Over the last few weeks, Computer Weekly has looked at how a number of organisations are combining automation and artificial intelligence (AI) to deliver measurable business benefits. During the Microsoft Future Decoded event in London, the use of Microsoft tools, and Thoughtonomy's intelligent automation platform at East Suffolk and North Essex NHS Foundation Trust, were used to demonstrate how AI and automation can combine to deliver benefits in the public sector. In logistics, Canandian transportation company Polaris Transportation is using AI and automation in a project to streamline the handling of scanned-in customs paperwork, enabling it to reduce many hours of manual work. The company used the WorkFusion intelligent automation platform to scan and "read" customs paperwork associated with cross-border shipping documents. According to Cindy Rose, CEO of Microsoft UK, more advanced organisations are accelerating their use of AI, which has enabled them to see its benefits on their bottom line.

The Sino-American relationship has been quite a roller coaster this year, courtesy of the belligerent occupant of the White House. With its technical and operational superiority in 5G mobile networks (the vital infrastructure for technologies like A.I. and the Internet of Things), Huawei might be an avatar for China itself: ambitious, future-focused, and a serious threat to U.S. exceptionalism. The deluge of American complaints about Huawei being a national security risk (due to its links to the Chinese state) should be recognized for what it is: cover for the United States to engage in economic warfare, throwing its significant weight around to help ensure Huawei is blacklisted across the globe. The desperation of the U.S. efforts reflects a cold truth about the international competition in technology, and A.I. in particular: China is opening up a lead. Following their public commitment in 2017 to develop world leadership in A.I. by 2030, China has backed up its strategy with several billion dollars' worth of funding and a cohesive bureaucratic effort to manage the plan's execution.

Adversarial training is by far the most successful strategy for improving robustness of neural networks to adversarial attacks. Despite its success as a defense mechanism, adversarial training fails to generalize well to unperturbed test set. We hypothesize that this poor generalization is a consequence of adversarial training with uniform perturbation radius around every training sample. Samples close to decision boundary can be morphed into a different class under a small perturbation budget, and enforcing large margins around these samples produce poor decision boundaries that generalize poorly. Motivated by this hypothesis, we propose instance adaptive adversarial training -- a technique that enforces sample-specific perturbation margins around every training sample. We show that using our approach, test accuracy on unperturbed samples improve with a marginal drop in robustness. Extensive experiments on CIFAR-10, CIFAR-100 and Imagenet datasets demonstrate the effectiveness of our proposed approach.

Mini-batch stochastic gradient descent (SGD) approximates the gradient of an objective function with the average gradient of some batch of constant size. While small batch sizes can yield high-variance gradient estimates that prevent the model from learning a good model, large batches may require more data and computational effort. This work presents a method to change the batch size adaptively with model quality. We show that our method requires the same number of model updates as full-batch gradient descent while requiring the same total number of gradient computations as SGD. While this method requires evaluating the objective function, we present a passive approximation that eliminates this constraint and improves computational efficiency. We provide extensive experiments illustrating that our methods require far fewer model updates without increasing the total amount of computation.

The use of technologies like artificial intelligence (AI) in government services has increased in the past few years. Governments are leveraging these technologies to achieve sustainable and inclusive socio-economic development. Governments have already been using technologies like AI, machine learning (ML), and big data to improve citizen services and overall administration. AI provides a wide array of uses in government services. Applications of AI in government services range from emergency response, healthcare, and defence to providing a platform to connect and interact with the common people.

The annual cost of cybercrime is estimated to rise to $6 trillion by 2021.[1] Artificial intelligence (AI), frequently mentioned for its potential to accelerate innovation, boost performance and improve decision-making, is already being applied to defend against cybercrime. Because AI works well with functions that use massive amounts of data and require analysis and judgment, integrating AI-based cybersecurity technology with other defenses is a natural choice for cybersecurity professionals. Today, AI is used more extensively in cybersecurity than in any other function, with 75% of companies using AI technology to detect and ward off cyberthreats, according to the results of a recent global executive survey on AI conducted by Protiviti. Cybersecurity usage of AI is expected to grow nearly 20% by 2021.[2] AI's significant and compelling benefits come with new risks that need to be managed.

NHS Digital wanted to make it easier for users to research and access published NHS health data. To achieve that, the agency partnered with IPsoft to provide users with their own data concierge whom they call ViDA (or Virtual Digital Assistant). ViDA is an always-on conversational agent based on our industry-leading digital colleague, Amelia. Users simply tell ViDA what information they are attempting to locate using everyday language, and ViDA can take it from there. You can read more about the project in detail here from our Cognitive Project Lead for UK Healthcare, David King.

In February 2019 Polish government added an amendment to a banking law that gives a customer a right to receive an explanation in case of a negative credit decision. This means that a bank needs to be able to explain why the loan wasn't granted if the decision process was automatic.

"For most real problems, there aren't perfect answers," writes Thomas W. Malone. "But when they are connected in the right ways, groups of people and computers together can often get closer to perfect intelligence than either could alone." Malone, who is the Patrick J. McGovern Professor of Management at the MIT Sloan School of Management and the founding director of the MIT Center for Collective Intelligence, explores the potential of such connections in his new book, Superminds: The Surprising Power of People and Computers Thinking Together, from which this excerpt is taken. Will general AI be a form of collective intelligence? We know that the human brain is itself a form of collective intelligence.

Artificial Intelligence was once the stuff of science fiction. Now it's here, and every publication from the Washington Post to Wired to the Wall Street Journal is full of articles and videos exploring it. Depending on whom you listen to, AI will be a job killer or a job creator; a tool to boost productivity or Skynet from the Terminator movies; a technology that will dramatically transform society or an overhyped nothingburger. To help prepare for this uncertain and potentially disturbing future, Gov. Phil Scott impaneled an Artificial Intelligence Task Force in 2018. Its mandate: to "investigate the field of artificial intelligence" in the state and make recommendations for how the technology can be responsibly applied in Vermont's economy and government.