Breakthrough made with computer that learns from memory
The DNC can take structured data like maps and family trees, and figure out the relationships between the items in it without being taught the different combinations that can be made. Image DeepMind
Google-owned DeepMind has made a major breakthrough in its artificial intelligence research. It has managed to develop a computer that learns from its own memory to answer questions.
Previously, AI needed to be fed all possible answers for it to be able to provide a response to a currently presented query. DeepMind’s new differentiable neural computer (DNC) can be fed standard information like a map or family tree and work out the relationships between the items listed on the data structure, reports The Next Web.
In essence, if the DNC is given a map and asked how to get somewhere, it will be able to “study” the map and answer the question without having prior knowledge of what the correct answer could be.
As the DNC’s name would suggest, the discovery is based on the concept of neural networks that mimic the human mind. Neural networks are currently being utilized for a number of things including Google Translate as well as popular electronic assistants from Apple, Microsoft and Google.
DeepMind is hoping that DNC will enable even further breakthroughs in AI research. It describes DNC as “a learning machine that, without prior programming, can organize information into connected facts and use those facts to solve problems.”
If a computer can learn to do any number of complex things by just learning from presented data, it would open the world to truly interactive computer systems. Alfred Bayle
