Over the past decades, researchers and ML practitioners have come up with better and better ways to build, understand and improve the quality of ML models, but mostly under the key assumption that the training data is distributed identically to the testing data. In many real-world applications, however, some potential training examples are unknown to the modeler, due to sample selection bias or, more generally, covariate shift, i.e., a distribution shift between the training and deployment stage. The resulting discrepancy between training and testing distributions leads to poor generalization performance of the ML model and hence biased predictions. We provide novel algorithms that estimate the number and properties of these unknown training examples---unknown unknowns. This information can then be used to correct the training set, prior to seeing any test data. The key idea is to combine species-estimation techniques with data-driven methods for estimating the feature values for the unknown unknowns. Experiments on a variety of ML models and datasets indicate that taking the unknown examples into account can yield a more robust ML model that generalizes better.

Kearns et al. [2018] recently proposed a notion of rich subgroup fairness intended to bridge the gap between statistical and individual notions of fairness. Rich subgroup fairness picks a statistical fairness constraint (say, equalizing false positive rates across protected groups), but then asks that this constraint hold over an exponentially or infinitely large collection of subgroups defined by a class of functions with bounded VC dimension. They give an algorithm guaranteed to learn subject to this constraint, under the condition that it has access to oracles for perfectly learning absent a fairness constraint. In this paper, we undertake an extensive empirical evaluation of the algorithm of Kearns et al. On four real datasets for which fairness is a concern, we investigate the basic convergence of the algorithm when instantiated with fast heuristics in place of learning oracles, measure the tradeoffs between fairness and accuracy, and compare this approach with the recent algorithm of Agarwal et al. [2018], which implements weaker and more traditional marginal fairness constraints defined by individual protected attributes. We find that in general, the Kearns et al. algorithm converges quickly, large gains in fairness can be obtained with mild costs to accuracy, and that optimizing accuracy subject only to marginal fairness leads to classifiers with substantial subgroup unfairness. We also provide a number of analyses and visualizations of the dynamics and behavior of the Kearns et al. algorithm. Overall we find this algorithm to be effective on real data, and rich subgroup fairness to be a viable notion in practice.

Scientists at the University of North Carolina at Chapel Hill have created a new method of particle tracking based on machine learning that is far more accurate and provides better automation than techniques currently in use. Single-particle tracking involves tracking the motion of individual particles, such as viruses, cells and drug-loaded nanoparticles, within fluids and biological samples. The technique is widely used in both physical and life sciences. The team at UNC-Chapel Hill that developed the new tracking method uses particle tracking to develop new ways to treat and prevent infectious diseases. They examine molecular interactions between antibodies and biopolymers and characterize and design nano-sized drug carriers.

Now that a new school year is upon us, it's the perfect time to get your dorm room set up with some sweet home tech. To make the decision easier, Amazon has discounted three of its best-selling Echo devices: the Fire TV with 4K Ultra HD is $40 (43 percent off), the 2nd-gen Echo is $85 (15 percent off), and the Fire TV Cube is $90 (25 percent off). You don't need to be a math major to see that those are fantastic prices. The Fire TV stick, Amazon's pendant-style streaming device, is compatible with 4K Ultra HD and High Dynamic Range TV. You'll be able to search through all your apps, from streaming services to social networks, with just your voice via Alexa, built in to the included remote. You can also connect to an existing Echo for hands-free control.

Artificial Intelligence means a lot of things to a lot of different people. It's difficult to nail down what exactly it is; it's easier to start with what it is not. AI is not a Jetsons-style robot that lives in your home and does all your domestic duties for you. It won't cook your bacon and eggs, nor will it serve you breakfast in the morning. AI also is not a Terminator-type robot that is intent on destroying the human race.

This new study is much more interested in the benefits of robot-assisted long-term coaching. As well as offering the benefit of bypassing any baggage associated with human interactions, this autonomous robot intervention allows for a system that supports and augments any work with other clinicians and teachers. Scassellati suggests further longer-term study will be necessary to better understand the benefits of the program but these results from just one month of work bodes well for future robot-assisted interventions helping autistic children develop social skills.

Artificial intelligence promises to make hiring an unbiased utopia. Employee referrals, a process that tends to leave underrepresented groups out, still make up a bulk of companies' hires. Recruiters and hiring managers also bring their own biases to the process, studies have found, often choosing people with the "right-sounding" names and educational background. Across the pipeline, companies lack racial and gender diversity, with the ranks of underrepresented people thinning at the highest levels of the corporate ladder. Fewer than five per cent of chief executive officers at Fortune 500 companies are women, and that number will shrink further in October when Pepsi CEO Indra Nooyi steps down.

Google appears to be preparing an AI-powered fitness coach for WearOS as it continues its mission to revitalise the platform. Fitness remains a key reason why people buy smartwatches. Beyond that, they're still really just an expensive place to see our smartphone notifications. WearOS' main competitor, the Apple Watch, has produced some clever features with regards to fitness. Among the most impressive is the ability to tap some gym equipment to bring up the relevant info and tracking for that specific machine.

For years, multiple-choice tests have allowed scanners to score results without human intervention. Now technology is coming directly into the exam hall. Coursera has patented a system to take images of students and verify their identity against scanned documents. There are plagiarism detectors that can scan essay answers and search the Web--or the work of other students--to identify copying. Webcams can monitor exam locations to spot malpractice.

Many companies are experimenting with AI on a small scale, and a few have made a commitment that their organizations will be "AI first" or "AI-driven." But what does this mean? What is AI doing or leading, and, in particular, what is the role of leadership in making organizations AI-driven? We see a lot of confusion around opportunity and action. In the 2018 Deloitte Global Human Capital Trends survey and report of business and HR leaders, 72% indicated that AI, robots, and automation are important -- but only 31% felt their organizations were prepared to address strategy to implement these technologies.