI systems known as “large language models” ingest immense amounts of digital text from news stories, social media posts, and other online sources. They then utilize that information to train software that can predict and generate content on its own when given a prompt or query. The models can be used to generate chatbot chats, write articles, compose tweets, and recommend computer program code.
As more businesses have started to develop them and launch tests of goods based on the models, the technology has gained popularity and drawn controversy in recent months, highlighting a new field of rivalry among tech titans. Microsoft has committed billions of dollars to OpenAI, the company that created the GPT-3 big language model that powers the ChatGPT chatbot. This month, the software developer introduced a beta version of its Bing search engine using OpenAI’s conversation technology, which immediately sparked worries about its occasionally improper responses.
LaMDA, or Language Model for Dialogue Applications, is a model created by Google. The industry leader in internet search and advertising is piloting Bard, an AI-powered chat-based search solution, which unfortunately has some bugs.
OPT-175B, a sizable language model that Meta had previously released, has been superseded by the more recent and sophisticated LLaMA. Galactica, another model that Meta introduced late last year, was soon taken back after researchers found that it consistently provided users with biased or false information.
On earnings conference calls and in interviews, Zuckerberg frequently discusses the value of AI in enhancing Meta’s products as he has made it a primary focus within the firm. Although LLaMA is not now utilized in Meta products, it might be in the future. For the time being, Meta uses AI for a variety of tasks, such as content moderation and ranking content that appears in user feeds.
Making the LLaMA model open-source enables outsiders to understand the system’s operation more clearly, modify it to meet their needs, and work together on related initiatives. BLOOM, an open-source Program that was created by Big Science and Hugging Face and aimed to make this kind of technology more accessible, was released last year.
