I was trying to organise the things I have discussed so far in a coherent format, but I don’t know how to start. It’s more natural to develop the theoretical aspects and then design some games based on that, but things are intertwined and confusing.
If platform users are compared to particles, then the first step is structure formation. Then structure formation in the universe should be mimicked. However, it cannot be imitated precisely because it’s not known exactly how the universe functions.
So things we already know should be imitated. This is a simple game based on this idea: particles (users) move around randomly. The parameters linked to every particle will be translated into money. Since particles move around randomly the amount of money they have goes up and down randomly. However, because the motions of all particles are pseudo-random, the sum of all the money in circulation is zero (some users have a negative sum of money, some have positive money and the overall sum is zero). Users should try to decode patterns in patches of particles, and use those patterns to form structures with other users and increase their money because they can move around in a predictable way to earn money. The mechanism is still very vague, but the main idea is that by decoding patterns in the motions of other particles, users can move around in a predictable way to earn money. By forming structures they increase the possibility of earning money. Particles collide together and construct new structures. That will be mimicking structure formation in the universe. Decoding patterns is similar to technical analysis in the stock market but the game structure is different.
This idea can be expanded to the wave language. In deep learning, there are several layers of neurons and feathers of something are extracted one by one. When wave language is added, the interaction between different elements will be done via waves (again a very vague idea). In deep learning values and parameters are transmitted from one neuron to another, but here waves are transmitted. Consider this example. Imagine the pixels of an image (of some object) are similar to the particles mentioned before. They can move around and deform the image. The wave deep learning method will let particles move around as long as the symmetry constructing that object is preserved. For example, in the real world if we see an apple we can imagine that the apple is rotated, etc. The brain can retain the symmetry of the apple and can modify that object while preserving the symmetry. The apple will still be apple although it is imagined in a new position. If pixels in the example can move around so that the object in the image is modified while the entirety of the image is retained, and if many many images of that object are fed into the algorithm, then the algorithm will associate a symmetry to that object. This adds a semantic layer to deep learning because computers will perceive symmetries that construct objects. Then symmetries will interact with each other via wave language.
The same idea is valid in language processing and LLM as well and adds a semantic layer to that. However, the idea here is very vague. The idea of construction here is based on evolutionary mechanisms. The perfect structure is defined at inf, and a sequence of structures will be produced that get better and better but since the perfection is defined at inf, this process never ends.