Quantum computers were received into the tech world with lots of hopes because of their capability to solve complex tasks that conventional computers can’t handle. In fact, it is believed that quantum technology will completely replace the traditional computer.
However, the tech is not yet grown to meet the expectations scientists project. The quantum states have been struggling because of being super sensitive to constant instability from their environment. Well, that’s what artificial intelligence is seeking to resolve – by establishing an active defense, based on what experts call quantum error correction.
The Concept Quantum Error Correction System
See, a team lead by Florian Marquardt says they have managed to develop a quantum error correction system which runs on AI’s principles. The new model learns from errors then becomes smarter and more stable in operations.
In other words, this concept taps from AI’s principle of machine learning, in this case, the AlphaGo system. Go, as a gaming program powered by machine learning got into records after winning 4 out of 5 games against the best human player in the world. It is one of the most complex games ever created and it is said that it possesses more combinations of moves than the existing atoms in the whole universe.
So for anything to master this game, it needs an exceptional seer ability and learning capability, which is what AlphaGo had an advantage of, powered by its artificial neural networks. With that the systems are able to practice and execute hundreds of thousands of moves in minutes, behind the scene, thereby surpassing the knowledge of the wisest human player. To be exact, the Erlangen-based researchers choose to use neural networks that are very similar to this to achieve their goal of making an error-correction self-learning system, for use with quantum computers.
What Goes on Inside Artificial Neural Networks?
By definition, ANNs are computer programs that somewhat replicate the brain, how the nerve cells interconnect and sent signals to execute tasks. So tapping from that, the research in Erlangen has deployed two thousand ANN’s interconnected to one another. “We consider this a great move, taking the latest development in computer science and merging it to physical systems, which is double progress in the AI technology,” said F. Marquardt.
Being the number one area of application the team demonstrates how AI networks with AlphaGo’s model and how they are capable of learning independently. And how they manage to perform tasks that are compatible with correcting future quantum computers. Now, looking at how this is coming up, the scientist believe that with sufficient training, systems of this nature will go beyond error-correction to deeper integration functions.
Training the New System
ANN’s are nothing without training. Like the human brain, the complexities of these networks draw their value from the data they learn from. However, the interesting thing with this work is that Marquardt and team focused on training only one network and then allowing it to be a teacher-network to the rest of the neural networks in the system.
That gives the teacher-network more chance to be smarter, which again flows to the student networks. With that alone the researchers say, in this would make error correction in quantum computers sharper and more reliable than can ever be imagined.