Let's say Alice has a book full of secrets that she wants to destroy. She tosses the book into a handy blackhole. Black holes, nature's fastest scramblers acting like garbage shredders and destroying large amounts of material, make Alice's secrets quite safe.Let's say Bob is her nemesis and has a quantum computer that is entangled with a black hole. (In entangled quantum system, actions taken on one particle affect their entangled companions regardless of distance, even if some go into a black hole.John Preskill and Patrick Hayden have done a famous thought experiment that says Bob can see a few particles light that leak from a black hole's edges. Bob then can use those photons to create qubits, the basic processing unit for quantum computing. This will reveal the particular physics behind Alice's text. He can then reconstruct the book.But it's not over yet.Recent research on quantum machine learning suggests that Alice's book may not be lost forever.QUANTUM COMPUTERS ARE USE TO STUDY QANTUM MECHANICSAlice may never be able to keep her secrets from the black hole. Our new no-go rule about information scrambling is still applicable to understanding random and chaotic systems within the rapidly expanding fields quantum machine learning, quantum thermodynamics and quantum information science.Richard Feynman was one of the greatest physicists in the 20th century. He introduced quantum computing to the world with a 1981 speech. They are notoriously difficult for other researchers to study.Along with other collaborators, our Los Alamos National Laboratory team has been studying algorithms for quantum computers, and in particular machine-learning algorithms, which some call artificial intelligence. This research sheds light upon which algorithms can work with existing noisy, intermediate-scale quantum computer systems and unresolved issues in quantum mechanics.We have focused on the training of variational quantum algorithms. The problem-solving landscape is composed of the peaks representing the highest-energy (undesirable), points in the system or problem and the valleys representing the lower-energy (desirable). The algorithm examines each feature of a mathematical landscape one by one to find the solution. The deepest valley holds the answer.SCRAMBLING IS KEPT BY ENTANGLEMENTWe wanted to see if quantum machine learning could be used to explain scrambling. This quantum phenomenon occurs when entanglement develops in a system composed of many particles or even atoms. Consider the initial conditions of this system to be a type of informationAlices book. The quantum system's entanglement between particles increases, and information spreads. This scrambling of information is crucial to understand quantum chaos, quantum science, random circuits, and many other topics.The ultimate scrambler is a black hole. We could explore it using a variational quantum algorithm and a theoretical quantum computer that is entangled with the blackhole. This would allow us to probe the scaleability and application of quantum machine learning. It could also help us learn more about quantum systems in general. We had an idea to create a variational quantum algorithm to exploit leaked photons in order to understand the dynamics of the blackhole. This would allow us to search through the mathematical landscape in order to find the lowest point.We would be able to reveal the dynamics of the black hole if we could find it. Bob could use this information to decode the scramblers code, and reconstruct Alice's book.Here's the problem. Hayden-Preskill's thought experiment assumes Bob is able to determine the blackhole dynamics that are scrambling information. We found instead that scrambling is not able to teach Bob these dynamics.STALLED OUT ON A BARREN PLATEAUHere's why: The algorithm stopped on a barren plateau. This is a common place in machine learning. A barren plateau is a problem-solving area that the algorithm can only see. The algorithm cannot find the down slope in this flat landscape. There is no path to the energy minimum. The algorithm simply spins, unable to find anything new. It does not find the solution.The resulting no-go theory states that any quantum machine learning strategy will face the barren plateau when it is applied to an unknown scrambling procedure.The good news is that most physical processes aren't as complicated as black holes and we will often have prior knowledge of their dynamics so the no-go rule doesn't invalidate quantum machine learning. It is up to us to choose the right problems to apply it to. We won't need quantum machine learning to see inside a blackhole to find out about Alices book or any other information anytime soon.Alice can therefore rest assured that all her secrets are secure.This article is an opinion and analysis piece. Scientific American does not necessarily endorse the views expressed in it.
