When you think of deep learning, you might not think of the latest breakthroughs in machine learning. However, deep learning algorithms are improving our understanding of the world around us. For instance, DeepMind’s AlphaFold software makes predictions of protein structures based on experimentally determined models.
They’ve spent years developing accurate metrics and ranking predictions. This can save us years of experiments and costly mistakes. What is DeepMind Working On?
DeepMind is part of Alphabet, a technology company owned by Google. It was founded in September 2010 and acquired by Google in 2014. The company is based in London with research centres in Canada, France, and the United States. In 2015, DeepMind became a wholly owned subsidiary of Alphabet Inc., the largest company in the world. It has a history of breakthroughs in artificial intelligence.
The latest research is already reshaping the world. DeepMind has successfully developed a version of Google Maps that applies graph networks to navigate the world. This application is now serving queries in Google Maps across the world. DeepMind is making huge advances in healthcare, so keep an eye on their work. You never know when breakthroughs will happen. It’s exciting times! DeepMind is already doing great things.
While DeepMind has successfully developed AlphaFold, they’re now seeking research that can make the software even better. These researchers are also seeking ways to break AlphaFold’s code, so that it can learn how to handle mistakes and improve itself. This marks a new chapter in Hassabis’ life, as he juggles his leadership role at DeepMind with his role as CEO of Isomorphic Labs, an Alphabet sister company focused on bringing AI to biotech and medicine.
A recent project launched by DeepMind Health uses machine learning to improve healthcare. It uses artificial intelligence (AI) and big data to digitize medical processes. One project, which was completed in February 2017, was to diagnose acute kidney injuries by analyzing blood tests. The work was then rolled out to Yeovil District Hospital NHS Foundation, Taunton and Somerset NHS Foundation Trust, and Imperial College Healthcare NHS Trust. It’s currently in the experimental stages and has not yet been fully implemented.
DeepMind is collaborating with the U.K. Department of Veterans Affairs to improve the detection of Acute Kidney Injury (AKI). Acute kidney injury, a leading cause of patient deterioration, is one of the most common causes of death in hospitals. Researchers are trying to develop an algorithm that can identify AKI and improve detection. These algorithms may eventually be applied to other parts of the body, including cancer.
DeepMind’s work is instructive for the development of artificial intelligence systems. Google has acquired Boston Dynamics and other companies to further the development of AI. Some companies are studying the topics of business intelligence development and cybersecurity, and experts are looking at ethics. Those who are concerned about AI’s ability to determine race, gender, and other factors have a good reason to be wary. DeepMind is working on a way to make artificial intelligence that isn’t biased.
The work of two researchers at DeepMind has a broader application. In the past, they developed a technique called “graph representation learning” that allows them to create a representation of graph-shaped data. Text and images, for example, can be seen as graphs of pixels linked together by proximity. Things in nature are naturally represented as graph structures. A recent breakthrough in deep learning has made these algorithms far more accurate.
In the same way, AlphaFold allows AI to focus on particular parts of data. The technology was developed to train AI to focus on a specific part of the input data. The system is currently used to make predictions of protein structure, which is a key task in drug discovery. The AI may one day even be able to predict the shape of proteins. The company has also been working on a new program that will allow computers to predict the structure of amino acids.