A wide-ranging, UK government-commissioned industrial strategy review of the life sciences sector, conducted by Oxford University's Sir John Bell, has underlined the value locked up in publicly funded data held by the country's National Health Service -- and called for a new regulatory framework to be established in order to "capture for the UK the value in algorithms generated using NHS data". The NHS is a free-at-the-point of use national health service covering some 65 million users -- which gives you an idea of the unique depth and granularity of the patient data it holds. And how much potential value could therefore be created for the nation by utilizing patient data-sets to develop machine learning algorithms for medical diagnosis and tracking. "AI is likely to be used widely in healthcare and it should be the ambition for the UK to develop and test integrated AI systems that provide real-time data better than human monitoring and prediction of a wide range of patient outcomes in conditions such as mental health, cancer and inflammatory disease," writes Bell in the report. His recommendation for the government and the NHS to be pro-active about creating and capturing AI-enabled value off of valuable, taxpayer-funded health data-sets comes hard on the heels of the conclusion of a lengthy investigation by the UK's data protection watchdog, the ICO, into a controversial 2015 data-sharing arrangement between Google-DeepMind and a London-based NHS Trust, the Royal Free Hospitals Trust, to co-develop a clinical task management app.

We consider the problem of noisy matrix completion, in which the goal is to reconstruct a structured matrix whose entries are partially observed in noise. Standard approaches to this underdetermined inverse problem are based on assuming that the underlying matrix has low rank, or is well-approximated by a low rank matrix. In this paper, we propose a richer model based on what we term the "permutation-rank" of a matrix. We first describe how the classical non-negative rank model enforces restrictions that may be undesirable in practice, and how and these restrictions can be avoided by using the richer permutation-rank model. Second, we establish the minimax rates of estimation under the new permutation-based model, and prove that surprisingly, the minimax rates are equivalent up to logarithmic factors to those for estimation under the typical low rank model. Third, we analyze a computationally efficient singular-value-thresholding algorithm, known to be optimal for the low-rank setting, and show that it also simultaneously yields a consistent estimator for the low-permutation rank setting. Finally, we present various structural results characterizing the uniqueness of the permutation-rank decomposition, and characterizing convex approximations of the permutation-rank polytope.

Most content-based websites, like Yahoo News, HuffPost, or any given news site, organize their stories according to subject matter or in some similar way. You can imagine that websites with a huge amount of stories must need an automated method to filter or categorize them as the content is ingested into their systems. For example, algorithms that power Yahoo News label news articles with tags (e.g., Military conflict, Nuclear policy, Refugees) as they are ingested, and then display the content by subject matter and/or on a personalized feed. This well-known process of labeling content with all its relevant tags is known as Multilabel Learning (MLL). Up to now, whenever scientists and engineers use MLL to create their own specific models to label content however they like, they have used datasets that have pre-computed features like bag-of-words, or dense representations like doc2vec.

Big data was the first step in creating value from transactions and accumulated data. Now that we have warehouses filled with data the challenge has turned to delivering actionable insights to not just customers, but also employees. Smart businesses have realized that their employees, across teams, can make better decisions and scale expertise by democratizing the access to real-time data insights. These six companies are just a small sampling of how AI is helping companies empower all their employees to be as good as their best ones. H&R Block trains its tax pros to create tax returns for its customers but understanding and interpreting the 74,000 page tax code is daunting.

Unseen data conditions can inflict serious performance degradation on systems relying on supervised machine learning algorithms. Because data can often be unseen, and because traditional machine learning algorithms are trained in a supervised manner, unsupervised adaptation techniques must be used to adapt the model to the unseen data conditions. However, unsupervised adaptation is often challenging, as one must generate some hypothesis given a model and then use that hypothesis to bootstrap the model to the unseen data conditions. Unfortunately, reliability of such hypotheses is often poor, given the mismatch between the training and testing datasets. In such cases, a model hypothesis confidence measure enables performing data selection for the model adaptation. Underlying this approach is the fact that for unseen data conditions, data variability is introduced to the model, which the model propagates to its output decision, impacting decision reliability. In a fully connected network, this data variability is propagated as distortions from one layer to the next. This work aims to estimate the propagation of such distortion in the form of network activation entropy, which is measured over a short- time running window on the activation from each neuron of a given hidden layer, and these measurements are then used to compute summary entropy. This work demonstrates that such an entropy measure can help to select data for unsupervised model adaptation, resulting in performance gains in speech recognition tasks. Results from standard benchmark speech recognition tasks show that the proposed approach can alleviate the performance degradation experienced under unseen data conditions by iteratively adapting the model to the unseen datas acoustic condition.

Anomaly detectors are often used to produce a ranked list of statistical anomalies, which are examined by human analysts in order to extract the actual anomalies of interest. Unfortunately, in realworld applications, this process can be exceedingly difficult for the analyst since a large fraction of high-ranking anomalies are false positives and not interesting from the application perspective. In this paper, we aim to make the analyst's job easier by allowing for analyst feedback during the investigation process. Ideally, the feedback influences the ranking of the anomaly detector in a way that reduces the number of false positives that must be examined before discovering the anomalies of interest. In particular, we introduce a novel technique for incorporating simple binary feedback into tree-based anomaly detectors. We focus on the Isolation Forest algorithm as a representative tree-based anomaly detector, and show that we can significantly improve its performance by incorporating feedback, when compared with the baseline algorithm that does not incorporate feedback. Our technique is simple and scales well as the size of the data increases, which makes it suitable for interactive discovery of anomalies in large datasets.

Twenty years ago today, Skynet, the fictional artificial intelligence network and antagonist of the "Terminator" movie franchise, became self-aware. In the films, a Silicon Valley tech company built the network of supercomputers for the U.S. military to replace the humans who control America's nuclear arsenal. The idea was that the program would offer better reaction times and fewer mistakes than humans and engage in strategic warfare unencumbered by emotion, political incentives or ego. According to dates cited in "Terminator 2: Judgment Day," the network was activated on Aug. 4, 1997 and started thinking on its own on Aug. 29, 1997. But Skynet had a problem.

Fact Checking Some of the clearest and most objective cases of "Fake News" include incorrect facts that could be checked. This could include lying about the number of immigrants in a country or clearly exaggerated the cost of the national health service for a country. There are more than 114 fact checking initiatives in almost 50 countries. These include Full Fact, a charity in the UK that among other things, fact checks comments made by politicians in the House of Commons and the Crosscheck project from First Draft News that introduces a framework for fact checking that is being tested in the context of the French elections. These organisations are doing an amazing job making sure people are accountable.

America seems gripped with what we might call Battlestar Galactica Syndrome. In the TV show, society developed a technology (supersmart robots) which then produced unintended and terrible results (the near destruction of humanity). Where the syndrome comes in is that in not wanting to make the same mistake twice, a society might decide to completely ban the technology. We can't have killer robots imperil our species if we don't have smart robots to begin with! The problem is you probably won't have autonomous cars and 1,000 other cool inventions, too.

Fifty years ago, an MIT professor created a chatbot that simulated a psychotherapist. Named Eliza, it was able to trick some people into believing it was human. But it didn't understand what it was told, nor did it have the capacity to learn on its own. The only test it had to pass was: Could it fool humans? These days, with robotics advancing to drive cars, beat humans at chess and Go!, and replace entire workforces, Eliza's smoke and mirrors is child's play.