Meta is refocusing its resources on developing AI products and features such as chatbotsbut about a third of its researchers who co-authored published AI research related to large language models have left the company, according to the WSJ.
Six of the 14 authors listed on the research paper for LLaMA have left or announced they will be departing, according to their LinkedIn profiles and people familiar with the matter. Eight of the 19 co-authors on the paper for OPT have left as well.
A new generative AI group overseen by top executives aims to integrate generative AI models into all of Meta’s products, which could help increase user engagement and improve the metaverse. Internally, Meta is rolling out an app called “Metamate.” This productivity assistant pulls information from internal sources to perform tasks at the request of employees.
The new generative AI group is part of a broader push to shift resources from AI research to generative AI for products. It received more than 2,000 internal applications and “quickly amassed hundreds of people from different teams,” the WSJ reports. Hardware resources are also being shifted from research to the more product-focused AI group.
Is Meta going the Android route?
Meta is stepping up its LLM game with LLaMA v2, which will be commercially deployable this time. Meta may try to play the Android game: make the base technology free for everyone, but commercialize the applications.
In doing so, Meta could take a leadership role in the overall AI ecosystem and gain some of the ground that Google, Microsoft, and OpenAI have recently gained over Meta by being faster to integrate AI technology into products.
The first LLaMA was disruptive for open-source AI development and became the basis for many projects such as Alpaca, Vicuna, OpenLLaMA, and OpenAssistant. According to Zuckerberg, the new model will become a core infrastructure product for Meta, serving many purposes in its products and “hopefully many others”. Meta itself is apparently planning to implement chatbots in WhatsApp and Instagram.
With projects like Blenderbot 3 and Galactica, Meta was at the forefront of generative AI for language tasks but was mostly criticized for the risks associated with the technology, such as false information. This was before everyone was talking about ChatGPT and LLMs became cool.