Google Owned DeepMind Unveils the Latest Avatar of Its Go-playing Program AlphaGo. A Read You Need to Read!

by Guest Author on December 15, 2017

in Tech News

There is no limit for Google’s increased interest in the future technology artificial intelligence. Yes, Google owned DeepMind has unveiled the latest version of its Go-playing program AlphaGo. AlphaGo is a powerful AI that makes use of more than thousand years of human knowledge of the ancient Chinese board game Go.

The AI program AlphaGo Zero comes with immense capabilities and can work from a blank slate, without being helped by human. This is Google’s big feat in general-purpose AI. The talent that AlphaGo Zero holds can be used to solve many real-world problems. AlphaGo Zero is working on proteins fold, a process in which a polypeptide folds into a specific, stable, functional, three-dimensional structure. In fact, proteins fold is a huge scientific challenge. Google’s this effort would give the much needed boost to drug discovery.

On the other hand, artificial intelligence companies USA who are working on healthcare AI have got big boost to build better AI solutions to serve global healthcare.

What DeepMind’s CEO Demis Hassabis Says?

The CEO of DeepMind, Demis Hassabis said it is our big step towards building general-purpose algorithms. In fact, the prevalent AIs are just narrow AIs and they can just perform just one specific task, such as translating languages or recognising faces. But general-purpose AI can go beyond humans’ capabilities and can think and perform on their own and do multiple tasks at a time. Hassabis believes the descendants of AlphaGo would work alongside humans as experts in science and medical fields.

The previous versions of AlphaGo used to take human help to learn and build itself, mainly from a large number of games played by amateurs and professional gamers. But the current version learned completely on its own. It learned playing itself millions of times. It started to learn by randomly placing stones on the Go board and improved itself with winning strategies.

What AlphaGo’s Lead Researcher David Silver Says?

AlphaGo’s lead researcher David Silver says “the current version of AlphaGo has more powerful approach than the previous ones. And it is able to create knowledge itself. We eliminated the limitations of human knowledge.”

The program builds its skill through reinforcement learning. The program works in the same way a human tries to master the art of bike riding, where he uses scuffed knees and hands in different ways. When this AI goes good, it would make a heavy impact in general AI, but when it goes bad, the impact may be disastrous.

The core of this AI application is a group of software “neurons” that together build an artificial neural network. This network analyzes the positions of each piece on the Go board and decides which moves might be right to move next. And the probability of them ensures win. The application gets stronger and sharper with the each game it played.

AlphaGo Zero can work on less data and run on a smaller computer. Some researchers well explained the journey of AlphaGo Zero in the journal Nature. Their write ups cover the story of how AlphaGo Zero started off horribly, then geared up to the level of a naive amateur, and finally how it was able to deploy deeply strategic moves used by grandmasters. AlphaGo Zero started off with one common play, called a joseki, which was from just 10 hours of the game started. Other names called “small avalanche” and “knight’s move pincer” were soon able to follow. Amazingly, after just 3 days of the game, the program explored some supreme ways that humans are still studying. The program comprehended some advanced moves such as a pattern called a ladder.

Commenting on the intelligence of AlphaGo Zero, Hassabis said “AlphaGo Zero explores some best plays, josekis, and then it works on to find something even better. You can even witness its capability to rediscover thousands of years of human knowledge.”

Eleni Vasilaki, a professor of computational neuroscience at Sheffield University, stated it as an “impressive feat”. She said AlphaGo Zero is able to discover better moves that goes beyond human intelligence. Today’s computers are capable enough to outplay humans at games that include complex calculations and precision.

AlphaGo Zero got appreciation from many industry experts. Tom Mitchell, a computer scientist at Carnegie Mellon University in Pittsburgh, stated this AI application as an “outstanding engineering accomplishment”. Being impressed by AlphaGo, Tom said computers teach humans how to play Go better.

The president of the American Go Association Andy Okun says I don’t know if computers will dominate us, but it is really fun to play the game with neural-network software.

Conclusion:

Google’s this move is really a big boost for AI initiatives the world is working on, especially for AI development companies USA and UK. On the other hand, AI has been a nice-to-have solution for businesses of every vertical. So there is an increased demand for best artificial intelligence companies USA, UK and across the world.

Guest article written by: Namratha is a Technical content writer at USM Business Systems – USM is a certified Minority Business Enterprise and global IT services and staffing company in USA. USM is the best data science companies USA and USM provide services on Artificial Intelligence, Machine Learning, Industrial Automation and Big Data Analytics. Connect with Google+, Linkedin, Twitter.

Comments & Leave a Comment

comments

{ 0 comments… add one now }

 

Leave a Comment

CommentLuv badge

Previous post:

Next post: