Artificial intelligence discovered alternative laws of physics

This exploratory work could one day lead to revolutionary developments in many disciplines.

Modern science is largely based on the principle of repetition. We start from some simple and verifiable assertions to build more complex theories that, once validated, will be used to create new models – and so on.

This approach has proven solid, and today we owe an enormous amount of progress to it, which has indisputably led to the progress of our civilization … But that does not necessarily mean that it was the only possible path. If the circumstances were different, our scientific method might have developed in a completely different way.

This is a question that most science fiction fans have already thought about; For example, many, many observers have wondered how extraterrestrial species might model what we call physics or mathematics.

Until very recently, all this reasoning was more about thought experiment; But the game is starting to change with the explosion of artificial intelligence. This technology is incredibly powerful when it comes to matching different elements that can be quite numerous and above all completely abstract. And that’s why AI works wonders in areas like computer vision.

Reinventing physics from scratch

So researchers at Columbia University decided to conduct a very original experiment: they AI asked to rediscover the laws of physics on its own that govern the behavior of matter. But above all, she had to do it only through concrete examples. She did not have access to There is no theoretical basis Like Newton’s theories, or any information about geometry.

© Dan Cristian Padureț – Unsplash

Their work is based on a camera that monitors the evolution of a physical system, like a pendulum. This is the only resource at his disposal. From these simplified visual examples, AI is responsible for determining the number of parameters needed to describe the behavior of the system in question. In a very exhilarating way, it is a bit like a genius scientist rediscovering physics in real time in a parallel dimension.

Take the well-known example of a double pendulum – a pendulum hanging from the end of another pendulum. It is necessary to describe it within the framework of physics as formulated by Newton four parameters – We’re talking about case variables – that is, the angle and the angular velocity of both arms.

So the researchers were curious to see if the AI ​​would also find four criteria, which could indicate that it would have followed the same logic as humans. But the proposed answer was very surprising: to describe the double pendulum, the system estimated that it would be necessary … 4.7 settings.

Artificial intelligence has its reasons that the mind ignores

At this point, the problem is exacerbated. Because the “thinking” process of these neural networks is inherently difficult for humans to decipher; One can understand the meaning of the proposed result, but it is oftenIt is impossible to accurately identify the algorithmic tricks that allowed the system to reach this conclusion.

So the researchers haven’t been able to figure out what corresponds to that number, to say the least. How can a number of parameters be anything other than an integer? What can this 0.7 mean in practice? For humans, does it make sense to think using fractional parameters?

In an effort to answer these questions, the researchers launched a set of new computer simulations. Objective: to compare these virtual parameters with those in real life. They were able to determine that two of the parameters proposed by the AI ​​correspond more or less to the angle of the arms … but for the others, they have no idea. And not because of my lack of sight.

We tried to associate the other variables with everything and absolutely anything Boyuan Chen, lead author of the study, explains. “ Angular and linear velocities, kinetic and potential energies, and various combinations of other known parameters… quotes. ” But nothing quite compares to it ‘, sorry. “We don’t yet understand the mathematical language spoken by AI.”summarizes.

This is where the problem becomes fascinating. Because even if the researchers did not understand the path of their algorithm, they were able to predict the behavior of the studied systems with great accuracy. conclusion: Whatever the reason behind it, it works fine. An alternative physics model built by artificial intelligence is With the same efficiency we have, Even if that is incomprehensible.

A real creator of “Eureka Moments”?

So the researchers repeated the experiment with other already well-documented mechanical systems. And each time, the result was the same: the algorithm consistently succeeded in predicting the evolution of a mechanical system based on entirely new variables that do not correspond to any parameter in Newtonian physics.

Without any prior knowledge of the physical mechanisms involved, our algorithm discovered the intrinsic dimensions of the observed dynamics and specific combinations of state variables. ‘, explained the researchers. In short, this AI doesn’t just think outside the box; it imagines new ways to get around.

This exploratory work may seem as futile as it is fiction, but its implications can actually be very profound. They reinforce the idea thatThere are likely many other ways to describe visible reality. Some of these methods may be more effective than those we know today.

So the challenge will be to hopefully explore these new approachesDetermine those that can be exploited by humans. This could generate major conceptual revolutions in already highly advanced disciplines, where the slightest advance would require enormous efforts of imagination and experimentation from humans.

© FLY: D – Unsplash

Tangible potential in certain areas

In all honesty, there is little chance that humanity will end up turning into a ” new physics » Formalization by AI; Blowing up the current foundations of science is likely to be counterproductive, at least in the short term. On the other hand, this approach can work wonders in some disciplines that work on rather obscure phenomena.

The most obvious example is surely Quantitative Statistics. Everyone agrees that this technology has huge potential, but it is still progressing rather slowly; Some of the underlying mechanisms are still poorly understood, often forcing researchers to feel their way, very empirically.

In a context of this kind, one can fully imagine that AI could provide very interesting leads which would allow humans toapproach these problems in a fundamentally different way – Enough to pave the way for revolutionary progress.

By starting from scratch each time, it will be possible to reinvent certain concepts from radically different and perhaps more relevant bases. In the case of this study, the parameters formulated by the AI ​​were related to the motion of physical systems, but the concept as a whole goes beyond this field.

This approach can also be used in More specific areas Like the Logistics, urban planning, climatology or public health, for example. These are the activities that AI has already caused a major disruption. But until then, only complementary elements made it possible to improve concepts imagined by humans.

On the other hand, a system of this kind can make it possible to highlight phenomena and approaches that would Researchers have so far eluded them… including the work of the artificial intelligence systems themselves!

Certainly, between AI systems that have already revolutionized scientific research, and those that write scientific articles about themselves and work of this nature, there is reason to be excited about the future of AI in research.

Technical documentation related to this work is available here.

Leave a Comment