Google seems to have taken another step forward with their progress in artificial intelligence as their new AI program can now navigate the London Underground system without repetitive feeding of data.
Most AI programs can do the same but the difference with the new Google AI agent is that it can learn the ropes in just one try. In addition, the same program also has the capability to answer several questions regarding a family tree.
Google DeepMind researchers developed the program without having to pre-program it to know what and how to learn. Once the map of the London Underground subway was given, it took care of the rest.
Lead Author of the study Alex Graves said that the unique thing about their new Google DeepMind project is how AI's retention of the methods and reasoning behind how it learns to navigate the complex map, WIRED.co.uk has learned. There were 20 researchers in total who worked on the new neural network program for Google DeepMind.
Graves said: "You can't give normal neural networks a piece of information and let them keep it indefinitely in their internal state - at some point it will be overwritten and they will essentially forget it."
The new AI program can then learn indefinitely without having to overwrite some information. Scientists believe that it is the next step for the improvement of virtual assistants that will benefit from neural networks and advanced machine learning capabilities to help humans.
Researchers were able to develop the new AI program by adding an external memory where the information was stored, The Guardian reported. If the machine needed to learn something based on the stored info, they can just grab it from the said external memory storage.
Google DeepMind's new AI program was able to quickly find the fastest route between the stops located in the map. It was also capable of answering simple questions that would otherwise render the current virtual assistants useless because they do not have such capabilities yet.
Further research could be made to improve the AI program altogether and it could be used in future Google products. The study was published in the Nature science and technology journal.