Artificial Intelligence (AI) is proving its powers to prevent and detect everything gripping them from routine employee theft, frauds, insider trading and business risks. Many large corporations, business enterprises have been employing AI to detect and prevent money laundering and widespread frauds. Machine learning has been increasingly deployed by social media platforms to block illicit content such as child pornography and fake news. Businesses have been using AI for higher risk management and responsive fraud detection towards prevention and prediction of crimes. The earlier monitoring systems used by the industries need manual interference and are often not cent percent accurate.

Note: This is part-4 of a series of articles on'Security and Privacy in Artificial Intelligence & Machine Learning'. In this article we will take a closer look at use of AI&ML in various security-related use cases. We will cover not only cybersecurity but also some general security scenarios and how solutions based on AI&ML are becoming increasingly prevalent in all such areas. Towards the end we will also explore ways that attackers are likely to circumvent these security techniques. So let us begin with a look at some interesting areas where security features are benefiting from ML&AI. When we consider cybersecurity, one of the most common areas where AI&ML gets a mention is addressing the'needle in the haystack' problem in security of a large scale environment.

Despite the emergence of apps like Slack and Yammer for internal employee communication, email is still the dominant form of external employee communication for enterprises. "In a similar way that computers, servers and devices communicate with one another through data packets transmitted via TCP/IP, employees communicate with one another through natural language and documents shared via email," says Bishop. Why are account takeovers on the rise? And how can businesses prevent this method of attack? Asaf Cidon, from Barracuda Networks, helps Information Age answer these questions. "When email was created in the early 1970s it was the first'killer app' for the web.

Famously, in 2014 Prof. Stephen Hawking told the BBC: "The development of full artificial intelligence could spell the end of the human race." The ethical questions around Artificial Intelligence were discussed at a meeting led by the BCS President Chris Rees in London on October 2nd. This is also an area covered by KuppingerCole under the heading of Cognitive Technologies and this blog provides a summary of some of the issues that need to be considered. Firstly, AI is a generic term and it is important to understand precisely what this means. Currently the state of the art can be described as Narrow AI.

Annotation guidelines used to guide the annotation of training and evaluation datasets can have a considerable impact on the quality of machine learning models. In this study, we explore the effects of annotation guidelines on the quality of app feature extraction models. As a main result, we propose several changes to the existing annotation guidelines with a goal of making the extracted app features more useful and informative to the app developers. We test the proposed changes via simulating the application of the new annotation guidelines and then evaluating the performance of the supervised machine learning models trained on datasets annotated with initial and simulated guidelines. While the overall performance of automatic app feature extraction remains the same as compared to the model trained on the dataset with initial annotations, the features extracted by the model trained on the dataset with simulated new annotations are less noisy and more informative to the app developers. Secondly, we are interested in what kind of annotated training data is necessary for training an automatic app feature extraction model. In particular, we explore whether the training set should contain annotated app reviews from those apps/app categories on which the model is subsequently planned to be applied, or is it sufficient to have annotated app reviews from any app available for training, even when these apps are from very different categories compared to the test app. Our experiments show that having annotated training reviews from the test app is not necessary although including them into training set helps to improve recall. Furthermore, we test whether augmenting the training set with annotated product reviews helps to improve the performance of app feature extraction. We find that the models trained on augmented training set lead to improved recall but at the cost of the drop in precision.

This article is part of Demystifying AI, a series of posts that (try) to disambiguate the jargon and myths surrounding AI. A while ago, while browsing through the latest AI news, I stumbled upon a company that claimed to use "machine learning and advanced artificial intelligence" to collect and analyze hundreds of data touch points to improve user experience in mobile apps. On the same day, I read about another company that predicted customer behavior using "a combination of machine learning and AI" and "AI-powered predictive analytics." Some people refer to AI and machine learning as synonyms and use them interchangeably, while other use them as separate, parallel technologies. In many cases, the people speaking and writing about the technology don't know the difference between AI and ML. In others, they intentionally ignore those differences to create hype and excitement for marketing and sales purposes.

To handle more-natural spoken interactions, Alexa must track references through several rounds of conversation. If, for instance, a customer says, "How far is it to Redmond?" and after the answer follows up by saying, "Find good Indian restaurants there", Alexa should be able to infer that "there" refers to Redmond. We call the task of reference tracking "context carryover," and it's a capability that is currently being phased in to the Alexa experience. At this year's Interspeech, the largest conference on spoken-language understanding, my colleagues and I will present a paper titled "Contextual Slot Carryover for Disparate Schemas," which describes our solution to the problem of slot carryover, a crucial aspect of context carryover. "Domain" describes the type of application -- or "skill" -- that the utterance should invoke; for instance, mapping skills should answer questions about geographic distance.

What happens when you put a neural network in charge of a rules-based marketing automation solution? Would the AI emerge victorious, or does the human driver still have a thing or two to show the talented mimicker? It's an interesting question, to be sure, but more importantly, and it's an experiment that the folks at SAS – which still uses rules-based approaches in some of its analytics offerings -- actually ran, and the results might surprise you. "It beat our system," SAS Executive Vice President Oliver Schabenberger said during the SAS Analytics Experience conference held last week in San Diego, California. The result forced Schabenberger, who also holds the title of CTO and COO, to inquire about the cause. "Why is it the AI system works better than what our best minds can put together?" he said.

CHINA'S "Sputnik moment" came on May 27th 2017. On that day an algorithm thrashed Ke Jie, the world's best player of Go, an ancient and demanding Chinese board game. Mr Ke's defeat by AlphaGo, an artificial intelligence (AI) system developed by DeepMind, a British firm that had been bought by Google, was as much a blow to China's psyche as the Soviet satellite was to America's self-esteem in 1957. Within months, China announced ambitious plans to dominate AI by 2030. Kai-Fu Lee thinks it will succeed. He is well placed to judge.

I first entered the informatics field in the late 1980s, at the tail end of the first era of artificial intelligence (AI) in medicine. Initial systems focused on making medical diagnoses using symbolic processing, which was appropriate for a time of relatively little digital data, both for individual patients and healthcare as whole, and underpowered hardware. Systems like MYCIN [1], INTERNIST-1/QMR [2], and DXPLAIN [3] provided relatively accurate diagnostic performance, but were slow and difficult to use. They also provided a single likely diagnosis, which was not really what clinicians needed. Because of these shortcomings, they never achieved significant real-world adoption, and their "Greek Oracle" style of approach was abandoned. There was also some early enthusiasm for neural networks around that time [5], although in retrospect those systems were hampered by lack of data and computing power.