Content：
Neural networks generate theorems. Under the cumulative effects of unhealthy lifestyles, cancer cells progress from 1 to 2, from 2 to hundreds, from hundreds to thousands. As cancer develops from early to mid and late stages, it forms a structure similar to a neural network, gaining "intelligence." This allows it to compete with and even overpower the immune system and immune cells, ultimately leading to death. However, a single cancer cell can be easily eliminated by the immune system.
When an earthworm is cut into two segments, both segments can survive and become two separate earthworms. The principle is similar: a neural network exists throughout the body, not just in the brain, making it difficult to eradicate.
In contemporary conflicts like the Israel-Hamas war, decapitation strikes targeting leadership are common. However, adversarial groups have since adopted collective and decentralized leadership structures to reduce the risk of such strikes.
During the Chinese Revolutionary Civil War and the Anti-Japanese War, especially before moving to Xibaipo, the Central Committee of the Communist Party of China divided into two leadership centers: one in the rear and one at the front led by Zhu De. This approach proved effective.
The essence of the world is contingency and uncertainty. As Yang Zhenning described Edward Teller (the "father of the hydrogen bomb"): Teller had ten ideas a day, but only half of one was useful – which was considered remarkable. AI also operates through trial and error. Since both involve trial and error, forming neural networks for this process could be more efficient and effective.
AI becomes a scientist
If the contingency of the world is incorporated into a neural network within a supercomputer 200 million years from now, it could better simulate the laws of the world. This year's Nobel Prize in Biology for protein structure prediction and design follows a similar logic. However, this only approximates truth or creates pseudo-truth. I consistently believe that true truth lies in contingency and uncertainty.
The Yang-Mills equations contain an element of conjecture. I am curious whether the unification of quantum mechanics and general relativity could also be achieved through trial and error, conjecture, or prediction by neural networks – essentially, AI conducting scientific research and invention.
Download the full text PDF for viewing and using it according to the license of this paper.
