Innovation

Google DeepMind: The smart person's guide

Bought by Google in 2014 for 400 million pounds, DeepMind is an AI company best known for beating the world champion in the game Go. TechRepublic's comprehensive guide explains how it works.

Google's AlphaGo beats Lee Sedol at the game of Go
Image: Screenshot by Max Taves/CNET


With the boom in AI affecting virtually every industry, there has been an explosion in the research and development of machine learning, a subfield of AI. And, perhaps, no company better illustrates what machine learning is capable of than Google's DeepMind.

Founded in London in 2010 by Demis Hassabis, Shane Legg, and Mustafa Suleyman, DeepMind has developed machine learning systems that uses deep neural networks, reinforcement learning, and systems neuroscience-inspired models. The startup was purchased in January 2014 by Google for a rreported 400 million, with Hassabis remaining CEO of DeepMind.

Instead of relying on explicit programming, DeepMind applies general-purpose learning algorithms to a large data set in order to "train" the system and make predictions.

So, how does DeepMind really work, and what is it capable of? This comprehensive guide explains what the program really does.

SEE: Smart Person's Guides

Executive summary

  • What it is: Google DeepMind is a machine learning system that uses algorithms based on deep neural networks and reinforcement learning to train on massive datasets to be able to predict outcomes.
  • Why it matters: Google DeepMind is a prominent example of machine learning that illustrates what advanced AI is capable of.
  • Who does this affect: Anyone from businesses to computer scientists to engineers to end-users will be impacted by machine learning. The principles used by DeepMind can be applied to businesses that want to improve efficiency, gamers who want to learn how to master certain games, like Go.
  • When is this happening: Google DeepMind came into the public eye in October 2015 when it beat the European champion of Go, marking a breakthrough in AI that came a decade earlier than many experts predicted.
  • How to take advantage of DeepMind: While Google isn't sharing all the details of its machine learning secrets, much of its code is open source, and it has also shared the code of its software called TensorFlow, a deep learning engine. DeepMind also publishes many academic papers about its work online.

What it is

DeepMind is a subsidiary of Google that focuses on AI. More specifically, it uses a branch of AI called machine learning, which can include approaches like deep neural networks and reinforcement learning to make predictions. This can rely on massive data sets, sometimes manual data labeling—but sometimes not.

Many other AI programs like IBM's DeepBlue, which defeated Garry Kasparov in chess in 1997, have used explicit, rule-based systems that rely on programmers to write the code. However, machine learning enables computers to teach themselves and set their own rules, through which they make predictions.

In March 2016, DeepMind's AlphaGo program beat world champion Lee Sedol in 4 out of 5 games of Go, a complex board game—a huge victory in AI that came much earlier than many experts believed possible. It did this through combining "Monte-Carlo tree search with deep neural networks that have been trained by supervised learning, from human expert games, and by reinforcement learning from games of self-play," according to Google.

But beyond mastering games, deep learning has other more practical applications. In 2012, it was used to recognize a million images with a 16% error rate—which is now at about 5.5%. Deep learning is also used in text-based searches and speech recognition. According to founder Mustafa Suleyman, it achieved a "30% reduction in error rate against the existing old school system. This was the biggest single improvement in speech recognition in 20 years, again using the same very general deep learning system across all of these."

Deep learning is also used for fraud detection, spam detection, handwriting recognition, image search, speech recognition, Street View detection, and translation. According to Suleyman, deep learning networks have now replaced 60 "handcrafted rule-based systems" at Google.

Additional resources:

Why it matters

While Google DeepMind's accomplishments in the gaming world are impressive, the implications of its machine learning platform are far-reaching. An announcement that DeepMind was able to slash Google's electricity bill by increasing energy-efficiency has big implications in both economic and environmental realms.

Also, DeepMind's partnership with the National Health Service, part of DeepMind Health, employs machine learning in spotting critical conditions in eye health. It hopes to eventually use algorithms to personalize health care treatments, determining which work best on patients, given their previous medical history.

Additional resources:

Who it affects

DeepMind's machine learning platform has implications for just about any organization that wants to capitalize on its data to gain insights, improve relationships with customers, increase sales, or be competitive at a specific task. It has applications in government, business, education—virtually anyone who wants to make predictions, and has a large enough data set, can use machine learning to achieve their goals. It also has the potential to create many jobs in data labeling as well as disrupt jobs that were traditionally done manually.

Additional resources:

When is it happening?

Machine learning, popular in the 1980s, has seen a recent resurgence. Google DeepMind is one step in the chain of AI platforms that use neural networks to make predictions. Here are some highlights:

  • 2010: DeepMind is founded in London, England.
  • 2011: Google Brain was created, which was a deep neural network that could identify and categorize objects.
  • 2014: Facebook's DeepFace algorithm was introduced, which could recognize people from a set of photos.
  • 2015: Amazon launched its machine learning platform and Microsoft offered a Distributed Machine Learning Toolkit.
  • 2016: Google's DeepMind program "AlphaGo" beat the world champion, Lee Sedol, at the complex game of Go.

Additional resources:

How to take advantage of DeepMind and machine learning

There are many online resources for machine learning. To get an overview of how to create a machine learning system, a series of YouTube videos by Google Developer has come in handy for me. There are also classes on machine learning from Coursera and many other institutions.And, to further integrate machine learning into your organization, you can use resources like Microsoft's Azure, Google Cloud Machine Learning, Amazon Machine Learning, IBM Watson, and free platforms like Scikit .

Additional resources:

About Hope Reese

Hope Reese is a Staff Writer for TechRepublic. She covers the intersection of technology and society, examining the people and ideas that transform how we live today.

Editor's Picks

Free Newsletters, In your Inbox