Kodzilla

IN CHESS, A HUMAN PAIRED WITH A MACHINE WAS THE STRONGEST FOR THIRTY YEARS. IN WORK, THAT WINDOW MAY CLOSE MUCH FASTER. That's the view of Sholto Douglas and Nick Marwell, lead researchers on reinforcement learning at Anthropic. They said it in a conversation with Joe Lonsdale. And it's the thought that won't leave me alone. Because we are in a pleasant phase right now. AI helps people work faster, so there is more work, not less. Companies are hiring, not firing. And this pleasant phase hides what is actually happening. When a technology makes every worker more valuable, it feels like it will always be that way. But chess showed something else. At first a human and a machine together beat everyone. Then the human in the pair became an unnecessary link. And nobody announced that moment in advance. That's what worries me most. Because full replacement doesn't arrive with a sign on it. It arrives disguised as a temporary boost. So I'd put the main question differently. Not whether work will remain. But what remains for the human once the pair with the model is no longer needed. I think the answer isn't execution. The machine will take execution. What remains is choice. What is worth doing. Who needs it. Who answers if something goes wrong. Look at chess today. Machines have long been stronger than any grandmaster. And we still watch matches between people. Because we don't care about the best move. We care how a person arrived at it. So maybe value is simply moving. From the result to the meaning that someone alive puts into it. But there is one more question, and it's sharper. The researchers say that if AI really makes the world far more productive, the main problem won't be creating abundance. It will be how to share it. So that profits don't flow only to those who already had capital. Imagine two worlds. In both, AI has made everything cheap. In the first, everyone has access to it. In the second, a few hands hold the key. The technology is identical. The only difference is what we managed to decide while the window was open. I think this is the most important thing. Not learning to work faster with AI. That will be useful, but it isn't enough. We must not miss the moment when we can still influence the rules. Because once the pair of human and machine stops mattering, the rules will already be written. And the only question is by whom.