The Wharton professor believed that orchestrating agents would require a lengthy process of human design. Personal agents, like swarms, proved him wrong.
In his latest essay published on October 1, Ethan Mollick returns to a conviction he defended: to take advantage of AI agents, we must carefully design their coordination, in the way we build a company. He now believes that the organization of work is just one more skill that AI can learn, a form of "bitter lesson" applied to management.
The point and the swarm
The "dot" is the personal agent. Mollick says that one agent spotted a project number error in an email he had sent himself, and another noticed the impending expiration of an airline credit before contacting the company to request an extension. The "swarm" is the thousands of agents that OpenAI got working together on the Navier-Stokes problem: 88 hours and 2.7 million messages, with minimal human coordination.
His conclusion goes against the usual fears: because organization is expensive, making this function cheap could facilitate the integration of agents into human structures and, in the end, create more work for people rather than eliminate it.
Read the full article on the original source
Go to the source →