As we increasingly delegate decision-making to algorithms, whether directly or indirectly, important questions emerge in circumstances where those decisions have direct consequences for individual rights and personal opportunities, as well as for the collective good. A key problem for policymakers is that the social implications of these new methods can only be grasped if there is an adequate comprehension of their general technical underpinnings. The discussion here focuses primarily on the case of enforcement decisions in the criminal justice system, but draws on similar situations emerging from other algorithms utilised in controlling access to opportunities, to explain how machine learning works and, as a result, how decisions are made by modern intelligent algorithms or 'classifiers'. It examines the key aspects of the performance of classifiers, including how classifiers learn, the fact that they operate on the basis of correlation rather than causation, and that the term 'bias' in machine learning has a different meaning to common usage.An example of a real world 'classifier', the Harm Assessment Risk Tool (HART), is examined, through identification of its technical features: the classification method, the training data and the test data, the features and the labels, validation and performance measures. Four normative benchmarks are then considered by reference to HART: (a) prediction accuracy (b) fairness and equality before the law (c) transparency and accountability (d) informational privacy and freedom of expression, in order to demonstrate how its technical features have important normative dimensions that bear directly on the extent to which the system can be regarded as a viable and legitimate support for, or even alternative to, existing human decision-makers.

We cast the problem of combinatorial auction design in a Bayesian framework in order to incorporate prior information into the auction process and minimize the number of rounds to convergence. We first develop a generative model of agent valuations and market prices such that clearing prices become maximum a posteriori estimates given observed agent valuations. This generative model then forms the basis of an auction process which alternates between refining estimates of agent valuations and computing candidate clearing prices. We provide an implementation of the auction using assumed density filtering to estimate valuations and expectation maximization to compute prices. An empirical evaluation over a range of valuation domains demonstrates that our Bayesian auction mechanism is highly competitive against the combinatorial clock auction in terms of rounds to convergence, even under the most favorable choices of price increment for this baseline.

Most video games set in the First or Second World War shoot for gritty realism. It's the visual fidelity, paired with addictive combat, that draws players in. The story-driven experience, set during the last two years of World War I, has a painterly art style inspired by artists such as Claude Monet and Joseph Mallord William Turner. Every scene is created with tiny brushstrokes that slowly move, transform and dissolve. As you move the twin protagonists forward, the paint will change again to reflect your position in the level.

Efforts to develop artificial intelligence (AI) are increasingly being seen as a global race, even a new Great Game. Apart from the race between countries to become more competent and establish a competitive advantage in AI, enterprises are also in a contest to acquire AI talent, leverage data advantages, and offer unique services. In both cases, success would depend on whether AI solutions can be democratized and distributed across sectors. The global AI race is unlike any other global competition, as the extent to which innovation is being driven by governments, the corporate sector or academia differs substantially from country to country. On average, though, the majority of innovations so far have emerged from academia, with governments contributing through procurement, rather than internal research and development.

Is there a'New Normal' looming on the horizon for how food and beverage decisions are made? Q: Could you provide our readers with a brief introduction to SteadyServ Technologies? A: The mission of SteadyServ is to provide a clear understanding of the best decision options available for retailers and suppliers to grow beverage sales and profitability. Today's retail environment is a chaotic and rapidly changing environment in which competition and costs are increasing at an astounding rate. Consumers and what they expect shifting faster than retailers are used to. Because of this some are beginning to use smarter digital tools to capture the maximum share of wallet of a consumer.

Time travel to the UK in 2025: Harry is a teenager with a smartphone and Pauline is a senior citizen with Alzheimer's who relies on smart glasses for independent living. Harry is frustrated his favourite online game is slow, and Pauline is anxious because her healthcare app is unresponsive. Forbes predicts that by 2025 more than 80 billion devices, from wearables and smartphones, to factory and smart-city sensors, will be connected to the internet. Something like 180 trillion gigabytes of data will be generated that year. Currently almost all data we generate is sent to and processed in distant clouds.

I was born with the usual set of limbs. When I was nine months old, I contracted meningococcal septicaemia, a dangerous infection of the blood, which very nearly killed me. I survived, but because I had sustained major tissue damage, it became necessary to amputate my right leg below the knee, all of the fingers on my left hand and the second and third digits on my right hand. I learned to walk on a prosthetic leg at the age of 14 months, and have gone through my life wearing a succession of artificial limbs. As time has passed and technology has advanced, so too have my limbs. Like our mobile phones, prostheses have become lighter, faster and more efficient. When I was nine, I was fitted with a lifeless silicone hand, a useless thing that was purely cosmetic, and so clumsy that I refused to wear it after the first day. Now, at 21, and a student in my third year at Edinburgh University, I wear a bionic arm with nimble fingers that move independently, which I operate using controlled muscle movements in my forearm, as well as an app on my phone. As a child I wore a stiff artificial leg attached with straps that frequently fell off; earlier this summer, I took delivery of a new dynamic right leg with shock absorption and carbon fibre blades. Prosthetics have been around for more than 3,000 years: wooden toes, which strapped on and were specifically designed to work with sandals, were found on the feet of Ancient Egyptian mummies.

The next generation of unmanned drones will act more like birds than machines, thanks to new study by researchers from RMIT University in Melbourne and ISAE-Supaéro in Toulouse. The study includes experiments with drones that can sense wind gusts and thermals, then use them to gain speed or altitude, just like birds do. Dr Abdulghani Mohamed, who leads a large research program into bio-inspired technology in RMIT's Unmanned Aircraft Systems (UAS) research team, said the world-first project had exceeded expectations. "The results of our gust soaring system were remarkable and represent a big leap in energy harvesting for drones," Mohamed said. "This technology not only allows a drone to gain kinetic energy to fly faster but also means less work and more efficiency for the propulsion system, potentially enabling the next generation of drones to increase their flight time on limited resources."

AI is a hotly debated topic in every conversation, so much so that we have moved from saying'there is an app for that' to'there is an AI for that'. Oliver Schabenberger, chief operating officer and chief technology officer at SAS, observes how AI has permeated everyday discourse in recent years. Yet, AI has not always been talked about this way. An overhype of the technology led to'AI winter' in the 1980s, he says in his keynote at the Analytics Experience conference this week in Milan. During cocktail gatherings, saying one worked in AI could kill a conversation.

In this paper, we introduce proximal gradient temporal difference learning, which provides a principled way of designing and analyzing true stochastic gradient temporal difference learning algorithms. We show how gradient TD (GTD) reinforcement learning methods can be formally derived, not by starting from their original objective functions, as previously attempted, but rather from a primal-dual saddle-point objective function. We also conduct a saddle-point error analysis to obtain finite-sample bounds on their performance. Previous analyses of this class of algorithms use stochastic approximation techniques to prove asymptotic convergence, and do not provide any finite-sample analysis. We also propose an accelerated algorithm, called GTD2-MP, that uses proximal "mirror maps" to yield an improved convergence rate. The results of our theoretical analysis imply that the GTD family of algorithms are comparable and may indeed be preferred over existing least squares TD methods for off-policy learning, due to their linear complexity. We provide experimental results showing the improved performance of our accelerated gradient TD methods.