Mark Zuckerberg just announced a new AI model 'LLaMA,' designed to help researchers make chatbots less 'toxic'
Getty Images
- Meta said its new model can help researchers improve and fix AI tools that promote "misinformation."
- CEO Mark Zuckerberg touted "a lot of promise" behind the technology underlying bots like ChatGPT.
- Microsoft and Google have adopted AI technology to boost their search engines, to mixed early reception.
Meta is releasing its generative AI model called "LLaMA" to drive research into what it called an "important, fast-changing field," in the wake of OpenAI's viral success with ChatGPT.
The social media company said that more research can help solve problems like "bias, toxicity, and the potential for generating misinformation" that generative AI tools can pose, according to Meta's blog post on Friday.
OpenAI's own chief technology officer acknowledged the pitfalls of its crown jewel tool, telling Time in an interview this month that ChatGPT "may make up facts." Meanwhile, Microsoft previously told Insider that its Bing chatbot, powered by OpenAI's technology, "may make mistakes" in the early phase of its rollout this month, which was marked by what some users described as Bing's sometimes strange, inaccurate, and combative responses.
Meta said on Friday that "there is still more research that needs to be done to address the risks of bias, toxic comments, and hallucinations in large language models."
The company's AI model, which stands for "Large Language Model Meta AI," is geared toward researchers, its CEO Mark Zuckerberg said in a Facebook post on Friday.
"LLMs 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," he wrote in the post, using the abbreviation "LLMs" to refer to large language models.
"Meta is committed to this open model of research and we'll make our new model available to the AI research community," he wrote.
The company alluded to the popularity of the generative AI tools without referring to any by name, and said that its model can help open up the means to study and develop such technology, which can require significant computing power to train.
"Smaller, more performant models such as LLaMA enable others in the research community who don't have access to large amounts of infrastructure to study these models, further democratizing access in this important, fast-changing field," Meta wrote in a blog post Friday.
For its part, Google is still testing its own Bard AI bot in order to open it up to users. Tech experts say the adoption of such technology by Big Tech could signal a shift toward chatbots potentially springing up across the internet.