Abstractive summarization systems aim to produce more coherent and concise summaries than their extractive counterparts. Popular neural models have achieved impressive results for single-document summarization, yet their outputs are often incoherent and unfaithful to the input. In this paper, we introduce SENECA, a novel System for ENtity-drivEn Coherent Abstractive summarization framework that leverages entity information to generate informative and coherent abstracts. Our framework takes a two-step approach: (1) an entity-aware content selection module first identifies salient sentences from the input, then (2) an abstract generation module conducts cross-sentence information compression and abstraction to generate the final summary, which is trained with rewards to promote coherence, conciseness, and clarity. The two components are further connected using reinforcement learning. Automatic evaluation shows that our model significantly outperforms previous state-of-the-art on ROUGE and our proposed coherence measures on New York Times and CNN/Daily Mail datasets. Human judges further rate our system summaries as more informative and coherent than those by popular summarization models.

Before you start reading, think of 3 possible scenarios for the future of Artificial Intelligence (AI). If I asked you to think of 3 possible scenarios for the future of AI, I am guessing you'd think of the bad first: Takeover scenario -- Terminator-style. Computers and robots dominate human species, take over our planet, and eventually wipe us off the face of Earth. Or, that the power of AI will be held, and used by a handful of tyrants whose sole purpose is to enslave the rest of us. You might've also thought of a hybrid scenario, where we lose some of our humanity to gain far superior computational and physical power. And finally, you might've even thought of brighter days where robots work for human species who now enjoy their Universal Basic Income (UBI), follow their "passions" or their "useless" creative endeavors, and live without a single worry in the world. Even though these are the most commonly talked about scenarios, I think we are missing the most probable scenarios somewhere in the "boring AI outcomes" section. First of all, AI, being as hyped of a topic as it is, attracts attention, and attention is usually not maintained by analyzing history and political philosophy and coming up with a possible outcome based on that. Attention is maintained by either fear or hope for a better tomorrow (i.e. That's why these'common scenarios' are not only the most written but also the most read about scenarios. If you haven't picked it up already, you'll be reading about one of the "boring AI outcomes".

Recently, law professor and AI activist Ryan Abbott filed patent applications for two innovative designs. The patents were filed on behalf of the inventor -- an algorithm named DABUS. But can AI really be an inventor? Doubling down on this move, researchers from the University of Surrey, including Abbott, established the Artificial Inventor Project, an initiative that seeks "intellectual property rights for the autonomous output of artificial intelligence." According to the patent bureau, DABUS met all of the criteria necessary to be considered a qualified applicant.

One of the biggest legal problems protecting AI users in the coming years will be accountability – dealing with the opacity of the black box and explaining decisions made by machine thinking. Understanding the logic behind an AI finding is not an issue where AI is assisting in spotting real-world risks that affect individuals – such as the current use of AI in radiology, where failure to use AI radiology analysis may soon be considered malpractice. As long as the AI is accurate and productive in showing where cancer may exist, we don't care how the machine picked that specific spot on the x-ray, we are just happy to have another tool that helps save lives. But where the AI proposes treatments or outcomes, your clients – healthcare and otherwise – will need to be ready to defend those decisions. This means an entirely different baseline organization and feature set for than the AI currently envisioned or in use.

As with other aspects of customer-technology, first there was the task, then came the data and then the need to manage and use it properly. Customer spreadsheets and databases became CRM tools, point of sales records became data for recommendation engines and marketing, and so on. The same is true with chatbots, as businesses look to leverage the data they provide to add value. Chatbots are rapidly changing the way that businesses interact with their clients, and how data is provided to their systems. It isn't a massive change from previous generations, when a sales database became a tool for the whole organization to use, but still represents a change in how companies will work.

Eric Schmidt, former CEO and chairman of Google, has donated money to Stanford University, and taught at its business school. But a group of current and former Google employees, academics, and human rights activists wants the university to cancel a talk he is scheduled to give next month at a conference on ethics and artificial intelligence. They say Schmidt is a poor ethical role model. In a letter to the conference organizers, the group says Schmidt's appearance would be inappropriate given "serious and credible" questions over his ethical conduct. Their petition was publicly released Tuesday with more than 40 signatories, including 20 current Google employees, but first sent to Stanford Sunday.

In the fall of 2018, the New York Times published a piece about an experiment in which they used an algorithm to produce Halloween costume ideas. The results were amusing: "baseball clown", "cat witch", "king dog." The algorithm combined random letters to make words, which it then compared to the set of real words in its training data. If it found a match, it kept the word and paired it with another one. In this example--and in many more serious ones like it--the algorithm was given pre-existing patterns and taught to replicate them.

All that automation yields data that can be used to analyze workers' performance. Those analyses, whether done by humans or software programs, may affect who is hired, fired, promoted and given raises. Some artificial intelligence programs can mine and manipulate the data to predict future actions, such as who is likely to quit their job, or to diagnose medical conditions. If your job doesn't currently involve these types of technologies, it likely will in the very near future. This worries me--a labor and employment law scholar who researches the role of technology in the workplace--because unless significant changes are made to American workplace laws, these sorts of surveillance and privacy invasions will be perfectly legal.

There are countless news stories and scientific publications illustrating how artificial intelligence (AI) will change the world. As far as law is concerned, discussions largely center around how AI systems such as IBM's Watson will cause disruption in the legal industry. However, little attention has been directed at how AI might prove beneficial for the field of private international law. Private international law has always been a complex discipline, and its application in the online environment has been particularly challenging, with both jurisdictional overreach and jurisdictional gaps. Primarily, this is due to the fact that the near-global reach of a person's online activities will so easily expose that person to the jurisdiction and laws of a large number of countries. Thus, online users ranging from individuals to the largest online companies are subject to unpredictable legal consequences when using the Internet.

In this paper, we study Censored Semi-Bandits, a novel variant of the semi-bandits problem. The learner is assumed to have a fixed amount of resources, which it allocates to the arms at each time step. The loss observed from an arm is random and depends on the amount of resource allocated to it. More specifically, the loss equals zero if the allocation for the arm exceeds a constant (but unknown) threshold that can be dependent on the arm. Our goal is to learn a feasible allocation that minimizes the expected loss. The problem is challenging because the loss distribution and threshold value of each arm are unknown. We study this novel setting by establishing its `equivalence' to Multiple-Play Multi-Armed Bandits (MP-MAB) and Combinatorial Semi-Bandits. Exploiting these equivalences, we derive optimal algorithms for our setting using existing algorithms for MP-MAB and Combinatorial Semi-Bandits. Experiments on synthetically generated data validate performance guarantees of the proposed algorithms.