Meta did a “terrible” job of creating a modern artificial intelligence division that will aim to “reinvigorate” a happier internal culture through better communication, career development and even snacks, top employees said Monday in an internal post seen by WIRED.
The comments by Andrew Bosworth, Meta’s chief technology officer, follow last week’s WIRED reporting that revealed widespread dissatisfaction within its applied artificial intelligence engineering division. In March, Meta formed a division of about 6,500 engineers and product managers to work on projects aimed at improving the company’s generative artificial intelligence models. But what workers described as the servile nature of the work led one to describe it as a “gulag.”
Bosworth cited recent employee feedback that influenced the changes he announced. “We have undermined your trust that your expertise and contributions will be valued, that you will grow and advance in your career, and that this will be a place where you can truly make an impact,” Bosworth wrote. “We shook up the management structure that had provided stability while rapid changes in strategy, including a boom-and-bust cycle in hiring, left entire teams in the lurch.”
Meta declined to comment for this story.
The unrest within the AI team is part of a broader decline in morale at Meta in the wake of mass layoffs, employee surveillance and other employee concerns. In recent days, several executives, including CEO Mark Zuckerberg, have posted internal messages acknowledging employees’ feelings and promising to make changes to address the issues.
In a lengthy memo, Bosworth, long considered a Zuckerberg loyalist, said employees would receive more personalized attention in the future. Meta plans to limit the number of managers to about 20 direct reports each, he wrote, and will try to limit the number of times employees are transferred to modern managers as part of the restructuring. Managers would focus primarily on management and secondarily on independent work, and employees would have access to “AI coaching” tools if they chose to apply them.
In response to a comment on his memo about the Applied AI team, Bosworth blamed himself and fellow executives for losing sight of the employee perspective in the rush to focus on broader strategic issues, such as how to better compete in the market for AI coding tools. “We have undoubtedly done a terrible job of explaining this vision, giving people a clear picture of how we will support them and their careers in this transition, and painting a picture of how it will change over time,” he wrote.
However, Bosworth also suggested that bringing people to the AI team in the name of speed is the way to go, and reminded employees that they may have to work on a project that “will not be personally fulfilling for them for some time” because “there will be times when the work will require sacrifices.”
In a separate post last Friday obtained by WIRED, Maher Saba, the vice president leading the Applied AI team, told employees who were forced to join that they would now be able to take on other roles at Meta if they were able to secure them. “We felt it was necessary to leverage what Meta had and others [AI] laboratories don’t do that: our scale and the knowledge of our employees,” Saba said of the decision to appoint people to his team. “But moving forward, we are returning to normal operations and giving people the freedom to apply for positions that interest them,” Saba wrote.