Microsoft AI research clash: John Langford quits, declares AI Frontiers over

John Langford, the mind behind Microsoft’s AI research group, AI Frontiers, has left the tech company. He declared that AI Frontiers is over. The move comes after the small research group was moved from Microsoft Research into Microsoft AI, a transition that Langford says left little room for the kind of innovation his team wanted to pursue. What began as an effort to make AI research more organised has now according to him ended in a rather uncomfortable split. And Langford's account offers a glimpse into a growing debate in the AI industry: should companies keep building bigger models, or find smarter ways to make them work?

Writing on his blog, Machine Learning (Theory), Langford described the group's journey, its reorganisation and his decision to leave Microsoft. His X post called the situation “a bit of a train wreck” that was “painful to express”, adding that the event was significant for him and several others involved.From research breakthroughs to bigger models

AI Frontiers was formed within Microsoft Research in 2023, following ChatGPT's arrival and the realisation that individual researchers could struggle to compete with better-funded, organised AI teams. With around 50 people, the group worked on projects such as AutoGen, an AI agent orchestration framework, and Fara, a computer-use agent. Some researchers also explored more fundamental ideas, including techniques that could make models more capable without simply increasing their size.

That approach, however, ran into a different philosophy after AI Frontiers moved into Microsoft AI in September. Langford describes the leadership's core philosophy as “scaling, and aggressively not research”. In other words, the focus was on making models bigger, improving training data and refining existing approaches rather than pursuing new research directions.

Langford does acknowledge why companies might prefer scaling. Bigger models often deliver better predictions, while the returns on scaling are more predictable from a business perspective than research breakthroughs. But he believes treating the two as competing priorities is a mistake. “Returns to research are real and significant,” he writes.Why Langford chose to walk away

For Langford, the disagreement was not just about research philosophy. He says he struggled to get innovations into actual use, or even seriously considered, while working in Microsoft AI.

“I am just not comfortable being at a company where the research I do has no path to real use,” he writes. He also questions whether Microsoft Research still offers the traditional promise that good work will be rewarded. Langford believes AI Frontiers had delivered meaningful research but was reorganised against its wishes, raising concerns about the organisation's future.

Microsoft seems to be going through a tough time. Along with layoffs, for which it often comes into the limelight, several senior leaders have left the company this year. The most recent is Ryan Roslansky, who ran LinkedIn as CEO before taking charge of Microsoft’s Office and Teams groups. He announced in October that he will leave Microsoft by the end of 2026. Earlier, in April, Eric Boyd, who led Microsoft’s AI platform, joined the AI startup Anthropic as its head of infrastructure. A few other names include Vishnu Nath, who headed the Office product group and now leads product for Google Chat; Phil Spencer, who retired in February after 38 years at the company; and Sarah Bond, who resigned as Xbox president the same day.- Ends

I’m a journalist who loves reading and exploring diverse realms.