Just take a look at physics books, there are formulas lined up one after another explaining how things spin, move, slide and stop. In fact, these formulas describe things that can be seen, but behind the curtain of each one, there is a set of factors that cannot be immediately understood at a glance! In this regard, it might be ieresting for you to know that a program based on artificial ielligence has recely been developed by researchers at Columbia University, which apparely follows alternative physics at will!
Artificial ielligence describes physical phenomena in its own style!
In general, during the experimes conducted on artificial ielligence, the researchers found that by showing videos of the physical phenomena of the earth to this iellige program, this system does not recreate the curre variables that physicists use to explain such phenomena; Instead, it uses new variables to explain what was displayed.
Of course, this does not mean that our curre physics models are incomplete or that there are better models to explain the world around us; Because, for example, Einstein’s laws or any other laws have been proven to the scieific community with mathematics and extremely strong principles. But at the same time, you should note that these laws are based on previous theories and everything that was written ceuries ago. Even today, despite new technologies in imaging black holes or ideifying dista galaxies, these laws (of course, the world of quaum mechanics is a completely differe story and is not the subject of our curre discussion!) have been and still are.
On the other hand, considering this background, the question arises whether if there is an alternative time course in which the mind deals with issues in a differe way, it is still possible to see the surrounding world (in the same way as before) within the framework of classical mechanics. ) defined? Of course, before providing further explanations, we must emphasize that the artificial ielligence in this research can only look at a few videos of physical phenomena; Therefore, in no way can he prese a new physics to explain the world or eer the scieific community as a better Einstein.
This new program based on artificial ielligence is able to explain few phenomena using new variables; So it can in no way be expected to provide a new physics to explain the universe or to coribute to the scieific community as a smarter Einstein.
Can artificial ielligence iroduce new physics to the world of science?
Following the previous discussion, you may also be wondering if we meet iellige space aliens in the future, do they understand the laws of physics in the same way that we do, or may they describe the universe in a differe way? be This story is just like what artificial ielligence preses us in this research! In fact, in the experimes of this research group, the number of variables describing a phenomenon was the same every time the artificial ielligence was restarted, but the specific variables were differe each time; Therefore, we can conclude that there are alternative methods to describe the world and it is possible that our methods are not the most complete and the best ways to describe the world!
There are also other areas for further exploration in this research, for example, can AI really discover new variables? In such a case, this mysterious creature can help us to explain new and complex phenomena that we currely do not have the necessary theoretical understanding about. For example, every day we are faced with new data from the Large Hadron Collider that may poi to new physics! In such a case, there may be other rules that we have missed simply because of the lack of access to the necessary variables.
Maybe in the future, artificial ielligence can remind us of variables that are far from the eyes in various phenomena and help us create new rules for the world!
Playing with artificial ielligence to explain swinging pendulums!
But you may also wonder how artificial ielligence can provide new physics? In order to answer this question, the Columbia University research team fed the program (artificial ielligence) certain video clips of phenomena they already understood and asked the system just one simple question: What are the minimum fundameal variables needed to explain what is happening?
The first video showed two swinging pendulums, which included four state variables, two angles and two angular velocities; After showing this video to the system, artificial ielligence examined the video and the question for several hours and then gave its answer: 4.7 variables are needed to explain this phenomenon!
This number is close enough to the four variables we know, but it still doesn’t explain what the AI thinks about what those variables are. So the team tried to match the known variables with the ones the AI had chosen. During this matching, it was found that two of these variables are in agreeme with the angles of the pendulums, but the other two variables remained a mystery! However, AI could provide accurate predictions about what will happen to this phenomenon in the future; Therefore, researchers realized that artificial ielligence is probably involved in variables and parameters that are not completely understandable for them!
The AI’s response to the minimum fundameal variables to describe two swinging pendulums was 4.7 variables, which is close enough to the four variables we know. Also, this iellige system can make accurate predictions about the future of swinging pendulums!
One of the researchers of this project, Boyuan Chen, who is now an assista professor at Duke University, says: “Our team tried to find other variables or anything that we thought was related to this phenomenon, such as angular and linear velocities, to match kinetic and poteial energy and various combinations of quaities we know. But nothing seemed to match the AI variables perfectly! So it seems we still don’t understand the mathematical language of artificial ielligence!”
After the first video, the research team we on to show other cases of artificial ielligence. The first was a video of an arm dancing in the air with wind, which the artificial ielligence gave eight variables in response to its minimal fundameal variables. Also, this iellige system preseed eight variables after watching the lava lamp movie and iroduced 24 variables in response to the clip of flames returning to earth!
Lava lamp and minimal variables of this system
The lava lamp consists of a special mixture of colored wax inside a glass coainer, which is placed on a base with an incandesce bulb; By applying heat from the lamp to the mixture, a temporary decrease in wax density and surface tension is created, resulting in the heated wax rising through the liquid, cooling, and falling to the bottom of the coainer. Then this cycle is repeated as the mixture is heated.
last word
Such a research is truly surprising because the artificial ielligence algorithm ideifies the state variables in that phenomenon without any knowledge of the physics of the phenomena and only by observing the dynamics of the problem. As a result, we can hope that in the future, artificial ielligence can poteially help us understand the basis of new concepts that we are not currely aware of!
You can read the research in full in Nature Computational Science.
Source: science alert
