After sweeping excel, artificial intelligence targeting quantum physics?

after sweeping excel, artificial intelligence targeting quantum physics?

artificial intelligence to build "chess" god AlphaGo and Master some time ago out of the limelight.Now, artificial intelligence, or will solve another problem - solving contains hundreds of millions of particles in quantum systems.

as early as January 2016, Google go ai AlphaGo emerged, Mr Goolsbee, a professor at the university of the Swiss federal industrial calello (Giuseppe Carleo) was initially formed by using machine learning techniques to solve the problem of quantum mechanics.

after sweeping excel, artificial intelligence targeting quantum physics?

now, he has set up a neural network, the goal is no longer a go, but understand quantum mechanics system.

artificial intelligence on the go against human than in chess against human more difficult due to go on board, the number of pieces may be more than the number of atoms in the universe.This means that the artificial intelligence method of exhaustion "violence" against human on the go.

it is interesting to note that go with a classic problem of quantum mechanics are very similar: a quantum system that contains billions of atoms, and the connection between the atoms through complex equation, according to demand characteristic of the system.

material science research needs solving quantum system

any material in the form of the macro, are essentially quantum system, thus solving the quantum system is the study of material properties, and the key to the research and development of new materials.

quantum mechanics limit, can't be a moment accurately calculate a particle in what position.In addition, many particles with spin properties - either up, or down.Consists of 100 particles of quantum system, the sum of its spin state number have more than 10 to the 30th.

at the moment, even if the use of the most powerful supercomputers, human can only solve the quantum system of 48 particles.Even if not consider computing time, only consider the calculation results of storage, calello, estimates that even if we turn the earth into a super big hard disk, the hard disk is used to solve the 100 particles of quantum system is only just enough!

however, artificial neural network could simplify the problem.Artificial intelligence can find out the game winning moves, perhaps it can eventually win over quantum system to solve the problem.

dream machine

calello said, artificial neural network is very good at from the limited information are summarized.With some photos of calello training neural networks, it can be from a photo of had never seen before in the calello identified.

calello and Microsoft's matias DE Troy (Matthias Troyer) cooperation, build a simple, used to solve multiple subsystems body wave functions (i.e., quantum state probability) neural network.At the same time, the neural network can also calculate the lowest energy state of the system, which is one of the fundamental problems of quantum mechanics.

they use multiple has obtained results of quantum mechanical system to test the tool, the results show that the tool is better than the existing multibody system of quantum mechanics to solve tool.Therefore, a larger artificial intelligence solution tool should be able to more effectively solve large-scale quantum system to solve the problem.

calello jokingly said that this system successfully solves the problem of "schrodinger's cat".The university of Texas at Austin Scott aronson said Scott Aaronson, calello team work very well.Artificial intelligence this huge change, powerful tools are many fields of quantum mechanics field is not exception.The success of artificial intelligence to solve more dimension subsystem, will lay a foundation for more exciting achievements in the future.

compile: ion heart

reference: https://www.newscientist.com/article/2120856-ai-learns-to-solve-quantum-state-of-many-particles-at-once/

after sweeping excel, artificial intelligence targeting quantum physics?

for

location: Beijing

contact: [email protected]

share to a circle of friends is endorsed by

(more financial information,)

The related content recommendation