Google has done it again. This company won’t stop with its strategy to make the world a better technological place for living. According to some reliable sources Google is on the verge of releasing a new Smart Computer. Yes, you’ve heard it right. Google’s researchers have revealed a computer that it could become a self-programing machine one day. The developers at DeepMind start-up, which was bought by Google for $400 million, are working on this computer. They want to include mimics of the short-term memory of the human’s brain. I don’t know how they will achieve this, but I think they will succeed in everything they have put as their goal. These researchers believe this will be a longer learning process as the combination of ordinary computer and the works of a human brain could be a complicated process. The computer is expected to learn to program by himself.
The computers will become us!
This computer was described as “Neural Turing Machine” that will be able to learn everything as it stores more and more memory. The stored memory will be retrieved every time the mighty computer will feel the need of it. According to the researchers, they have introduced this Neural Turing Machine as a network architecture that has really big inspiration from both the design of digital computers and models of the biological working memory. All of their experiments have already brought some results they can rely on and demonstrate the capability of learning some simple algorithms from a simple data and using them to generalize all of them outside its regime for training.
This new computer will be a type of neural network that will adopt working with an external memory. The result will come as a new computer that learns and retrieve the data for performing some logical tasks. The team said that they are extending its neural networks by coupling all of them to external sources for memory. The interaction between the computer and the memory is done by attentional processes. This combined system is similar to a Turing Machine, but it allows to be trained with gradient descent. The first results are showing that this new smart computer can read and perform some simple algorithms such as sorting, copying and recall from other output and input examples.
I won’t comment whether this new smart computer is really a smart thing to release in the near future because we, the people, are not ready to be replaced with computers.