Somehow, people seem opposed to the idea that simple, brute-force scaling can result in something approximating intelligence. But the difference between goofy i

Somehow, people seem opposed to the idea that simple, brute-force scaling can result in something approximating intelligence. But the difference between goofy inchoate text ramblings and effectively writing a coherent essay was just scale. https://www.gwern.net/Scaling-hypothesis# https://www.gwern.net/Scaling-hypothesis# Yet, accuracy on AI tasks looks a lot like site reliability (90% is garbage, 99% is flaky, each 9 after that starts getting you a reputation for reliability).