Knowledgeable people typically judge their abilities accurately, but the incompetent routinely overestimate their skill. This all-too-familiar glitch in our mental makeup is called the Dunning-Kruger effect. It's a failure of metacognition: the ability to gauge what you know and what you don't. Research suggests bees do not have this flaw. In a recent experiment led by Andrew Barron, associate professor in biological sciences at Macquarie University in Australia, researchers trained honeybees to determine which of two horizontal lines was above or below the other.

The microscope is mainly used for imaging applications to analyze terabytes of data per day. These applications can profit by late advances in computer vision and profound learning. Now, in collaboration with robotic microscopy applications, Google engineers have assembled high-quality image datasets that separate signal from noise. In "Assessing Microscope Image Focus Quality with Deep Learning", researchers trained a deep neural network to rate the focus quality of microscopy images with higher accuracy than previous methods. They added the pre-trained TensorFlow model with plugins in Fiji (ImageJ) and CellProfiler, two leading open source scientific image analysis tools to use with the graphical user interface or invoked via scripts.

We contribute to recent efforts in relating two approaches to automatic synthesis, namely, automated planning and discrete reactive synthesis. First, we develop a declarative characterization of the standard "fairness" assumption on environments in non-deterministic planning, and show that strong-cyclic plans are correct solution concepts for fair environments. This complements, and arguably completes, the existing foundational work on non-deterministic planning, which focuses on characterizing (and computing) plans enjoying special "structural" properties, namely loopy but closed policy structures. Second, we provide an encoding suitable for reactive synthesis that avoids the naive exponential state space blowup. To do so, special care has to be taken to specify the fairness assumption on the environment in a succinct manner.

In this work, we investigate Batch Normalization technique and propose its probabilistic interpretation. We propose a probabilistic model and show that Batch Normalization maximazes the lower bound of its marginalized log-likelihood. Then, according to the new probabilistic model, we design an algorithm which acts consistently during train and test. However, inference becomes computationally inefficient. To reduce memory and computational cost, we propose Stochastic Batch Normalization -- an efficient approximation of proper inference procedure. This method provides us with a scalable uncertainty estimation technique. We demonstrate the performance of Stochastic Batch Normalization on popular architectures (including deep convolutional architectures: VGG-like and ResNets) for MNIST and CIFAR-10 datasets.

Even todays most advanced machine learning models are easily fooled by almost imperceptible perturbations of their inputs. Foolbox is a new Python package to generate such adversarial perturbations and to quantify and compare the robustness of machine learning models. It is build around the idea that the most comparable robustness measure is the minimum perturbation needed to craft an adversarial example. To this end, Foolbox provides reference implementations of most published adversarial attack methods alongside some new ones, all of which perform internal hyperparameter tuning to find the minimum adversarial perturbation. Additionally, Foolbox interfaces with most popular deep learning frameworks such as PyTorch, Keras, TensorFlow, Theano and MXNet and allows different adversarial criteria such as targeted misclassification and top-k misclassification as well as different distance measures. The code is licensed under the MIT license and is openly available at https://github.com/bethgelab/foolbox . The most up-to-date documentation can be found at http://foolbox.readthedocs.io .

Networks of coupled dynamical systems provide a powerful way to model systems with enormously complex dynamics, such as the human brain. Control of synchronization in such networked systems has far reaching applications in many domains, including engineering and medicine. In this paper, we formulate the synchronization control in dynamical systems as an optimization problem and present a multi-objective genetic programming-based approach to infer optimal control functions that drive the system from a synchronized to a non-synchronized state and vice-versa. The genetic programming-based controller allows learning optimal control functions in an interpretable symbolic form. The effectiveness of the proposed approach is demonstrated in controlling synchronization in coupled oscillator systems linked in networks of increasing order complexity, ranging from a simple coupled oscillator system to a hierarchical network of coupled oscillators. The results show that the proposed method can learn highly-effective and interpretable control functions for such systems.

Many people see the potential benefits of artificial intelligence technologies used for government services – but many also aren't convinced governments will use AI tech responsibly, according to a new survey from Accenture. The online survey of more than 6,000 citizens from US, Australia, the UK, Singapore, France and Germany found that more than half (54%) of citizens said they are willing to use AI services delivered by government, with even more expressing willingness when presented with the potential benefits derived from artificial intelligence. For instance, three-quarters (74%) of respondents said they would be willing to use artificial intelligence if it would increase pension or retirement income (such as by improving their personal investment strategy and/or pension scheme), and two-thirds (66%) said they would use a chatbot if it would guarantee faster processing of a tax refund or social service benefits. However, that doesn't mean citizens aren't worried about the government using artificial intelligence responsibly – two-thirds (66%) of respondents indicated a lack of confidence in government's ethical and responsible use of AI. Specifically, only one-third (34%) said they're "confident/very confident" that government would be ethical and responsible in its use of AI; fewer than one in three (29%) said they are "not at all confident" in government using AI ethically and responsibly, and slightly more than one-third (37%) said they are neutral on the point. The survey also determined that regardless of where they lived, citizens have concerns about the use of artificial intelligence in government, including in areas of job security and personal data security.

Tinder has introduced a new feature that it claims will help you come up with a better chat-up line. 'Feed' is a timeline that shows the activity of all your past matches, including new photo they have uploaded and changes to their favourite songs on Spotify. Tinder says the new feature is'an exciting new way to see more of what someone is all about by giving you a true glimpse into their world.' But many users claim the service is'creepy' and allows their matches to'stalk' them online without them realising. The new'Feed' tab shows all updates that matches make to their profile such as new photos (left) and even what they are listening to on Spotify (right).

Wow, it's been a super busy month, as shown by the size of this month's newsletter! Visiting New Zealand for last month's Digital Nations conference, we welcomed Tomáš Ižo, Engineering Director at the Google Research Machine Perception Group. Tomáš presented to a full house in Wellington with a wide ranging talk covering technical and societal issues associated with AI. This was followed by a lively Q&A session too. No doubt, as you will have seen in the media, the team have been going all out preparing for AI-DAY 2018 on Wednesday 28 March, which is less than two weeks away now! We have been busy talking to some of the key speakers at the conference.

A fresh wave of artificial intelligence rolling through Microsoft's language translation technologies is bringing more accurate speech recognition to more of the world's languages and higher quality machine-powered translations to all 60 languages supported by Microsoft's translation technologies. The advances were announced at Microsoft Tech Summit Sydney in Australia on November 16. "We've got a complex machine, and we're innovating on all fronts," said Olivier Fontana, the director of product strategy for Microsoft Translator, a platform for text and speech translation services. As the wave spreads, he added, these machine translation tools are allowing more people to grow businesses, build relationships and experience different cultures. Microsoft's research labs around the world are also building on top of these technologies to help people learn how to speak new languages, including a language learning application for non-native speakers of Chinese that also was announced at this week's tech summit. The new Microsoft Translator advances build on last year's switch to deep neural network-powered machine translations, which offer more fluent, human-sounding translations than the predecessor technology known as statistical machine translation.