Deep Mind as Android improves by beating Go players – Android Planet

Google’s Deep Mind beats artificial intelligence AlphaGo one after the other human Go champion. Training this intelligence also has implications for the future of Android.

The robots Googles Boston Dynamics are becoming more sophisticated and scarier. How long will it be until these down kicked and pushed beings will turn against us and the destruction of mankind is near? So quickly will not go, because as yet the robots are kept pretty stupid. For the time being, because Google is working with Deep Mind also very advanced artificial intelligence.

At this time centering the progress of this intelligence around the complex oriental board game Go. Earlier this week, the artificial intelligence AlphaGo Go world champion Lee Sedol reports. Sedol claims that he was speechless after his loss: “I clearly lost. From the beginning there was not a moment I thought I was first. “This in itself is already very special, but the success of AlphaGo also has significance for the future of Android.

Go

the fight between the intelligence of computer and human being fought over the game. Mainly complex board with many rules. This allows the ways in which games can run almost innumerable. The game of chess was an excellent candidate for the computer used. As developed IBM’s Deep Blue chess computer and won the opposite grandmaster Garry Kasparov. It has been twenty years ago. The level of conflict between computer and human is now much higher and the game is now played is Go

"" align center wp-image-266 923 "http:. // www.androidplanet.nl/wp-content/uploads/2016/03/Screenshot-2016-03-11-at-12.33.54.png "alt =" Deep Mind Android "width =" 560 "height =" 288 " srcset="http://www.androidplanet.nl/wp-content/uploads/2016/03/Screenshot-2016-03-11-at-12.33.54-150x77@2x.png 300w, 768w http://www.androidplanet.nl/wp-content/uploads/2016/03/Screenshot-2016-03-11-at-12.33.54-768x395.png, http://www.androidplanet.nl /wp-content/uploads/2016/03/Screenshot-2016-03-11-at-12.33.54-225x116.png 225W, http://www.androidplanet.nl/wp-content/uploads/2016/03/ screenshot-2016-03-11-at-12.33.54-150x77.png 150w, http://www.androidplanet.nl/wp-content/uploads/2016/03/Screenshot-2016-03-11-at-12.33 .54-882x453.png 882w, http://www.androidplanet.nl/wp-content/uploads/2016/03/Screenshot-2016-03-11-at-12.33.54-78x40.png 78W, http: / /www.androidplanet.nl/wp-content/uploads/2016/03/Screenshot-2016-03-11-at-12.33.54.png 948w, http://www.androidplanet.nl/wp-content/uploads/2016/03/Screenshot-2016-03-11-at-12.33.54-300x154@2x.png 600w, http://www.androidplanet.nl/wp-content/uploads/2016/03/Screenshot-2016-03-11-at-12.33.54-225x116@2x.png 450w, http://www.androidplanet.nl/wp-content/uploads/2016/03/Screenshot-2016-03-11-at-12.33.54-78x40@2x.png 156w "sizes =" (max-width: 560px) 100vw, 560px "/>

What Go is so fascinating is that it’s much more complex than chess In taking turns game put the players. pebble on a plate, at a point where two lines intersect. the goal is to conquer territories on the board by terrain and surround the other player with your own color. like chess, it is therefore a strategy game. Players are constantly to respond to what the opponent while trying to work out strategies themselves.

AlphaGo

the number of moves and scenarios in chess is already enormous, but computers can this relatively easy to remember. The question is whether this has something to do with intelligence. with GO’s all remember a hopeless task, so why AlphaGo need to think smart and creative to beat human opponents.

For this AlphaGo uses neural networks, allowing the intelligence reacts to what happens on the board. AlphaGo plays the course of the game in his imagination, the most successful virtual paths are followed. The artificial intelligence then choose the move that proved the most successful in the simulation. AlphaGo include trained by putting thirty million of learning human experts so that it gets better able to predict the next move. Then AlphaGo taught himself new strategies by playing a lot.

Google Photos

The battle for beating the best Go players is fascinating, but what we have as Android users to? Give it a few years and our smartphones through this learning behavior much more intelligent. In limited form we use all the neural network of Deep Mind Google Photos. We can search all our photos to a cat or a particular person, then we have only the photos that we want to see presented. This is because the system learns which itself is a cat, and how a cat looks like, so that it can recognize a. Creative thinking that Deep Mind is developing can be later deployed more fascinating ways.

Personal Assistant

The key technology that Deep Mind will be used for the personal assistant. Although there already exist assistants like Siri and Cortana, these are still not at the level of an assistant, as in the movie Her. Actually, that’s where Assistant makers all now strive towards and Deep Mind thinks here. In Her assistant has a deep understanding about what you say and what you want to do. The assistant can learn from your behavior and so improve themselves. The assistants that now exist, a number of programmed paths that can be hiked, but once you go down that path, nothing happens. This is why we do not use much these assistants and Google’s focus is not the focus lays on in the current version of Android.

The goal is a system that can adapt freely. This is Deep Mind now learning with AlphaGo. Because of the incredible amount of possibilities and the fact that a game of Go through each move can turn the AI ​​must constantly adapt to a new reality. How AlphaGo learn to play, can also be used to learn to move an assistant than the programmed path.

Medical Applications

Another angle where Deep Mind puts his focus is health. Deep Mind Health will focus on making medical apps that, for example, doctors can see whether a patient is at risk for complications. At this time of Deep Mind artificial intelligence is not yet used in the development of apps, but in the future will be used AI to better predict the needs of the patient.

>

Although certain diseases the symptoms and course of the disease for many patients will be the same, each patient is different. People react differently to medications and have symptoms that usually is not associated with the disease. A creative thoughtful intelligence in our smartphone can recognize this, there to respond and doctors take better informed.

Lapse

In an interview with The Verge proposes Demis Hassabis Deep Mind that we will already notice subtle improvements over the next three years with our smartphone usage. This does not necessarily have to get the latest Android devices because Deep Mind works through servers. Older devices can thus be a lot smarter. In four or five years Hassabis expect we really going to see great strides in the capabilities of our devices. We can not wait to get started with their own smart assistant.