Recent work has shown that state-of-the-art models are highly vulnerable to adversarial perturbations of the input. We propose cowboy, an approach to detecting and defending against adversarial attacks by using both the discriminator and generator of a GAN trained on the same dataset. We show that the discriminator consistently scores the adversarial samples lower than the real samples across multiple attacks and datasets. We provide empirical evidence that adversarial samples lie outside of the data manifold learned by the GAN. Based on this, we propose a cleaning method which uses both the discriminator and generator of the GAN to project the samples back onto the data manifold. This cleaning procedure is independent of the classifier and type of attack and thus can be deployed in existing systems.

We present a study on reinforcement learning (RL) from human bandit feedback for sequence-to-sequence learning, exemplified by the task of bandit neural machine translation (NMT). We investigate the reliability of human bandit feedback, and analyze the influence of reliability on the learnability of a reward estimator, and the effect of the quality of reward estimates on the overall RL task. Our analysis of cardinal (5-point ratings) and ordinal (pairwise preferences) feedback shows that their intra- and inter-annotator $\alpha$-agreement is comparable. Best reliability is obtained for standardized cardinal feedback, and cardinal feedback is also easiest to learn and generalize from. Finally, improvements of over 1 BLEU can be obtained by integrating a regression-based reward estimator trained on cardinal feedback for 800 translations into RL for NMT. This shows that RL is possible even from small amounts of fairly reliable human feedback, pointing to a great potential for applications at larger scale.

The main objective of explanations is to transmit knowledge to humans. This work proposes to construct informative explanations for predictions made from machine learning models. Motivated by the observations from social sciences, our approach selects data points from the training sample that exhibit special characteristics crucial for explanation, for instance, ones contrastive to the classification prediction and ones representative of the models. Subsequently, semantic concepts are derived from the selected data points through the use of domain ontologies. These concepts are filtered and ranked to produce informative explanations that improves human understanding. The main features of our approach are that (1) knowledge about explanations is captured in the form of ontological concepts, (2) explanations include contrastive evidences in addition to normal evidences, and (3) explanations are user relevant.

Masayoshi Son, with a net worth of $21.5bn, is reputed to be Japan's wealthiest person. He is also one of the world's slickest and smartest salesmen. In September 2016, Son met then Saudi Deputy Crown Prince Mohammed bin Salman. By the end of the 45-minute meeting, bin Salman, MBS, as he is known, had committed $45bn from Saudi Arabia's Public Investment Fund (PIF) to Son's Vision Fund. That's nearly half the value of what has become the biggest investment fund the world has ever seen.

As of January 2017, MIT started offering a variety of course in deep learning.

There is a lot you can do with Alexa. Here's how you can add skills. It was every Amazon Echo owner's nightmare. Alexa, the connected speaker, really, truly, was listening in on your conservations, and behind your back, passed on the recording of a private chit-chat to someone on your Echo contact list. This actually happened this week, according to Seattle TV station KIRO, which told the story of a Portland woman's privacy gone amuck.

The Global Artificial Intelligence In Insurance Market Report includes a comprehensive analysis of the present market. The report starts with the basic Artificial Intelligence In Insurance Market overview and then goes into each and every detail. Houston, TX -- (SBWIRE) -- 05/26/2018 -- Artificial Intelligence In Insurance Market research report offers comprehensive insight into the key growth drivers, notable challenges, prominent trends, recent technological advancements, and the competitive landscape. The study presents a critical assessment of the scope of key applications and the innovations in products brought about by key players. It further takes a closer look at prevailing regulatory landscape in major regions and identifies promising avenues.

Demand for technological, social and emotional, and higher cognitive skills will rise by 2030. How will workers and organizations adapt? Skill shifts have accompanied the introduction of new technologies in the workplace since at least the Industrial Revolution, but adoption of automation and artificial intelligence (AI) will mark an acceleration over the shifts of even the recent past. The need for some skills, such as technological as well as social and emotional skills, will rise, even as the demand for others, including physical and manual skills, will fall. These changes will require workers everywhere to deepen their existing skill sets or acquire new ones. Companies, too, will need to rethink how work is organized within their organizations.

Eric Schmidt, the former chairman of Google's parent company Alphabet and now its technical adviser, joined the list of people Friday who oppose Tesla and SpaceX CEO Elon Musk's views about the future of artificial intelligence. Musk has warned that AI, if unregulated, will eventually become an existentialist threat to humanity, and his opinion has both famous supporters, like the late Stephen Hawking, and dissenters like Schmidt. Speaking at the VivaTech conference in Paris on Friday, Schmidt's comments were in response to a question about Musk's dire warnings about AI. "I think Elon is exactly wrong. The fact of the matter is that AI and machine learning are so fundamentally good for humanity," Schmidt said, adding he shared Musk's concerns about the potential for misuse of technology. Alphabet's former executive chairman Eric Schmidt speaks on the phone during the World Economic Forum (WEF) annual meeting in Davos, Switzerland Jan. 24, 2018.

John Ludwig is an electrical engineer and the president of Xevo's Artificial Intelligence (AI) Group. Xevo is a tier-one OEM software company, located in Seattle, that manages automotive software for driver assistance, engagement, and in-vehicle entertainment. Its main product is the Xevo Market, a merchant-to-driver commerce platform that uses a vehicle's infotainment screen to make purchases and transations from inside the car. Xevo Market launched at the end of 2017 and is already available in millions of vehicles. Prior to working with Xevo, Ludwig was a software manager with Microsoft, overseeing operating systems and online service projects.