Читаем The Three-Body Problem полностью

At the time, I didn’t. Yes, I know that someone studying math should know about a master like Poincaré, but I didn’t worship masters and I didn’t want to become one, so I didn’t know his work. But even if I had, I would have continued to pursue the three-body problem.

Everyone seems to believe that Poincaré proved that the three-body problem couldn’t be solved, but I think they’re mistaken. He only proved sensitive dependence on initial conditions, and that the three-body system couldn’t be solved by integrals. But sensitivity is not the same as being completely indeterminable. It’s just that the solution contains a greater number of different forms. What’s needed is a new algorithm.

Back then, I thought of one thing: Have you heard of the Monte Carlo method? Ah, it’s a computer algorithm often used for calculating the area of irregular shapes. Specifically, the software puts the figure of interest in a figure of known area, such as a circle, and randomly strikes it with many tiny balls, never targeting the same spot twice. After a large number of balls, the proportion of balls that fall within the irregular shape compared to the total number of balls used to hit the circle will yield the area of the shape. Of course, the smaller the balls used, the more accurate the result.

Although the method is simple, it shows how, mathematically, random brute force can overcome precise logic. It’s a numerical approach that uses quantity to derive quality. This is my strategy for solving the three-body problem. I study the system moment by moment. At each moment, the spheres’ motion vectors can combine in infinite ways. I treat each combination like a life form. The key is to set up some rules: which combinations of motion vectors are “healthy” and “beneficial,” and which combinations are “detrimental” and “harmful.” The former receive a survival advantage while the latter are disfavored. The computation proceeds by eliminating the disadvantaged and preserving the advantaged. The final combination that survives is the correct prediction for the system’s next configuration, the next moment in time.

“It’s an evolutionary algorithm,” Wang said.

“It’s a good thing I invited you along.” Shi Qiang nodded at Wang.

Yes. Only much later did I learn that term. The distinguishing feature of this algorithm is that it requires ultralarge amounts of computing power. For the three-body problem, the computers we have now aren’t enough.