What Is Google DeepMind?

How deep learning is embedded into products you use

google deepmind
DeepMind uses neural networks and deep learning to train artificial intelligences. Yuichiro Chino / Moment / Getty

DeepMind can refer to two things: the technology behind Google’s artificial intelligence (AI), and the company that’s responsible for developing that artificial intelligence. The company called DeepMind is a subsidiary of Alphabet Inc., which is also Google’s parent company, and DeepMind's artificial intelligence technology has found its way into a number of Google projects and devices.

If you use Google Home or Google Assistant, then your life has already intersected with Google DeepMind in some surprising ways.

How and Why Did Google Acquire DeepMind?

DeepMind was founded in 2011 with the goal of “solving intelligence, and then using that to solve everything else.” The founders tackled the problem of machine learning armed with insights about neuroscience with the goal of creating powerful general-purpose algorithms that would be able to learn rather than needing to be programmed.

Several large players in the AI field saw the massive amount of talent that DeepMind put together, in the form of artificial intelligence experts and researchers, and Facebook made a play to acquire the company in 2012.

The Facebook deal fell apart, but Google swooped in and acquired DeepMind in 2014 for about $500 million. DeepMind then became a subsidiary of Alphabet Inc. during the Google corporate restructuring that took place in 2015.

Google’s main reason behind buying DeepMind was to jump-start their own artificial intelligence research. While DeepMind’s main campus remained in London, England after the acquisition, an applied team was dispatched to Google’s headquarters in Mountain View, California to work on integrating DeepMind AI with Google products.

What is Google Doing With DeepMind?

DeepMind’s goal of solving intelligence didn’t change when they handed the keys over to Google. Work continued on deep learning, which is a type of machine learning that isn’t task-specific. That means DeepMind isn’t programmed for a specific task, unlike earlier AIs.

For instance, IBM’s Deep Blue famously defeated chess Grandmaster Gary Kasparov. However, Deep Blue was designed to perform that specific function and was not useful outside of that one purpose. DeepMind, on the other hand, is designed to learn from experience, which theoretically makes it useful in many different applications.

DeepMind’s artificial intelligence has learned how to play early video games, like Breakout, better than even the best human players, and a computer Go program powered by DeepMind managed to defeat a champion Go player five to zero.

In addition to pure research, Google also integrates DeepMind AI into its flagship search products and consumer products like Home and Android phones.

How Does Google DeepMind Affect Your Daily Life?

DeepMind’s deep learning tools have been implemented across the entire spectrum of Google’s products and services, so if you use Google for anything, there’s a good chance that you have interacted with DeepMind in some way.

Some of the most prominent places DeepMind AI has been used include speech recognition, image recognition, fraud detection, detecting and identifying spam, handwriting recognition, translation, Street View, and even Local Search.

Google’s Super-Accurate Speech Recognition

Speech recognition, or the ability of a computer to interpret spoken commands, has been around for a long time, but the likes of Siri, Cortana, Alexa, and Google Assistant have brought it more and more into our daily lives.

In the case of Google’s own voice recognition technology, deep learning has been employed to great effect. In fact, machine-learning has allowed Google’s voice recognition to achieve an astounding level of accuracy for the English language, to the point where it’s just as accurate as a human listener.

If you have any Google devices, like an Android Phone or Google Home, this has a direct, real-world application to your life. Every time you say, “Okay, Google” followed by a question, DeepMind flexes its muscles to help Google Assistant understand what you’re saying.

This application of machine-learning to speech recognition has an additional impact that applies specifically to Google Home. Unlike Amazon’s Alexa, which uses eight microphones to better understand voice commands, Google Home’s DeepMind-powered voice recognition only needs two.

Google Home and Assistant Voice Generation

Traditional speech synthesis uses something called concatenative text-to-speech (TTS). When you interact with a device that uses this method of speech synthesis, it consults a database full of speech fragments and assembles them into words and sentences. This results in strangely inflected words, and it’s usually pretty clear that there isn’t a human behind the voice.

DeepMind tackled voice generation with a project called WaveNet. This allows artificially-generated voices, like the one you hear when you talk to your Google Home or Google Assistant on your phone, to sound much more natural.

WaveNet also relies on samples of real human speech, but it doesn’t use them to synthesize anything directly. Instead, it analyzes the samples of human speech to learn how the raw audio waveforms work. This allows it to be trained to speak different languages, use accents, or even be trained to sound like a specific person.

Unlike other TTS systems, WaveNet also generates non-speech sounds, like breathing and lip-smacking, which can make it seem even more realistic.

If you want to hear the difference between a voice generated through concatenative text-to-speech, and one generated by WaveNet, DeepMind has some very interesting voice samples that you can listen to.

Deep Learning and Google Photo Search

Without artificial intelligence, searching for images relies on context clues like tags, surrounding text on websites, and file names. With DeepMind’s deep learning tools, Google Photos search was actually able to learn what things look like, allowing you to search your own images and get relevant results without needing to tag anything.

For example, you might search “dog” and it will pull up pictures of your dog that you took, even though you never actually labeled them. This is because it was able to learn what dogs look like, in much the same way that humans learn what things look like. And, unlike Google’s dog-obsessed Deep Dream, it’s more than 90 percent accurate at identifying all sorts of different images.

DeepMind in Google Lens and Visual Search

One of the most stunning impacts that DeepMind has made is Google Lens. This is essentially a visual search engine that allows you to snap a picture of something out in the real world and instantly pull up information about it. And it wouldn’t work without DeepMind.

While the implementation is different, this is similar to the way that deep learning is used in Google+ image search. When you take a picture, Google Lens is able to look at it and figure out what it is. Based on that, it can perform a variety of functions.

For instance, if you take a picture of a famous landmark, it will supply you with information about the landmark, or if you take a picture of a local store, it can pull up information about that store. If the picture includes a phone number or email address, Google Lens is also able to recognize that, and it will give you the option to call the number or send an email.