Mass Intelligence

Making capable AI easier to access also means reducing the knowledge needed to select it. Ethan Mollick's Mass Intelligence essay argues that model menus had become a barrier: people could use a chatbot regularly without discovering its stronger reasoning option.

OpenAI's GPT-5 was meant to remove that choice. As Mollick explains, a request to GPT-5 went to a router that was supposed to decide whether a smaller, faster model could handle it or whether it needed a more powerful reasoning model. That moves a technical choice into the product itself. He also reports that the router did not work well at first, so similar users could receive very different answers. The access benefit therefore depended partly on how well that selection worked.

“When you pick ChatGPT 5 you are actually picking Auto mode, which selects among the various ChatGPT 5 models, some of which are among the best models in the world, some of which are much weaker.”

“According to OpenAI, less than 7% of paying customers selected o3 on a regular basis, meaning even power users were missing out on what Reasoners could do.”

“Google has reported that energy efficiency per prompt has improved by 33x in the last year alone.”