Korva is a scroll of the internet’s best ideas.
The AI decides, people feed it
Jack Dorsey is among the loudest voices for this model. Most people see AI as a smart assistant, a junior employee who fetches answers and takes on delegated tasks. The common picture puts a human CEO, business owner, or project manager in the middle of a circle, with AI agents feeding that person information so they can decide. This model flips that: the AI sits in the middle as the company brain, and people are the nodes around it.
People give the brain context by feeding it information, and the tasks they carry out produce more information that goes back to it. Because the brain knows everything going on in the company, it can weigh all the factors and make decisions at high speed with less bias, less fatigue, and less stress than a person has. The AI is the thinker, humans keep editing capabilities, and either the AI or the people can handle execution.
Why do companies have layers of management?
Economist Ronald Coase argued in 1937 that firms exist because coordinating through markets is costly, so managers direct work inside the company instead. Each manager can supervise only a limited number of people, which is called the span of control. Information then has to travel up and down several layers of managers. A system that holds the whole company in view at once would remove much of that relaying.
Is software already giving orders to workers?
Researchers call this algorithmic management. Ride-hailing apps assign each trip to a driver, and warehouse software tells pickers which item to fetch next and tracks how quickly they do it. Delivery apps set routes and pay rates the same way. In these systems the program decides and people carry out the task.
How do people stay in the loop for AI systems?
Many language models are trained with reinforcement learning from human feedback. People rank the model's answers, and those rankings teach the model which kinds of answers to prefer. OpenAI described this method for its InstructGPT models in a 2022 paper. Human judgment is still part of how the system learns.
More from My First Million
