Is there another physics to describe the universe?

The following article comes from the WeChat public account: principle, author takeko

The famous science fiction novelist Ted Chiang conceives a story about the contact of human and alien "seven limbs" in "The Story of your Life".

Faced with the same basic laws of the universe, humans and seven -limb barrels have a different understanding, which produces different "physical" to interpret the same natural phenomenon. This leads to the essential difference between the two civilized language and even the way of thinking.

Stills of "Advent", the film is adapted from "The Story of Your Life". (Photo / Paramram Studios)

This is a very disruptive imagination, but it also brings an interesting question: if one day, we encounter a sufficiently wise alien race, will they find the same physical law as us, or they may be different from different from our different different ways. How to describe the universe?

This question can also be changed, that is, in the face of a system, whether there is only one physical method to describe it, or can we find other feasible and even better description methods?

This is exactly the problem of a group of researchers using a new artificial intelligence (AI). This program can observe physical phenomena through a camera, and then try to find the minimum basic variable set that can fully describe the dynamic dynamic observed. The paper has been published in "Nature · Computing Science".

Determine related variables

Energy (E), Quality (M), speed (C), these three variables constitute Einstein's most representative equation E = MC². But how did Einstein know these concepts at first?

One of the pilot steps to understand physics is to determine the relevant variables. Without the concept of energy, quality, and speed, even Einstein cannot find the theory of relativity.

For thousands of years, people know that objects move quickly or slowly, but only when the concept of speed and acceleration is formally quantified can Newton find his famous exercise law F = MA. Similarly, variables describing temperature and pressure need to be determined before the thermodynamic law is officially established.

This is the case in every corner of the scientific world, and variables are the prelude to theory. But can these variables "automatically" be found? If you can get it, it is expected to accelerate scientific discovery.

In new studies, researchers began to provide the original video clip to the system that they already knew the answer. For example, they have entered a swinged dual videos, and it is known that it has 4 physical state variables, that is, the angle and angle speed of each of the two arms.

After several hours of analysis, AI output its answer: 4.7.

Researchers believe that for AI, this answer is close enough, especially this is the answer that AI can read without any physics or geometric knowledge.

But the number of variables is not enough, and what makes scientists curious is what these variables are said. They then started to visualize the actual variables determined by the program.

But the extraction variable itself is not easy, because computer programs cannot describe them in an intuitive way of human beings. After some investigation, they found that the two variables of the program selection seem to correspond to the angle of the arm, but the choice of the other two variables is still a mystery.

Researchers try to connect other variables with anything they can think of, such as angle speed and line speed, kinetic energy and potential energy, and various well -known combinations, but there seems to be no perfect match.

They are very confident that AI has indeed found a set of effective four variables because it makes good predictions, but they do not understand the mathematical language it speaks.

Explore the unknown complex system

After verifying some other known physical systems, the researchers provided it with other videos, including the physical systems they did not know the clear answer.

For example, one of the videos is a "inflatable dancer" "dancing" in front of a second -hand car market. After several hours of analysis, AI returned 8 variables. A video of a lava lamp made the program judge 8 variables. Subsequently, they entered a holiday fire furnace flame video clip, and the program returned 24 variables.

Part of the video fragments provided to AI in the experiment, including an inflatable dancer, a video fragment of a lava lamp and fireplace flame. (Picture / Columbia University)

This brings an extra interesting question: Is the variable set of each system be unique, or does it have different variable sets every time it is restarted?

In the experiment, the number of variables obtained at each time the AI ​​was restarted, but each specific variable was different. It can be said that there are indeed different methods to describe the universe.

And it is likely that some phenomena may look complicated and puzzling because we are trying to understand and describe them with the wrong variable set. Many times, our choices may not be perfect.

Help understand complex phenomena

In the past, software systems can be refined from the experimental data to the law of free forms, but the premise is to determine the variables in advance. Researchers believe that this new AI can help scientists discover complex phenomena. For these phenomena, theoretical understanding may not keep up with the pace of data torrents, such as the field of biology and cosmology.

Although the team uses video data in this study, theoretically, any type of array data source can be used, including radical arrays or DNA arrays.This may accelerate the footsteps of scientific discovery in the future.#Groupe OD creation:

Written article: takeko

Capture: Wenwen

#Reference source:

https://www.engineering.columbia.edu/news/lipson-chen- ai-alternative-Physics

#Image Source:

Cover map & First Figure: Boyuan Chen, Columbia Engineering

This article is authorized to reprint from WeChat public account: Principle (ID: Principia1687) Author: takeko

Reprinted content only represents the author's point of view

---

It does not represent the high energy office of the Chinese Academy of Sciences

Edit: Yuan Xinyang

---

- END -