Neural network-based systems can now learn to locate the referents of words and phrases in images, answer questions about visual scenes, and execute symbolic instructions as first-person actors in partially-observable worlds. To achieve this so-called grounded language learning, models must overcome challenges that infants face when learning their first words. While it is notable that models with no meaningful prior knowledge overcome these obstacles, researchers currently lack a clear understanding of how they do so, a problem that we attempt to address in this paper. For maximum control and generality, we focus on a simple neural network-based language learning agent, trained via policy-gradient methods, which can interpret single-word instructions in a simulated 3D world. Whilst the goal is not to explicitly model infant word learning, we take inspiration from experimental paradigms in developmental psychology and apply some of these to the artificial agent, exploring the conditions under which established human biases and learning effects emerge. We further propose a novel method for visualising semantic representations in the agent.

I keep hearing excuses for not working on difficult problems: "Eventually AI will solve this so there's no point working on it now." First, we should be cautious about putting too much expectation on artificial intelligence which, by most metrics, is really today's machine learning (ML) and neural networks. There is no doubt these systems have produced truly remarkable results. The GO game story of AlphaGo and AlphaZero from Deep Mind is by now a classic example of the surprisingly powerful results these systems produce. A chess-playing version of AlphaZero learned quickly and demonstrated choices of moves unlike those of traditional chess players.

Sign in to report inappropriate content. Patrick Ehlen, Chief Scientist at Loop AI Labs Cognitive Computing, explains what is the next frontier to better approximate human capacity with quantum probability theory.

Plenty has been written about deep learning frameworks such as Keras and PyTorch, and how powerful yet simple to use they are for constructing and playing with wonderful deep learning models. There are so many tutorials/articles already written about model architecture and optimizers-- the concept of convolution, max pooling, optimizers such as ADAM or RMSprop. What if, all you wanted, is a single function to pull automatically images from a specified directory on your disk, and give you back a fully trained neural net model, ready to be used for prediction? Therefore, in this article, we focus on how to use a couple of utility methods from the Keras (TensorFlow) API to streamline the training of such models (specifically for a classification task) with a proper data pre-processing. In the end, we aim to write a single utility function, which can take just the name of your folder where training images are stored, and give you back a fully trained CNN model.

As organizations look to AI to solve their business problems, they want a trusted partner that offers analytics expertise and domain knowledge. AI has been an integral part of SAS software for years. We continue to move beyond the hype to make AI real for our customers, building on our strong foundation in advanced analytics. From machine learning, deep learning, computer vision and natural language processing to forecasting and optimization, SAS helps customers in many industries benefit from AI and the intelligence it creates.

This month, the NHS signed its first deals with Google. Five NHS trusts have agreed contracts with Google Health, after it swallowed up its UK sister firm DeepMind Health, nearly a year after signalling its intention to do so. New Scientist first revealed the extent of DeepMind's access to the sensitive data of more than a million National Health Service patients back in 2016, in a deal that the UK's data watchdog later found breached the law. The partnership has yielded interesting research, including using artificial intelligence to detect eye disease from scans with an accuracy that matches or exceeds human experts. But is there a material difference now the deals are with the US tech giant rather than DeepMind, and should people who use the NHS be concerned at the change?

The healthcare industry has always been a leader when it comes to innovation. The constant mutating of diseases and viruses make for a difficult industry to stay ahead of the curve, but with the help of artificial intelligence and machine learning algorithms, it continues to advance with creating new treatments and helping people live longer and healthier lives. A study released this week by The Lancet Digital Health examined all the data between 2012 and 2019 of testing that has been involved with artificial intelligence and deep learning in the discovery of disease diagnosis through medical imaging. The study found that AI has become more accurate in the past years of identifying disease diagnosis in these images and has become a more viable source of diagnostic information. According to Lancet Digital, out of 14 studies conducted, the systems were able to correctly identify disease 87% of the time while healthcare professionals were correct 86% of the time.

Time is deeply woven into how people perceive, and communicate about the world. Almost unconsciously, we provide our language utterances with temporal cues, like verb tenses, and we can hardly produce sentences without such cues. Extracting temporal cues from text, and constructing a global temporal view about the order of described events is a major challenge of automatic natural language understanding. Temporal reasoning, the process of combining different temporal cues into a coherent temporal view, plays a central role in temporal information extraction. This article presents a comprehensive survey of the research from the past decades on temporal reasoning for automatic temporal information extraction from text, providing a case study on how combining symbolic reasoning with machine learning-based information extraction systems can improve performance. It gives a clear overview of the used methodologies for temporal reasoning, and explains how temporal reasoning can be, and has been successfully integrated into temporal information extraction systems. Based on the distillation of existing work, this survey also suggests currently unexplored research areas. We argue that the level of temporal reasoning that current systems use is still incomplete for the full task of temporal information extraction, and that a deeper understanding of how the various types of temporal information can be integrated into temporal reasoning is required to drive future research in this area.

In this paper, we propose a novel model RevGAN that automatically generates controllable and personalized user reviews based on the arbitrarily given sentimental and stylistic information. RevGAN utilizes the combination of three novel components, including self-attentive recursive autoencoders, conditional discriminators, and personalized decoders. We test its performance on the several real-world datasets, where our model significantly outperforms state-of-the-art generation models in terms of sentence quality, coherence, personalization and human evaluations. We also empirically show that the generated reviews could not be easily distinguished from the organically produced reviews and that they follow the same statistical linguistics laws.

This paper presents a family of backpropagation-free neural architectures, Gated Linear Networks (GLNs),that are well suited to online learning applications where sample efficiency is of paramount importance. The impressive empirical performance of these architectures has long been known within the data compression community, but a theoretically satisfying explanation as to how and why they perform so well has proven difficult. What distinguishes these architectures from other neural systems is the distributed and local nature of their credit assignment mechanism; each neuron directly predicts the target and has its own set of hard-gated weights that are locally adapted via online convex optimization. By providing an interpretation, generalization and subsequent theoretical analysis, we show that sufficiently large GLNs are universal in a strong sense: not only can they model any compactly supported, continuous density function to arbitrary accuracy, but that any choice of no-regret online convex optimization technique will provably converge to the correct solution with enough data. Empirically we show a collection of single-pass learning results on established machine learning benchmarks that are competitive with results obtained with general purpose batch learning techniques.