Google DeepMind Affects Our Future Right Now

31 March 2021

DeepMind affects our future but also the present. We explore how the new AI technology from Google is changing lives. 

DeepMind was founded in 2010 by Mustafa Suleyman and university buddies Demis Hassabis and Shane Legg and gained by Google in 2015. It was a collateral acquisition. Google had just purchased Alphabet Inc., and since DeepMind was a daughter company of Alphabet, it became Google’s property, respectively.

So who controls the DeepMind now? Still, all the AI enthusiasts, pioneers of deep learning. All they wanted was to take DeepMind to the next level, that’s why they found common ground with Google pretty quickly.

The Early Days

Since then, Google has improved DeepMind substantially. The first breakthrough was in 2016 when DeepMind created AlphaGo, which eventually beat table game Go’s world champion Lee Sedol. Soon after, Google in collaboration with DeepMind developed a program known as AlphaZero, which can play games like chess, shogi, and Go mastering them via self-play.

The bunch has since received a lot of backing, financially and otherwise. Among others, Horizons Ventures, Founders Fund, and even Elon Musk are investing billions of dollars in the company. These investors believe that artificial intelligence has a big future in many industries. A founder says he believes DeepMind solutions can reach human-level intelligence when a program can play multiple games.

AI investments

Now and the Future  

The investors are likely right. Evidence shows that DeepMind could affect our lives a lot immediately and in the long term. The following are five potential applications of the technology;

          1. A new tool in cancer diagnosis

Cancer kills 500,000 people worldwide every year, partly because of challenges in detecting and diagnosing the disease. Mammograms used in present-day diagnosis miss thousands of cancers every year, often resulting in false alarms and misdiagnosis. DeepMind can help diagnose two specific cancers–breast cancer and cancer of the neck and head.

 DeepMind works together with Google’s AOI health analytics and research team and several institutions, led by the Cancer Research Center at the Imperial College and the UK NHS, to improve breast cancer diagnosis and cancers of the neck and head. The technology identifies cancerous tissues more effectively than mammograms.

Clinicians spend at least four hours prepping patients before cancer treatments to prevent delicate tissue damage. Then they fed the information into the radiotherapy machine so the doctors can target affected areas without damaging healthy tissue. DeepMind can cut the prepping time down to one hour.

          2. Game-changer in solving protein structures

DeepMind has also made a significant leap in solving one of biology’s most complex challenges–determining a protein’s 3D shape from its amino acid sequence. The ability to predict protein structures from their amino acid sequences would be a groundbreaking achievement for medicine and the life sciences. It’s widely considered the key to understanding the building blocks of cells, thus enabling faster and more advanced drug discovery.

DeepMind recently created a program known as AlphaFold that has outperformed hundreds of other teams in a biennial protein-structure prediction challenge. The Critical Assessment of Structure Prediction (CASP) challenge was announced virtually in November last year.

          3. Heart Failure Prediction 

The NIX company elaborated the use case of deep learning techniques for Heart Surgery Hospitals. Our team uses СNN to develop a system of predicting heart failures by turning 2D data into 4D flow MRI types. The project is still in progress and we are constantly working to make it more precise. Here we facing the fundamental challenge in recording and analyzing blood flow data. But when successful it will be significant progress in treating people with heart disease and keeping track while the patient is in remission.

          4. Detecting acute kidney injury 

DeepMind recently developed a patient safety alert app called Streams, which tests results whenever a patient enters data and sends medical staff instant alerts if an urgent assessment is necessary. It displays the results of scans, x-rays, and blood tests at the touch of a button.

One of the most extensive uses of Streams to date is the diagnosis of acute kidney injury. In 2017, when it was first introduced at the Royal Free NHS Trust in London, it proved that by assessing blood tests it could identify kidney injuries. Many other hospitals in the UK, including the Yeovil District Hospital NHS Foundation and Somerset NHS Foundation Trust uses this innovation since then.

Nurses at the said hospitals note that Streams makes it a lot easier to diagnose acute kidney failure and potentially save lives. It also saves them up to two hours a day dealing with other diseases.

         5. Closing the text-to-speech gap

DeepMind is contributing to this area in two principal ways. First, most notably, the company has teamed up with WaveNet to help patients with speech impairment who can’t get back their original voice. Text-to-speech systems often produce robotic or unnatural sounding voices. The two companies are working together and speech-impaired individuals to improve the situation.

The results are very positive, with Tim Shaw, an Amyotrophic Lateral Sclerosis patient, being the first beneficiary. While the standard approaches in recreating natural voices often require hours of recorded videos, DeepMind only needs a handful of such recordings.

The WaveNet technology also uses neural networks to generate Google Assistant voices for US English and Japanese speakers.  

         6. DeepMind for Google Maps

Finally, Google is also using DeepMind to improve its products, and Maps could be one of the big beneficiaries. The ability of DeepMind to predict upcoming traffic jams especially makes the technology indispensable.

A recent blog post from Google explains how all this is possible. Essentially, Google and DeepMind take data from multiple sources and feed it into machine learning models. Often, this data includes live traffic information gathered anonymously from users’ Android devices, speed limit information, historical traffic data, and information on constructions from local governments. It also considers the quality, size, and direction of the road. Then this data moves into neural networks developed by DeepMind. Neural networks generate patterns that can help experts predict future traffic.

AI in healthcare

Endless Possibilities

DeepMind has many other futuristic applications that are already proving valuable even now. These include the potential to improve wind farm efficiency, optimize recommendations in Google Play, and predict patient deterioration in military veterans.

Nix Solutions is following the trends keenly to determine ways small and medium-sized enterprises can benefit from the technology. Contact us today to learn more.