DeepMind has attracted mixed headlines since Google paid £400 million for the artificial intelligence (AI) startup in January 2014. When an AI system built by DeepMind won a series of Go matches against the world's best player of the Chinese board game, the possibilities of the technology inspired wonder. But that was tempered by criticisms of its controversial access to personal health records, with concerns raised over data privacy.
From their lab in Kings Cross, London, 140 DeepMind researchers develop software systems capable of independent learning through AI algorithms. DeepMind is working with a small number of organisations to implement its technology so far, but there are already some intriguing real-life examples emerging.
Read next: Google DeepMind: What is it, how does it work and should you be scared?
1. DeepMind use cases: Royal Free NHS Trust
In February Google launched DeepMind Health, with the goal of using its machine learning systems to improve healthcare treatment and digitise processes.
The first project announced is a partnership with the Royal Free NHS Trust in north London, involving the creation of a mobile app that provides clinicians with cutting-edge analytics. Known as Streams, the app was designed to improve the detection of acute kidney injury by instantly reviewing blood test results before sending an alert to the most appropriate clinician via a handheld device.
The data-sharing agreement gives access to encrypted healthcare data from an estimated 1.6 million patients who use the three hospitals run by the trust annually: the Royal Free, Barnet, and Chase Farm. The project attracted controversy for providing Google with confidential patient information, despite assurances from the tech giant that the information would only be used to inform diagnosis and treatment.
Read next: Google’s DeepMind promises openness as it begins public consultation over healthcare plans
2. DeepMind use cases: Moorfields Eye Hospital
The second NHS collaboration announced by Google was a partnership with Moorfields Eye Hospital. The project was set up to develop a machine learning–based system that can recognise sight-threatening eye diseases from a digital scan of the eye. While its initial collaboration with the NHS at the Royal Free focused on patient care, this is the first that is entirely dedicated to medical research.
The programme involves the analysis of more than one million anonymous eye scans to produce an algorithm that detects early signs of emerging eye conditions and increases the speed of diagnosis.
A Moorfields ophthalmologist called Pearse Keane was credited with first introducing the idea. He contacted the company after seeing their technology help computers learn how to play video games and believed it could be applied to images of the eye.
3. DeepMind use cases: University College London Hospital
DeepMind's latest research partnership with the NHS aims to tackle head and neck cancers. Before radiotherapy can begin, clinicians currently spend around four hours preparing a detailed map of each patient's body to avoid targeting the delicate surrounding tissue that can be damaged in treatment. The information is then fed into a radiotherapy machine to target the cancer without harming the healthy tissue.
Researchers at DeepMind believe machine learning can cut this time down to an hour. The team will analyse anonymised scans from UCLH patients to develop a radiotherapy segmentation algorithm that can automate parts of the process. They hope to eventually apply the algorithm to other parts of the body.
4. DeepMind use cases: WaveNet
While healthcare technology dominates the current DeepMind developments, its machine learning systems have also been extended to audio. Talking machines have a long history in science fiction and have already entered mainstream usage through commercial products such as Apple's personal voice assistant Siri. But the gap between __computer and human speech remains substantial.
DeepMind has developed a text-to-speech system that can close that gap by more than 50 per cent. Known as WaveNet, it uses a neural network to replicate the soundwaves produced by human speakers rather than the copying the language they use. The computing power required to power Wavenet limits its practical applicability for now, but Google has already presented samples of its automatically generated piano pieces.
5. DeepMind use cases: Google
Google already uses machine learning algorithms in a range of its own products, including Maps, Gmail, YouTube and Android. It believes that DeepMind’s technology could also be extended to search, robots and the internet of things in future. A DeepMind agent has already matched human performance at 49 Atari games including Pac-Man and Space Invaders, and developed the aforementioned AlphaGo, the first __computer programme to ever beat a professional player at Go.
Google has even used DeepMind to cut its electricity bills at its huge data centres. DeepMind algorithms were used to predict the air conditioning required to cool the vast number of servers powering its services, which can vary depending on the user demand. The results were efficiency savings of 40 percent in the cooling systems along, and a reduction to the overall energy used in the data centres of 15 percent.
0 comments:
Post a Comment