Song He and Xiang Xiaotong's hands, holding their chopsticks, came to a synchronized halt.
"Take too big a step, and you'll pull something down there." Song He said seriously, "Have you thought it through?"
"What did you decide? If you're not using binary, what base are you planning to use?" Xiang Xiaotong asked.
"In theory, base e is the most efficient. After all, e is 2.71828. In practice, we'll definitely have to use an integer base." Yang Yu said, "ln3/3 is closer to lne/e than ln2/2 is, so the current mainstream view is that we should move toward ternary."
"Ternary is more efficient than binary when processing large datasets. That means the larger the supercomputer, the better ternary should perform in theory. We ran some rough calculations on certain problems, and ternary will very likely be more than twice as fast as binary—far exceeding the industry's prediction of a mere one-and-a-half times faster." Yuan Tiangang said, "One group has already converted binary logical operations into ternary ones. At the same time, ternary allows for more types of logical operations, laying the groundwork for improving the intelligence of brain-mimicking supercomputers."
"Binary computers dominate the field today. We believe that's mainly a historical legacy." Yuan Tiangang said, "There's a joke about path dependence. It says that in ancient times, the width of two horses' rear ends determined the width of roads, which then determined the width of coal wagons, which in turn determined the standard width of railways. Finally, after humanity entered the rocket age, completed rockets still had to be transported by rail, so the width of horses' rear ends in ancient times determined the width of rockets in later generations."
"When computers first emerged, materials technology was still very backward. The hardware we could make was crude, and having components