An artificial intelligence has discovered alternative laws of physics

This exploratory work may one day lead to revolutionary advances in many disciplines.

Modern science is largely based on the principle of iteration. We start from certain simple and verifiable propositions to build even more intricate theories, which, when validated again, will be used to establish new models – and so on.

This approach has proven to be solid and today we owe a huge amount of progress to it which has undeniably advanced our civilization… but that doesn’t necessarily mean it was the only possible clue. If circumstances had been different, our scientific method might well have developed in a very different way.

This is a question that most science fiction fans have already pondered; for example, many, many observers have wondered how an extraterrestrial species could have conceived of what we call physics or mathematics.

Until very recently, all this reasoning was more a matter of thought experiment; but the game is starting to change with the explosion of artificial intelligence. This technology is incredibly powerful when it comes to juggling different elements that can be very numerous and above all quite abstract. It is for this reason that AI works wonders in fields like computer vision.

Reinventing physics from the ground up

Researchers at Columbia University therefore decided to conduct a very original experiment: they asked an AI to rediscover the laws of physics on its own that governs the substance’s behavior. But above all, it had to do so only on the basis of concrete examples. She didn’t have access to no theoretical basis like Newton’s theorems or any information about geometry.

© Dan Cristian Pădureț – Unsplash

Their work is based on a camera that observes the evolution of a physical system, such as a pendulum. And this is the only resource at his disposal. From these simplified visual examples, the AI ​​is responsible for determining the number of parameters needed to describe the behavior of the system in question. In a very colorful way, it’s a bit like a brilliant scientist rediscovering physics in real time in a parallel dimension.

Take the well-known example of the double pendulum – a pendulum hanging from the end of another pendulum. Describing it within the framework of physics as it was formalized by Newton is necessary four parameters — we’re talking about state variables — namely the angle and angular velocity of each of the two arms.

The researchers were therefore curious to see if the AI ​​would also find four parameters that could possibly indicate that it would have followed the same reasoning 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.

AI has its reasons that reason ignores

At this point, the problem thickens. Because the “reasoning process” in these neural networks is by nature very difficult for humans to decipher; one can understand the meaning of the proposed result, but it is often i.aIt is impossible to determine exactly what algorithmic tricks allowed the system to reach this conclusion.

The researchers were therefore quite unable to know what this, to say the least curious, corresponded to. How on earth can an array of parameters be anything other than an integer? What can this 0.7 mean in practice? Does it make sense for humans to reason with fractional parameters?

In an attempt to answer these questions, researchers have launched a slew of new computer simulations. The goal: to compare these virtual parameters with those in real life. They were able to determine that two of the parameters suggested by the AI ​​corresponded more or less to the arm angle … but for the others they have no idea. And it’s not because of a lack of looks.

We tried to correlate the other variables with absolutely anything and everything “, explains Boyuan Chen, lead author of the study. “ Angular and linear velocities, kinetic and potential energies, various combinations of other known parameters… he quotes. ” But nothing was a perfect fit «, he laments. “We don’t yet understand the mathematical language that AI speaks”he sums up.

And this is where the problem becomes fascinating. Because even if the researchers do not understand the way to their algorithm, they still managed to predict the behavior of the studied systems with great precision. Conclusion: whatever the reason behind it, it works fine. The alternative physics model built by AI is as efficient as ours, even if it is incomprehensible.

A true generator of “Eureka moments”?

The researchers therefore 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 the mechanical system based on completely new variables that did 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 identified sets of state variables “, the researchers explain. In short, this AI doesn’t just think outside the box; she even imagines new ways of getting around.

This highly exploratory work may seem as pointless as it is anecdotal, but its implications could actually be extremely profound. They reinforce the ideathere are potentially many other ways of describing observable reality. And some of these approaches could be even more effective than the ones we know today.

The challenge will therefore be to explore these new approaches in the hope ofidentify those that would be exploitable by humans. This can generate major conceptual revolutions in already very advanced disciplines where the slightest progress requires enormous efforts of human imagination and experimentation.

© FLY:D – Unsplash

Concrete potential in certain areas

Frankly, there is little chance that humanity will end up converting to a ” new physics » formalized by an AI; blasting the current foundation of science would likely be counterproductive, at least in the short term. On the other hand, this approach could work wonders in certain disciplines that work on rather obscure phenomena.

The most obvious example is surely that quantum computing. Everyone agrees that this technology has enormous potential, but it is still progressing rather slowly; some of the underlying mechanisms remain poorly understood, often forcing researchers to probe, very empirically.

In a context of this kind, one can well imagine that an AI could offer very interesting leads, which would then give people the opportunity toattack these problems in a radically different way — enough to pave the way for revolutionary progress.

By starting from scratch each time, it would be possible to reinvent certain concepts from radically different and potentially more relevant foundations. In the case of this study, the parameters formalized by AI related to the movement of physical systems, but the concept as a whole goes far beyond this area.

This approach can also be used in much more specific areas just like logistics, urban planning, climatology or public health, e.g. These are activities where AI has already brought great disruption. But until then, only complementary elements made it possible to optimize concepts imagined by humans.

A system of this kind, on the other hand, could make it possible to highlight phenomena and approaches that would have scientists have so far completely evaded… including on the functioning of the AIs themselves!

Certainly, between the AIs that are already revolutionizing scientific research, those that are writing scientific papers about themselves, and work of this nature, there is much to be excited about the future of AI in research.

The technical documentation relating to this work is available here.

Leave a Comment