More than a hundred startup teams are beginning a training and mentoring program this month at the Berkeley SkyDeck Accelerator--the largest group ever accepted to the UC Berkeley program since it was founded in 2012. Aside from the funding, free office rent, and other resources offered by the startup accelerator located near the edge of the Berkeley campus, the program is also setting an example for its young companies--as an organization going all out for growth. SkyDeck, originally launched as free office space where startups could roost and benefit from the advice of mentors from the university community, now partners with an affiliated venture capital firm, Berkeley SkyDeck Fund, which invests $100,000 in each company entering the program's formal six-month session as a "Cohort" member. That's comparable to the $120,000 invested in each startup nurtured by the influential Silicon Valley accelerator program Y Combinator. For the fall session, 22 Cohort startups were chosen, and an additional 80 teams were admitted as "HotDesk" members who can attend workshops, consult mentors, use office desks as available, and prepare for their company's next stage.

Machine learning packages take care of the building and implementing the top machine learning algorithms, creating workflows, and in general helping to solve machine learning problems. They provide the primary toolkit for different classification, regression, and other problems. As an integral part of data science, data manipulation and analysis field represents libraries that carry out data scraping, ingestion, cleaning, pre-processing and other operations that allow you to "play with the data" and as a result to perform the analysis itself. With the help of visualization packages, you can display the data visually which is necessary for better understanding and interpreting the data. These packages contain numerous visualization charts as well as different options for representation.

Engineering students at the University of Leicester (www.le.ac.uk) have developed a miniature autonomous?drive-by-wire? It is a 1:6 scale version of the LUTZ Pathfinder Pod developed by the Transport Systems Catapult (TSC).

IBM's Project Debater can have meaningful conversations with humans. As the adage goes, "to err is human." Mistakes and bias are built into the human condition. We can try our very best to maintain objectivity, but more often than not we allow personal biases to creep into our everyday decision making. While most of the time these biases are harmless, this can become a huge problem when your job is hiring people.

Two years ago, Sanjaya Wijeratne--a computer science PhD student at Wright State University--noticed something odd in his research. He was studying the communication of gang members on Twitter. Among the grandstanding about drugs and money, he found gang members repeatedly dropping the emoji in their tweets. Wijeratne had been working on separate research relating to word-sense disambiguation, a field of computational linguistics that looks at how words take on multiple meanings. The use of jumped out as a brand new problem.

A study by Stanford University indicates that virtual reality, adaptive learning or analytical learning will be common in the classroom within fifteen years. Although Artificial Intelligence (AI) is already part of our lives, it is still strange to hear about it in areas such as education, where the reality of the classroom advances at a much slower pace than that of technology. However, it is precisely the educational field that could be reinforced and transformed the most thanks to the new artificial intelligence systems and their capacity to contribute to the personalisation of learning. This is what a group of researchers and academics believe that, backed by Standford University, published last September the report Artificial Intelligence and Life in 2030. According to the study, virtual reality, adaptive learning, analytical learning and online teaching will be common in classrooms in just fifteen years.

"AI is the new electricity" said Andrew Ng, computer scientist and co-founder of online university, Coursera. Artificial intelligence (AI) has indeed come a long way from being confined in the research laboratories or science fiction movies. It is a persistent reality in today's world, which has permeated every layer of business, causing disruptions of unprecedented magnitude. In such a short span of time, AI has an incredible impact in reshaping the entire retail landscape by boosting productivity, increasing accuracy and improving reliability of business intelligence. AI and machine learning have already begun to make sweeping changes to the entire retail ecosystem.

We introduce a new family of margin-based regret guarantees for adversarial contextual bandit learning. Our results are based on multiclass surrogate losses. Using the ramp loss, we derive a universal margin-based regret bound in terms of the sequential metric entropy for a benchmark class of real-valued regression functions. The new margin bound serves as a complete contextual bandit analogue of the classical margin bound from statistical learning. The result applies to large nonparametric classes, improving on the best known results for Lipschitz contextual bandits (Cesa-Bianchi et al., 2017) and, as a special case, generalizes the dimension-independent Banditron regret bound (Kakade et al., 2008) to arbitrary linear classes with smooth norms. On the algorithmic side, we use the hinge loss to derive an efficient algorithm with a $\sqrt{dT}$-type mistake bound against benchmark policies induced by $d$-dimensional regression functions. This provides the first hinge loss-based solution to the open problem of Abernethy and Rakhlin (2009). With an additional i.i.d. assumption we give a simple oracle-efficient algorithm whose regret matches our generic metric entropy-based bound for sufficiently complex nonparametric classes. Under realizability assumptions our results also yield classical regret bounds.

Learning to infer Bayesian posterior from a few-shot dataset is an important step towards robust meta-learning due to the model uncertainty inherent in the problem. In this paper, we propose a novel Bayesian model-agnostic meta-learning method. The proposed method combines scalable gradient-based meta-learning with nonparametric variational inference in a principled probabilistic framework. During fast adaptation, the method is capable of learning complex uncertainty structure beyond a point estimate or a simple Gaussian approximation. In addition, a robust Bayesian meta-update mechanism with a new meta-loss prevents overfitting during meta-update. Remaining an efficient gradient-based meta-learner, the method is also model-agnostic and simple to implement. Experiment results show the accuracy and robustness of the proposed method in various tasks: sinusoidal regression, image classification, active learning, and reinforcement learning.

Modern AI and robotic systems are characterized by a high and ever-increasing level of autonomy. At the same time, their applications in fields such as autonomous driving, service robotics and digital personal assistants move closer to humans. From the combination of both developments emerges the field of AI ethics which recognizes that the actions of autonomous machines entail moral dimensions and tries to answer the question of how we can build moral machines. In this paper we argue for taking inspiration from Aristotelian virtue ethics by showing that it forms a suitable combination with modern AI due to its focus on learning from experience. We furthermore propose that imitation learning from moral exemplars, a central concept in virtue ethics, can solve the value alignment problem. Finally, we show that an intelligent system endowed with the virtues of temperance and friendship to humans would not pose a control problem as it would not have the desire for limitless self-improvement.