Microsoft executive warns of AI-generated content “doom loop”

Ryan Roslansky, Executive Vice President at Microsoft overseeing LinkedIn, Office, and Teams, has warned against the rise of low-quality, unoriginal AI-generated content, describing the phenomenon as a “doom loop.” Roslansky shared an experience where he received a document clearly produced by generative AI and attempted to use AI to summarize it, only to find that the summary provided no new insights or perspectives.

He cautioned that relying solely on AI without adding human experience could lead to a sea of uniform, characterless text. This issue is linked to “model collapse,” a technical challenge where AI models trained on AI-generated data suffer from declining quality and reduced diversity.

To combat a culture of “token masking”—where employees use excessive AI tokens to appear productive—Microsoft executive Jay Parik has reportedly instructed staff to avoid unnecessary AI consumption and instead focus on meaningful outcomes for customers and the business. Roslansky emphasized that AI should be used to enhance and develop work through unique human perspectives rather than replacing them.