With all eyes on OpenAI, Meta drags its Responsible AI team to the recycle bin

Meta has disbanded its Responsible AI team and moved staff into other areas of Mark Zuckerberg’s empire to focus on generative AI.

The Facebook and Instagram owner launched its Responsible AI (RAI) team in 2019 in hope this internal group of interdisciplinary experts could help technical teams design and build machine-learning models ethically. 

In 2021, Meta listed five core concerns the team would focus on, including: privacy and security; fairness and inclusion; robustness and safety; transparency and control; and accountability and governance. 

One year later, the group was folded into the WhatsApp parent’s Social Impact unit. Now, the Responsible AI team is being scrapped altogether, as The Information was first to point out. Most of the employees have been reassigned to Meta’s Generative AI arm – the tech that outputs content based on user requests and prompts – with some going to its machine-learning infrastructure unit.

A Meta spokesperson confirmed the decision in a statement: “We continue to prioritize and invest in safe and responsible AI development, and these changes will allow us to better scale to meet our future needs. While the majority of the RAI team will sit in the Gen AI org, it will continue to support relevant cross-Meta efforts on responsible AI development and use.”

The shift raises questions over Meta’s dedication to building technology safely and responsibly, especially since it reportedly cut its Responsible Innovation team in September 2022. That crew was another faction working under Meta’s Social Impact department.

On the one hand, perhaps it’s not a bad idea to embed ethics experts in engineering teams; on the other, one hopes they’ll be effective in policing Meta while scattered across two sections of the corporation.

The reshuffle follows a series of layoffs at Meta; thousands of employees have been axed this year so far. CEO Mark Zuckerberg has since slashed many technical roles in low-priority projects, such as its AI protein-folding team in its Fundamental AI Research department, and most recently its custom silicon unit in its Reality Labs team. 

Zuck warned of the upcoming cuts to fulfill his “Year of Efficiency” vision to refocus his empire’s direction towards generative AI that can build the fabled metaverse – a set of connected 3D virtual-reality worlds in which people are expected to work, play, and spend.

“Our single largest investment is in advancing AI and building it into every one of our products,” Zuckerberg previously said in a statement.

“We have the infrastructure to do this at unprecedented scale and I think the experiences it enables will be amazing. Our leading work building the metaverse and shaping the next generation of computing platforms also remains central to defining the future of social connection.”  

Since then Meta has announced a basket of generative AI products, including ML services for its advertising business, its Llama large language models, AI-generated stickers, and more. Last week, it introduced Emu, a system that creates animated images. 

Meta launched two tools named Emu Edit, that allows users to direct the model using written instructions, and Emu Video, a text-to-video diffusion model. It hinted these could be deployed across its social media apps one day for users to edit photos and videos, or create custom GIFs in messages or posts. ®