Samsung kept its word that it will launch the Galaxy S9 with a "reimagined" camera this Sunday. Part of the reimagined technology is the introduction of a new feature called AR Emoji that pretty much functions like Apple's Animoji on iPhone X. In a press release, Samsung said that it developed AR Emoji "to offer users a visual communication tool that allows them to express themselves in a way that's unique to them." According to the South Korea giant the feature's machine-learning algorithm analyzes over 100 distinct facial features when generating a personalized 3D avatar or emoji. From eyes, nose, mouth, cheeks and just about any facial feature, AR Emoji copies everything to come up with an animated emoji version of the user.

For years now, some of the smartest and most influential people on Earth have been warning about the dangers of artificial intelligence, laying out nightmarish scenarios that sound like they were pulled from the pages of a Hollywood script. Stephen Hawking, the renowned theoretical physicist, has given humanity a tight deadline for escaping the planet. Disease-fighting business magnate Bill Gates, meanwhile, has said he doesn't understand why "some people are not concerned" about the threat posed by super-intelligent machines. However, Kevin Kelly, the executive editor of Wired magazine, is offering a decidedly optimistic answer to Gates's question. Contrasting humans with technology ignores something that has been true for the past 10,000 years or so -- something there's no coming back from, Kelly told a reporter at the World Government Summit in Dubai earlier this month.

In the latest issue of the biology journal Cell, they describe a way of using artificial intelligence and machine learning to recognise macular degeneration and diabetic macular oedema, the two most common causes of irreversible blindness. They say this new technology could be used anywhere in the world and could prove especially important in rural areas, and places like China, India, and Africa. Traditional computer programs work by defining actions to take given certain circumstances - such as typing in some numbers and doing calculations. Type in something it isn't programmed to react to - like a word instead of a number - and it won't know what to do. AI is a part of computer science where machines are designed to have some intelligent behaviour.

Samsung has officially unveiled its successor to last year's Galaxy S8 flagship, and it appears the Galaxy S9 has a bit of a disappointing aspect. Apparently, its Intelligent Scan feature isn't as good as Apple iPhone X's Face ID even though the South Korean giant already combined facial recognition with its iris scan technology. Two years ago, Samsung unveiled the now defunct Galaxy Note 7 that came with the company's first commercial implementation of iris scanning technology. The company's 2016 phablet may have received a lot of negative press, but the iris scanning technology was definitely a stunning feat at the time. When Samsung introduced the Galaxy S8 and Note 8 last year, it did not miss the chance to bring back the iris scanner.

Music genre classification is one example of content-based analysis of music signals. Traditionally, human-engineered features were used to automatize this task and 61% accuracy has been achieved in the 10-genre classification. However, it's still below the 70% accuracy that humans could achieve in the same task. Here, we propose a new method that combines knowledge of human perception study in music genre classification and the neurophysiology of the auditory system. The method works by training a simple convolutional neural network (CNN) to classify a short segment of the music signal. Then, the genre of a music is determined by splitting it into short segments and then combining CNN's predictions from all short segments. After training, this method achieves human-level (70%) accuracy and the filters learned in the CNN resemble the spectrotemporal receptive field (STRF) in the auditory system.

Statistical methods protecting sensitive information or the identity of the data owner have become critical to ensure privacy of individuals as well as of organizations. This paper investigates anonymization methods based on representation learning and deep neural networks, and motivated by novel information theoretical bounds. We introduce a novel training objective for simultaneously training a predictor over target variables of interest (the regular labels) while preventing an intermediate representation to be predictive of the private labels. The architecture is based on three sub-networks: one going from input to representation, one from representation to predicted regular labels, and one from representation to predicted private labels. The training procedure aims at learning representations that preserve the relevant part of the information (about regular labels) while dismissing information about the private labels which correspond to the identity of a person. We demonstrate the success of this approach for two distinct classification versus anonymization tasks (handwritten digits and sentiment analysis).

Predictive models that generalize well under distributional shift are often desirable and sometimes crucial to building robust and reliable machine learning applications. We focus on distributional shift that arises in causal inference from observational data and in unsupervised domain adaptation. We pose both of these problems as prediction under a shift in design. Popular methods for overcoming distributional shift make unrealistic assumptions such as having a well-specified model or knowing the policy that gave rise to the observed data. Other methods are hindered by their need for a pre-specified metric for comparing observations, or by poor asymptotic properties. We devise a bound on the generalization error under design shift, incorporating both representation learning and sample re-weighting. Based on the bound, we propose an algorithmic framework that does not require any of the above assumptions and which is asymptotically consistent. We empirically study the new framework using two synthetic datasets, and demonstrate its effectiveness compared to previous methods.

Variational Bayesian neural nets combine the flexibility of deep learning with Bayesian uncertainty estimation. Unfortunately, there is a tradeoff between cheap but simple variational families (e.g.~fully factorized) or expensive and complicated inference procedures. We show that natural gradient ascent with adaptive weight noise implicitly fits a variational posterior to maximize the evidence lower bound (ELBO). This insight allows us to train full-covariance, fully factorized, or matrix-variate Gaussian variational posteriors using noisy versions of natural gradient, Adam, and K-FAC, respectively, making it possible to scale up to modern-size ConvNets. On standard regression benchmarks, our noisy K-FAC algorithm makes better predictions and matches Hamiltonian Monte Carlo's predictive variances better than existing methods. Its improved uncertainty estimates lead to more efficient exploration in active learning, and intrinsic motivation for reinforcement learning.

Interpretation of a machine learning induced models is critical for feature engineering, debugging, and, arguably, compliance. Yet, best of breed machine learning models tend to be very complex. This paper presents a method for model interpretation which has the main benefit that the simple interpretations it provides are always grounded in actual sets of learning examples. The method is validated on the task of interpreting a complex regression model in the context of both an academic problem -- predicting the year in which a song was recorded and an industrial one -- predicting mail user churn.

Avian Influenza breakouts cause millions of dollars in damage each year globally, especially in Asian countries such as China and South Korea. The impact magnitude of a breakout directly correlates to time required to fully understand the influenza virus, particularly the interspecies pathogenicity. The procedure requires laboratory tests that require resources typically lacking in a breakout emergency. In this study, we propose new quantitative methods utilizing machine learning and deep learning to correctly classify host species given raw DNA sequence data of the influenza virus, and provide probabilities for each classification. The best deep learning models achieve top-1 classification accuracy of 47%, and top-3 classification accuracy of 82%, on a dataset of 11 host species classes.