Artificial Intelligence breaks quantum physics
Artificial Intelligence is breaking the mould of quantum experimentation, scaling entanglement and superposition of states to unimaginable levels, without the direct input of scientists.
Eduardo Martínez de la Fe
Septiember 14, 2021
Artificial Intelligence (AI) is developing experiments that transcend any human ideas and making astonishing progress at the frontiers of experimental quantum physics, reports Scientific American.
Behind this feat is quantum physicist Mario Krenn, who this month started a new research group at the Max Planck Institute for the Science of Light in Germany to use AI algorithms as a source of inspiration in quantum physics.
These algorithms are aimed at developing new technologies, including improved quantum telescopes or microscopes, as well as new sources for quantum computers.
The main focus of this team is to understand new ideas and concepts in artificial intelligence systems, to develop new quantum technologies and to gain new insights into quantum physics itself.
The group will apply two different technologies to the design of new quantum experiments and quantum hardware: firstly, a powerful new graph-theoretic artificial intelligence algorithm, which will extract the conceptual kernels of solutions to difficult scientific problems, reports the Institute.
Secondly, the team will use Deep Learning technologies, which have been successfully applied in other fields such as materials design, to discover how machine learning models 'think' about the most complex scientific problems.
Complex intertwined states
The challenge is considerable and his work promises interesting results. In 2016, long before landing at the German institute, Krenn had solved the problem of creating highly complex entangled states involving multiple photons.
A machine learning algorithm developed by Krenn, called Melvin, had achieved the feat on its own, without anyone having provided it with the necessary instructions to generate such complex states.
What was most striking was that the algorithm had found a way to create complex entangled states without prior knowledge: it had even improved on an experimental (human) solution proposed in the 1990s.
This revelation has triggered an escalation of experiments by other teams aimed at testing the conceptual foundations of quantum mechanics in different ways, harnessing the power of Artificial Intelligence.
Krenn has not held back either and has improved Melvin with a new machine learning algorithm called Theseus, which is much more powerful than its predecessor: it will be the star of the new research team's developments at the German institute.
All this technological progress revolves around quantum entanglement, one of the most puzzling phenomena in quantum mechanics: it involves two particles, each occupying several states at the same time, an experience known as superposition.
When two particles, such as atoms, photons or electrons, become entangled, they also experience an inexplicable bond that is maintained even if the particles are on opposite sides of the universe. As long as they are entangled, the behaviour of the particles is bound to each other.
Quantum entanglement and superposition of states have become increasingly complex with various attempts to explore them with not two, but more particles, especially photons or particles of light.
Krenn is one of those who have scaled up the effects of quantum entanglement by involving not only more photons, but also by increasing the number of superpositions of quantum states thanks to AI.
Quantum superposition occurs when an elementary particle simultaneously possesses two or more states, as happens for example with photons: they can stay in two different places at the same time, something unimaginable in the ordinary physical world.
These superimposed states can be scaled with more photons and promise safer and faster quantum communications: three photons could be in a superposition of three states and achieve cubits in a three-dimensional quantum state.
Si un bit es la unidad básica de información en la computación clásica binaria (basada en 1 y 0), y un cúbit es su equivalente cuántico (utiliza la superposición para gestionar estados simultáneos de 1 y 0), en los nuevos estados superpuestos complejos se habla de sistemas ternarios (trits) cuánticos llamados cútrits, que conllevan una superposición de al menos tres estados básicos. La escalada cuántica no tiene límites teóricos.
If a bit is the basic unit of information in classical binary computing (based on 1 and 0), and a cubit is its quantum equivalent (it uses superposition to handle simultaneous states of 1 and 0), in the new complex superposed states we speak of quantum ternary systems (trits) called cubits, which involve a superposition of at least three basic states. Quantum scaling has no theoretical limits.
The Melvin algorithm achieved this ternary complexity of states and showed that its configuration could be used to generate high-dimensional entangled states.
Theseus surpasses it in capability and promises to reach more complex levels of quantum entanglement, with photons, cubits and AI quatrains thinking on their own.
These sophisticated algorithms do not wait for human instructions to break the current mould of quantum research, surpassing even the human ability to devise new experiments, as has already happened.
New technological phase
The design of new devices and experiments has historically relied on the intuition of human experts, Krenn et al. explain in Nature Reviews Physics.
The new phase is, however, inspired by the design of computers that are increasingly empowering scientists, particularly in the quantum domain, they add.
This field is uniquely important because complex computing can solve two major challenges of quantum experiments: that quantum phenomena are not intuitive, and that the number of possible configurations of quantum experiments is exponentially increasing with AI, conclude the authors of this paper.