Solving Super Mario using D-Wave’s quantum computer and machine learning.
The useful usage of a quantum computer is not only a quantum speedup what is expected. We now focused on the expected low consumption of the electricity power because the main superconducting qubit is an analogue device that the consumption of the chip is almost zero ideally.
We now use a machine learning technique which the mario learn the most efficient way gradually from the action and environment which is so called reinforcement learning.
We mainly use the generative model of machine learning which generate samples from the structure of the model. Usually when we use D-Wave machine the main model is called ising model from the physics model.
We update the ising model as a formulation of cost function and get the new generated sample of qubit value 0 and 1 (or -1 and 1)
Reward getting from the environment
The each agent get the reward from the environment and update the ising model cost function.
For the future perspectives
For the future perspectives, the consumption of the energy power is the most attractive feature of the machine learning on D-Wave machine because the consumption of the electrical power is almost zero from the superconducting qubit.
In near future the machine learning will be used a lot in our daily life. The more we use machine learning as a automation process of the social society, the more the importance of the superconducing qubit will be bigger than now.