Deep neural networks (DNN) are powerful models for many pattern recognition tasks, yet their high computational complexity and memory requirement limit them to applications on high-performance computing platforms. In this paper, we propose a new method to evaluate DNNs trained with 32bit floating point (float32) accuracy using only low precision integer arithmetics in combination with binary shift and clipping operations. Because hardware implementation of these operations is much simpler than high precision floating point calculation, our method can be used for an efficient DNN inference on dedicated hardware. In experiments on MNIST, we demonstrate that DNNs trained with float32 can be evaluated using a combination of 2bit integer arithmetics and a few float32 calculations in each layer or only 3bit integer arithmetics in combination with binary shift and clipping without significant performance degradation.

Program synthesis from natural language (NL) is practical for humans and, once technically feasible, would significantly facilitate software development and revolutionize end-user programming. The proposed architecture relies exclusively on neural components, and is built upon a tree2tree autoencoder trained on abstract syntax trees, combined with a pretrained word embedding and a bidirectional multi-layer LSTM for NL processing. The decoder features a doubly-recurrent LSTM with a novel signal propagation scheme and soft attention mechanism. When applied to a large dataset of problems proposed in a previous study, SAPS performs on par with or better than the method proposed there, producing correct programs in over 90% of cases. In contrast to other methods, it does not involve any non-neural components to post-process the resulting programs, and uses a fixed-dimensional latent representation as the only link between the NL analyzer and source code generator. Program synthesis consists in automatic or semiautomatic (e.g., interactive) generation of programs (or other executable structures) from specifications. This task can be posed in several ways. It is most common to assume that specification has the form of input-output pairs (tests), in which case synthesis resembles learning from examples.

Restricted Boltzmann machines (RBMs) and their extensions, called 'deep-belief networks', are powerful neural networks that have found applications in the fields of machine learning and artificial intelligence. The standard way to training these models resorts to an iterative unsupervised procedure based on Gibbs sampling, called 'contrastive divergence' (CD), and additional supervised tuning via back-propagation. However, this procedure has been shown not to follow any gradient and can lead to suboptimal solutions. In this paper, we show an efficient alternative to CD by means of simulations of digital memcomputing machines (DMMs). We test our approach on pattern recognition using a modified version of the MNIST data set. DMMs sample effectively the vast phase space given by the model distribution of the RBM, and provide a very good approximation close to the optimum. This efficient search significantly reduces the number of pretraining iterations necessary to achieve a given level of accuracy, as well as a total performance gain over CD. In fact, the acceleration of pretraining achieved by simulating DMMs is comparable to, in number of iterations, the recently reported hardware application of the quantum annealing method on the same network and data set. Notably, however, DMMs perform far better than the reported quantum annealing results in terms of quality of the training. We also compare our method to advances in supervised training, like batch-normalization and rectifiers, that work to reduce the advantage of pretraining. We find that the memcomputing method still maintains a quality advantage ($>1\%$ in accuracy, and a $20\%$ reduction in error rate) over these approaches. Furthermore, our method is agnostic about the connectivity of the network. Therefore, it can be extended to train full Boltzmann machines, and even deep networks at once.

In a quiet corner of rural Hampshire, a robot called Rachel is pootling around an overgrown field. With bright orange casing and a smartphone clipped to her back end, she looks like a cross between an expensive toy and the kind of rover used on space missions. Up close, she has four USB ports, a disc-like GPS receiver, and the nuts and bolts of a system called Lidar, which enables her to orient herself using laser beams. She cost around £2,000 to make. Every three seconds, Rachel takes a closeup photograph of the plants and soil around her, which will build into a forensic map of the field and the wider farm beyond. After 20 minutes or so of this, she is momentarily disturbed by two of the farm's dogs, unsure what to make of her.

These four new solution accelerators help financial services and insurance firms solve complex business challenges by discovering meaningful relationships between events that impact one another (correlation) and cause a future event to happen (causation). Following the success of Synechron's AI Automation Program – Neo, Synechron's AI Data Science experts have developed a powerful set of accelerators that allow financial firms to address business challenges related to investment research generation, predicting the next best action to take with a wealth management client, high-priority customer complaints, and better predicting credit risk related to mortgage lending. The Accelerators combine Natural Language Processing (NLP), Deep Learning algorithms and Data Science to solve the complex business challenges and rely on a powerful Spark and Hadoop platform to ingest and run correlations across massive amounts of data to test hypotheses and predict future outcomes. The Data Science Accelerators are the fifth Accelerator program Synechron has launched in the last two years through its Financial Innovation Labs (FinLabs), which are operating in 11 key global financial markets across North America, Europe, Middle East and APAC; including: New York, Charlotte, Fort Lauderdale, London, Paris, Amsterdam, Serbia, Dubai, Pune, Bangalore and Hyderabad. With this, Synechron's Global Accelerator programs now includes over 50 Accelerators for: Blockchain, AI Automation, InsurTech, RegTech, and AI Data Science and a dedicated team of over 300 employees globally.

Typically, deep learning approaches to voice recognition -- systems that employ layers of neuron-mimicking mathematical functions to parse human speech -- lean on powerful remote servers for bulk of processing. But researchers at the University of Waterloo and startup DarwinAI claim to have pioneered a strategy for designing speech recognition networks that not only achieves state-of-the-art accuracy, but which produces models robust enough to run on low-end smartphones. They describe their method in a paper published on the preprint server Arxiv.org It builds on work by Amazon's Alexa Machine Learning team, which earlier this year developed navigation, temperature control, and music playback algorithms that can be performed locally; Qualcomm, which in May claimed to have created on-device voice recognition models that are 95 percent accurate; Dublin, Ireland startup Voysis, which in September announced an offline WaveNet voice model for mobile devices; and Intel. "In this study, we explore a human-machine collaborative design strategy for building low-footprint [deep neural network] architectures for speech recognition through a marriage of human-driven principled network design prototyping and machine-driven design exploration," the researchers wrote.

Elon Musk is engaged in a bizarre fight with Fortnite fans after he appeared to call them "virgins". The SpaceX and Tesla boss – who has been repeatedly criticised over his tweets – posted a picture of a fake news article that seemed to suggest fans of the video game are "virgins". "Elon Musk buys Fortnite and deletes it," the hoax news story, shared on Mr Musk's Twitter, read. It claimed that the billionaire had said he had to remove the game to protect players from "eternal virginity". The I.F.O. is fuelled by eight electric engines, which is able to push the flying object to an estimated top speed of about 120mph.

Doctors could use artificial intelligence to diagnose dementia more accurately and give better treatment, scientists say. Researchers have invented a computer algorithm which can analyse MRI brain scans and learn how to recognise different types of dementia. They say that although many types of the brain-destroying condition have similar symptoms, they respond differently to treatment. Being able to correctly identify which type someone has means patients could be helped earlier on in their illness or given more targeted therapy. Experts say the research is'pioneering' and has'huge potential' in the future of treating dementia, expected to affect one million Britons by 2025.

Artificial intelligence is adapting to protect children from the perils of cyber-bullying and social media. Researchers trained a machine-learning algorithm to detect bullying posts on social media and hide them from view. The AI detected words and phrases in this dataset that were typically associated with bullying and filtered out more than two-thirds of threats, insults and instances of sexual harassment. Gilles Jacobs at Ghent University in Belgium built the programme and tasked it with filtering real-life posts from AskFM. A team of professional linguists went through the same set of data and picked out the offensive posts from almost 200,000 posts.

Just this year, researchers at Columbia created a self-replicating neural network that can predict its future growth path -- not unlike a human planning their career and learning new skills. Even if complex programming projects will still require humans, there's a chance that database experts and lower-level, AI-related jobs will be phased out. Microsoft and the University of Cambridge recently released an algorithm that could solve simple equations, such as Excel formulas. Uniquely, for such a compact program, it could augment its abilities by using the brute force approach on a smaller scale: trying different chunks of code until it finds the winning solution.