Self-learning Artificial Intelligence Copernicus recently discovered through its data comprehension that the Earth orbits the Sun and we’re waiting for it to find out what happened before Big Bang. Is AI Copernicus up to the task?
For centuries, world astronomers were doubly sure that the Earth was a stationary body at the center of the universe. And that all other bodies revolved around it. Then came Nicolaus Copernicus in the 1500s and challenged the age-old wisdom. He brought out a new set of formulas that were much simpler and more accurate. Since then we’ve been certain that the Earth and all other planets orbit around the Sun.
With a similar style of calculation to the legendary Copernicus is a newly developed self-learning neural network AI Copernicus.
It recently re-discovered the fact that the earth orbits the sun. The AI was designed to believe that the Sun is at the center of the Solar System. It then based its calculations on that constant.
The feat might not come across with some startling new information. But it’s stunning, considering the fact that it took mankind more than a millennium and a half of astronomy. Before one man found out the real truth. Now, the scientific community has high hopes from the AI version of Copernicus. It is touted to play a significant role in advancing quantum physics. Consequently, physicists hope it brings us close to understanding the origins of the universe.
Understanding how AI Copernicus calculates
The neural network is self-learning i.e. it teaches itself the laws of physics. It analyses the new inputs to solve quantum-mechanics mysteries. Scientists are hoping Copernicus can discover new laws of physics.
Physicists hope that AI Copernicus formulates quantum mechanics by bringing out new unforeseen patterns in large data sets.
The idea behind the neural network’s creation and algorithmic structure comes from the work of another great scientist Albert Einstein. Einstein simplified Relativity with a beautifully concise equation E=mc2. The physicists at the Swiss Federal Institute of Technology (ETH) in Zurich took the same approach. They designed an algorithm to sift through large sets of data and also condense the information into a few basic formulae.
Inspired by the Human Brain
In order to achieve something close to human ingenuity with unbridled computing capacity, the researchers have designed Copernicus as a new type of machine-learning system the mimics the structure of the human brain. While conventional AIs learn to recognize objects and calculate findings through mathematical nodes (artificial neurons), instead of simplifying the information into simpler laws, conventional neural networks work like black boxes and spread their new knowhow to millions of mathematical nodes. There complex functioning then becomes impossible to interpret.
So, scientists had overcome this complexity and enable the neural network to think like a human brain. Thus, the physicists designed a kind of ‘lobotomized’ neural network. This means two sub-networks are connected to each other only through a small number of links. This enables the two sub-networks to work in tandem. While one learns from the huge data, the other uses the newfound ‘experience’ to predict and test those predictions. The condensed number of links forces the network to only pass on basic information.
How AI Copernicus rediscovered that Earth orbits the Sun?
Furthermore, the physicists put their new robotic brain to test. They provided simulated data about the movements of Mars and the Sun in the sky, as visualized from Earth. As per one of the physicists, while the algorithm derived the formulae, a human eye is needed. Human eyes can interpret equations and also understand how they relate to the movement of planets around the Sun.
The critical part the human eye plays is to single out the crucial parameters that describe a physical system. Roboticist Hod Lipson at Columbia University in New York City said, “these kinds of techniques are our only hope of understanding and keeping pace with increasingly complex phenomena, in physics and beyond.”
What’s in store for Copernicus to discover in the future – the origins of the Universe?
One of the leaders of the project, Physicist Renato Renner opines that his team aims to develop Copernicus’ machine-learning technologies to be able to solve ‘apparent contradictions in quantum mechanics’. Quantum mechanics is still as much a mystery to even the best of physics brains. The theory keeps on producing conflicting predictions of experiments’ outcomes.
Renner said, “It’s possible that the current way [quantum mechanics is] formulated is in some way just a historical artifact.”
So, the team at the ETH hopes that the neural network is able to produce a new Quantum mechanics formula free of the current contradictions. However, Renner acknowledges that the current techniques haven’t become sophisticated enough for the task.
So, the next objective for the physicists is to develop a new version of AI Copernicus. The new version will learn from experimental data. It will also come up with new experiments to test its own hypotheses of the origins of the universe.