Pixar founder George Lucas might have given the world the word "Droid" before he sold the company to Apple's Steve Jobs but now Apple CEO Tim Cook is making no secret of the fact that he wants the company to play its part in bringing Artificial Intelligence to the world. This week he's been spilling at least some of the beans on Apple's plans for AI, although his vision seems more connected to machine learning rather than the AI technologies that would help to bring droids like C3PO and R2D2. "We see AI as being horizontal in nature and running across all our platforms and products," said Cook, following on to say, "we see it being used in ways that most people don't even think about." That sounds grandoise – and, at first sight, it sounds like they have something amazing up their sleeves. So just what kind of unexpected use cases was he talking about?

Sketch a picture of a cat, or a castle, or a crab and most people will get what you're trying to convey -- but unless you have a little talent, the drawing probably doesn't look a lot like the real thing. That's not a problem for this system created by Belgian computer scientists. Their system can recognize toddler-level sketches of 250 categories of objects. This has been done a couple of times before, but one interesting aspect of this approach is that the machine learning system is exposed to the drawing as it's created, seeing it at various fractions of completeness. Turns out that can help identify the object; after all, you ever see anyone draw the chimney on the house first?

Despite its potentially widespread applications, creating a speech recognition software capable of cutting through the nuances and variations in the spoken word has been a task fraught with patchy success, at best. The aim many companies have striven toward is to create a software that can recognize the words in a conversation as well as a human would -- a key requisite for a truly immersive artificial intelligence experience. In a major breakthrough in this endeavor, Microsoft announced Tuesday that it had created a technology that enabled speech recognition systems to transcribe a conversation with the same error rate as their human counterparts. "We've reached human parity," Xuedong Huang, Microsoft's chief speech scientist, said in a statement. "This is an historic achievement."

We now have a deep learning model that is able to deliver valuable results, but how can we apply it easily to new data where and when we need to?

Along with taking part in the social experiment, Gemma will front the factual programme and delve into the advances that have been made into artificial intelligence. She will look at technology - such as driverless cars - and speak to British A.I. researcher Demis Hassabis and his DeepMind project, which is working toward creating machines that can learn even more by themsleves than ever before. "This film pushes the boundaries of what is possible using the technology that is increasingly influential in our lives," said Tom Porter, Channel 4's acting Commissioning Editor, Science.

DENSO Corporation and Toshiba Corporation have reached a basic agreement to jointly develop an artificial intelligence technology called Deep Neural Network-Intellectual Property (DNN-IP), which will be used in image recognition systems which have been independently developed by the two companies to help achieve advanced driver assistance and automated driving technologies. This Smart News Release features multimedia. DNN, an algorithm modeled after the neural networks of the human brain, is expected to perform recognition processing as accurately as, or even better than the human brain. To achieve automated driving, automotive computers need to be able to identify different road traffic situations including a variety of obstacles and road markings, availability of road space for driving, and potentially dangerous situations. In image recognition based on conventional pattern recognition and machine learning, objects that need to be recognized by computers must be characterized and extracted in advance.

Analysis Research in AI is expanding quickly, and the UK and US governments have begun to notice. Official reports about the new technology and future strategies were dropped by both governments this month. Blighty's Science and Technology Committee released Robotics and Artificial Intelligence, while the White House delivered Preparing for the Future of Artificial Intelligence and National Artificial Intelligence Research and Development Strategic Plan. The titles of the British and American reports provide a clue as to how both governments are responding. There is no "preparing" or "strategic plan" in the UK's reports.

Bright Computing, the leading provider of hardware-agnostic cluster and cloud management software, today announced that it will be exhibiting at Huawei Connect, Paris, October 20-21, 2016. Huawei Connect, a conference for Huawei's European ICT ecosystem, focuses on business innovation and open partnerships. This year's theme is "Shape the Cloud", and policy makers, industry leaders, academics and technology elites will gather together to share, discuss and debate the future technologies and new business models that are driving the world's digital transformation. At the event, Bright Computing will showcase its latest solution, Bright for Deep Learning, which makes it easy to build an enterprise-grade deep learning environment, quickly and efficiently, enabling organizations to focus on gaining actionable insights from rich, complex data. Bright will explain how it helps to find, configure, and deploy all of the dependent pieces needed to run deep learning libraries and frameworks, in order to gain advantage from the deep learning evolution.

Meta-learning consists in learning learning algorithms. We use a Long Short Term Memory (LSTM) based network to learn to compute on-line updates of the parameters of another neural network. These parameters are stored in the cell state of the LSTM. Our framework allows to compare learned algorithms to hand-made algorithms within the traditional train and test methodology. In an experiment, we learn a learning algorithm for a one-hidden layer Multi-Layer Perceptron (MLP) on non-linearly separable datasets. The learned algorithm is able to update parameters of both layers and generalise well on similar datasets.

A major hurdle to clinical translation of brain-machine interfaces (BMIs) is that current decoders, which are trained from a small quantity of recent data, become ineffective when neural recording conditions subsequently change. We tested whether a decoder could be made more robust to future neural variability by training it to handle a variety of recording conditions sampled from months of previously collected data as well as synthetic training data perturbations. We developed a new multiplicative recurrent neural network BMI decoder that successfully learned a large variety of neural-to- kinematic mappings and became more robust with larger training datasets. When tested with a non-human primate preclinical BMI model, this decoder was robust under conditions that disabled a state-of-the-art Kalman filter based decoder. These results validate a new BMI strategy in which accumulated data history is effectively harnessed, and may facilitate reliable daily BMI use by reducing decoder retraining downtime.