Google DeepMind AI can now learn things on its own

Thanks to their new hybrid system called Differential Neural Computer (DNC), it can pair a neural network with the data storage capacity of conventional computers, and learn from this external storage without any instructions.

Google took over British artificial intelligence company DeepMind back in 2014 and today it has taken machine learning to new heights. According to a recent report, the DeepMind AI can now teach itself based on information it has already been fed.

Thanks to their new hybrid system called Differential Neural Computer (DNC), it can pair a neural network with the data storage capacity of conventional computers, and learn from this external storage without any instructions.

The DNC combines external memory with the neural network approach of AI, where a massive number of interconnected nodes work dynamically to work like a brain. There is a controller inside the DNC that can optimise responses as well as compare its results with the desired and correct ones.

Over time it becomes more accurate and can figure out how to use its memory data banks at the same time. Similar to the brain, the neural network utilises an interconnected series of nodes to stimulate specific centers that would be needed to complete a certain task. Here, the AI is optimising the nodes to find the quickest solution to deliver the desired outcome.

"These models can learn from examples like neural networks, but they can also store complex data like computers," wrote DeepMind researchers Alexander Graves and Greg Wayne in a blog post.

For instance, after being told about relationships in a family tree, the DNC was able to figure out additional connections on its own all while optimising its memory to find the information more quickly in future searches.

Find our entire collection of stories, in-depth analysis, live updates, videos & more on Chandrayaan 2 Moon Mission on our dedicated #Chandrayaan2TheMoon domain.

Loading...




also see

science