It seems insects might be able to see much more than we previously thought. New research has revealed that both the honeybees and wasps are able to learn achromatic (black and white) images of human faces. Despite having tiny brains made up of just one million brain cells – compared to the 86 billion that make up a human brain – they appear to visually process faces in a similar way to how we do. This is despite them having no evolutionary reason for doing so, writes Dr Adrian Dyer, an associate professor from RMIT University in Australia for The Conversation. Understanding how this developed could help researchers create smarter artificial intelligence, Dr Dyer says.

"How shall I compare thee to a classic Windows Bliss wallpaper? While we may not be able to master the classic Shakespearean sonnet, scientists have been working on AI that can. A deep learning artificial intelligence bot created by researchers at IBM Research Australia, the University of Melbourne, and the University of Toronto has been trained on 2,600 real sonnets, according to Digital Trends. It reflects the rhyming pattern and pentameter of the poetic format used by the bard. "When we started the project, a research question that we wanted to address was, 'how do we build machines that can produce a coherent narrative that spans multiple sentences?'

Climate change has been bleaching coral reefs, decimating the local marine species that call them home, since at least the first major observations were recorded in the Caribbean in 1980. Thankfully, new A.I. cataloguing designed to identify the geographic regions where coral is still thriving hopes to reverse the trend, saving some of the world's most dense and varied aquatic ecosystems from all-but-certain extinction. There are numerous reasons why we need to care about saving coral reefs, from the ethical to the economic. In addition to housing about a quarter of marine species, these reefs provide $375 billion USD in revenue to the world economy, according to the Guardian, and food security to half a billion people. Without them, researchers say countless species and the entire ocean fishing industry that depends on them would simply evaporate.

Neural networks with ReLU activations have achieved great empirical success in various domains. However, existing results for learning ReLU networks either pose assumptions on the underlying data distribution being e.g. Gaussian, or require the network size and/or training size to be sufficiently large. In this context, the problem of learning a two-layer ReLU network is approached in a binary classification setting, where the data are linearly separable and a hinge loss criterion is adopted. Leveraging the power of random noise, this contribution presents a novel stochastic gradient descent (SGD) algorithm, which can provably train any single-hidden-layer ReLU network to attain global optimality, despite the presence of infinitely many bad local minima and saddle points in general. This result is the first of its kind, requiring no assumptions on the data distribution, training/network size, or initialization. Convergence of the resultant iterative algorithm to a global minimum is analyzed by establishing both an upper bound and a lower bound on the number of effective (non-zero) updates to be performed. Furthermore, generalization guarantees are developed for ReLU networks trained with the novel SGD. These guarantees highlight a fundamental difference (at least in the worst case) between learning a ReLU network as well as a leaky ReLU network in terms of sample complexity. Numerical tests using synthetic data and real images validate the effectiveness of the algorithm and the practical merits of the theory.

Optical coherence tomography (OCT) based measurements of retinal layer thickness, such as the retinal nerve fibre layer (RNFL) and the ganglion cell with inner plexiform layer (GCIPL) are commonly used for the diagnosis and monitoring of glaucoma. Previously, machine learning techniques have relied on segmentation-based imaging features such as the peripapillary RNFL thickness and the cup-to-disc ratio. Here, we propose a deep learning technique that classifies eyes as healthy or glaucomatous directly from raw, unsegmented OCT volumes of the optic nerve head (ONH) using a 3D Convolutional Neural Network (CNN). We compared the accuracy of this technique with various feature-based machine learning algorithms and demonstrated the superiority of the proposed deep learning based method. Logistic regression was found to be the best performing classical machine learning technique with an AUC of 0.89. In direct comparison, the deep learning approach achieved a substantially higher AUC of 0.94 with the additional advantage of providing insight into which regions of an OCT volume are important for glaucoma detection. Computing Class Activation Maps (CAM), we found that the CNN identified neuroretinal rim and optic disc cupping as well as the lamina cribrosa (LC) and its surrounding areas as the regions significantly associated with the glaucoma classification. These regions anatomically correspond to the well established and commonly used clinical markers for glaucoma diagnosis such as increased cup volume, cup diameter, and neuroretinal rim thinning at the superior and inferior segments.

Designing a reliable natural language (NL) interface for querying tables has been a longtime goal of researchers in both the data management and natural language processing (NLP) communities. Such an interface receives as input an NL question, translates it into a formal query, executes the query and returns the results. Errors in the translation process are not uncommon, and users typically struggle to understand whether their query has been mapped correctly. We address this problem by explaining the obtained formal queries to non-expert users. Two methods for query explanations are presented: the first translates queries into NL, while the second method provides a graphic representation of the query cell-based provenance (in its execution on a given table). Our solution augments a state-of-the-art NL interface over web tables, enhancing it in both its training and deployment phase. Experiments, including a user study conducted on Amazon Mechanical Turk, show our solution to improve both the correctness and reliability of an NL interface.

A recent scientific survey off the coast of Sulawesi Island in Indonesia suggests that some shallow water corals may be less vulnerable to global warming than previously thought. Between 2014 and 2017, the world's reefs endured the worst coral bleaching event in history, as the cyclical El Niño climate event combined with anthropogenic warming to cause unprecedented increases in water temperature. But the June survey, funded by Microsoft co-founder Paul Allen's family foundation, found the Sulawesi reefs were surprisingly healthy. In fact they were in better condition than when they were originally surveyed in 2014 - a surprise for British scientist Dr Emma Kennedy, who led the research team. "After several depressing years as a coral reef scientist, witnessing the worst-ever global coral bleaching event, it is unbelievably encouraging to experience reefs such as these," she said.

One year ago, we wrote about the world's first robot lawyer. It is a website with a chatbot that started off with a single and free legal service: helping to appeal unfair parking tickets. When the article was published, the services was available in the UK, and in New York and Seattle. At the time, it had helped overturn traffic tickets to the value of 4 million dollars. Apart from appealing parking tickets, the website could already assist you, too, in claiming compensation if your flight was delayed.

I recently had the opportunity to travel to South Korea to look over LG's work in both the AI and robotics fields, including some detailed time with its LG CLOi Airport Guide Robot. That's a design that LG has iterated on over time, and I had the chance to sit down for an interview (via a translator) with Hyungjn Choi, LG's Leader of Life support Robot Biz. That's a fancy title to say that he's in charge (in his own words) "of robot business development and product planning" at LG. Robots in industry are nothing new, but people-centric robots are a tough challenge. Mr Choi is quite clear that the first robot was the toughest. "Technically speaking, the most difficult one is the first one that you can see when you arrive (at Seoul's Incheon International Airport), the Airport guide robot. Because we have to deal with a spacious indoor area, the environment is very noisy. There are a lot of people travelling around. The robot needs to autonomously drive around and move around these high traffic areas. Also it needs to interpret and understand different and diverse pronunciations and ways to respond to customers' needs. That was the most technologically challenging area, and in order to take on that challenge, we've specifically taken on airports. But that doesn't mean that other robots were easy, technologically! So for example, the lawnmower robot is different from other lawnmower robots you might see. Our one moves around in a specific pattern, so it can move around in a zig zag motion, so even if your ground is bumpy, it can make a very calculated movement to ensure that every part of your lawn is well managed."

Digital transformation is driving a new breed of retail that will take consumers where they've never been before, writes Microsoft's Christi Olson. The following is a guest post by Christi Olson, head of evangelism for search at Microsoft. The views are the author's own. Like many industries today, retail is in an extreme state of disruption. In five years, 25% of malls will be gone.