Google's artificial intelligence (AI) platform DeepMind revolutionizes the field, being now capable of learning based on information already possessed.
Google's DeepMind AI
DeepMind is able of learning, or better said of teaching itself, based on data it already possesses. According to The Next Web, this is a significant step forward for artificial intelligence, a real breakthrough that revolutionizes the field. DeepMind technology is based on Alphabet's hybrid system called Differential Neural Computer (DNC). The system uses smart AI and a neural net capable of quickly parsing it, paired with the existing data storage capacity of conventional computers.
Researchers Alexander Graves and Greg Wayne wrote on the DeepMind blog that DeepMind models can store complex data like computers, but also learn from examples like neural networks. These neural networks use an interconnected series of nodes in order to stimulate specific centers required to complete a complex task, much like the brain. In order to deliver the desired outcome, the AI is optimizing the nodes to find the quickest solution. Over time, The DeepMind AI system can get more efficient at finding the correct answer over time by using the data already acquired.
The DeepMind team gave two examples to further explain the process. The DeepMind system was able to figure out on its own additional connections after being told about relationships in a family tree. The system optimized its memory to speed up finding the information in future searchers.
In another example given, the DeepMind system found on its own additional routes and the relationship between them after being given the basics of the London Underground public transportation system. In both these examples, the DeepMind AI was able to derive an answer from prior experience, instead of having to learn every possible outcome to find a solution, in the way traditional computing systems do. By using this revolutionary process, the DeepMind system was able to beat a human champion at "Go," a game with an infinite number of combinations and millions o potential moves.
There are different points of view on DeepMind's artificial intelligence performance. Some of the AI researchers consider that this could be a serious breakthrough for ever-smarter AI that might one day be capable of learning and thinking as humans do. But other tech experts are concerned that might lead to a "Skynet" kind of threat, inspired from the "Terminator" movie franchise.
Hybrid Differentiable Neural Computing
According to Hot Hardware, Google's DeepMind artificial intelligence system has accomplished a lot and the progress in the field is expected to continue. This year alone DeepMind not only was able to win over a Go champion but it also improved speech recognition, attempted to cure blindness, got started on an AI kill switch and even helped to reduce a data center's power bill. All these applications would not be possible without the new hybrid system called Differentiable Neural Computer (DNC), a large external dataset paired with a neural network.
The DeepMind computer's AI is smart enough to analyze the given data in order to create its own connections based on what's stored in memory. What makes this DNC system different is that DeepMind is using its own memory for storing and processing information in a way much similar to the way our brains do. DeepMind can begin to discover patterns and give us accurate answers based on the information stored in memory.
The applications for this kind of artificial intelligence based on self-learning could be various. With a computerized brain evaluating all of the possibilities, we could see in the future a system like DeepMind being able to find the best solutions to various problems, from forecasting global weather patterns to medicine, economy or finances. This latest development from Google is pushing the boundaries. With deep learning artificial intelligence the sky is the limit.
