The AI Bookie column documents highlights from AI Bets, an online forum for the creation of adjudicatable predictions, in the form of bets, about the future of AI. While it is easy to make broad, generalized, or off-the-cuff predictions about the future, it is more difficult to develop predictions that are carefully thought out, concrete, and measurable. This forum was created to help researchers craft predictions whose accuracy can be clearly and unambiguously judged when the bets come due. The bets will be documented both online and regularly in this column. We encourage bets that are rigorously and scientifically argued. We discourage bets that are too general to be evaluated or too specific to an individual or institution. The goal is not to continue to feed the media frenzy and outsized pundit predictions about AI, but rather to curate and promote bets whose outcomes will provide useful feedback to the scientific community. For detailed guidelines and to place bets, visit sciencebets.org.

The AAAI 2017 Fall Symposium Series was held Thursday through Saturday, November 9–11, at the Westin Arlington Gateway in Arlington, Virginia, adjacent to Washington, DC. The titles of the six symposia were Artificial Intelligence for Human-Robot Interaction; Cognitive Assistance in Government and Public Sector Applications; Deep Models and Artificial Intelligence for Military Applications: Potentials, Theories, Practices, Tools and Risks; Human-Agent Groups: Studies, Algorithms and Challenges; Natural Communication for Human-Robot Collaboration; and A Standard Model of the Mind. The highlights of each symposium (except the Natural Communication for Human-Robot Collaboration symposium, whose organizers did not submit a report) are presented in this report.

With the increasing ubiquity of artificial intelligence and machine learning applications, systems are emerging that require non-ML experts to interact with machine learning at the training step, not just the final system. These users may not have the skills, time, or inclination to familiarize themselves with the way machine learning works, so training systems must be developed that can communicate the necessary information and facilitate effortless collaboration with the user. We consider how to utilize techniques from qualitative coding, a human-centered approach for manual classification, and build better user experience for ML training.

Deep neural network (DNN) architecture based models have high expressive power and learning capacity. However, they are essentially a black box method since it is not easy to mathematically formulate the functions that are learned within its many layers of representation. Realizing this, many researchers have started to design methods to exploit the drawbacks of deep learning based algorithms questioning their robustness and exposing their singularities. In this paper, we attempt to unravel three aspects related to the robustness of DNNs for face recognition: (i) assessing the impact of deep architectures for face recognition in terms of vulnerabilities to attacks inspired by commonly observed distortions in the real world that are well handled by shallow learning methods along with learning based adversaries; (ii) detecting the singularities by characterizing abnormal filter response behavior in the hidden layers of deep networks; and (iii) making corrections to the processing pipeline to alleviate the problem. Our experimental evaluation using multiple open-source DNN-based face recognition networks, including OpenFace and VGG-Face, and two publicly available databases (MEDS and PaSC) demonstrates that the performance of deep learning based face recognition algorithms can suffer greatly in the presence of such distortions. The proposed method is also compared with existing detection algorithms and the results show that it is able to detect the attacks with very high accuracy by suitably designing a classifier using the response of the hidden layers in the network. Finally, we present several effective countermeasures to mitigate the impact of adversarial attacks and improve the overall robustness of DNN-based face recognition.

A major challenge in Entity Linking (EL) is making effective use of contextual information to disambiguate mentions to Wikipedia that might refer to different entities in different contexts. The problem exacerbates with cross-lingual EL which involves linking mentions written in non-English documents to entries in the English Wikipedia: to compare textual clues across languages we need to compute similarity between textual fragments across languages. In this paper, we propose a neural EL model that trains fine-grained similarities and dissimilarities between the query and candidate document from multiple perspectives, combined with convolution and tensor networks. Further, we show that this English-trained system can be applied, in zero-shot learning, to other languages by making surprisingly effective use of multi-lingual embeddings. The proposed system has strong empirical evidence yielding state-of-the-art results in English as well as cross-lingual: Spanish and Chinese TAC 2015 datasets.

Reports on the 2016 AAAI Fall Symposium Series

The AAAI 2016 Fall Symposium Series was held Thursday through Saturday, November 17–19, at the Westin Arlington Gateway in Arlington, Virginia adjacent to Washington, DC. The titles of the six symposia were Accelerating Science: A Grand Challenge for AI; Artificial Intelligence for Human-Robot Interaction, Cognitive Assistance in Government and Public Sector Applications, Cross-Disciplinary Challenges for Autonomous Systems, Privacy and Language Technologies, Shared Autonomy in Research and Practice. The highlights of each (except Acceleration Science) symposium are presented in this report.

A challenging Earth-observing satellite scheduling problem was recently studied in (Frank, Do and Tran 2016) for which the best resolution approach so far on the proposed benchmark is a time-indexed Mixed Integer Linear Program (MILP) formulation. This MILP formulation produces feasible solutions but is not able to prove optimality or to provide tight optimality gaps, making it difficult to assess the quality of existing solutions. In this paper, we first introduce an alternative disjunctive MILP formulation that manages to close more than half of the instances of the benchmark. This MILP formulation is then relaxed to provide good bounds on optimal values for the unsolved instances. We then propose a CP Optimizer model that consistently outperforms the original time-indexed MILP formulation, reducing the optimality gap by more than 4 times. This Constraint Programming (CP) formulation is very concise: we give its complete OPL implementation in the paper. Some improvements of this CP model are reported resulting in an approach that produces optimal or near-optimal solutions (optimality gap smaller than 1%) for about 80% of the instances. Unlike the MILP formulations, it is able to quickly produce good quality schedules and it is expected to be flexible enough to handle the changing requirements of the application.

We introduce a new class of models called multiresolution recurrent neural networks, which explicitly model natural language generation at multiple levels of abstraction. The models extend the sequence-to-sequence framework to generate two parallel stochastic processes: a sequence of high-level coarse tokens, and a sequence of natural language words (e.g. sentences). The coarse sequences follow a latent stochastic process with a factorial representation, which helps the models generalize to new examples. The coarse sequences can also incorporate task-specific knowledge, when available. In our experiments, the coarse sequences are extracted using automatic procedures, which are designed to capture compositional structure and semantics. These procedures enable training the multiresolution recurrent neural networks by maximizing the exact joint log-likelihood over both sequences. We apply the models to dialogue response generation in the technical support domain and compare them with several competing models. The multiresolution recurrent neural networks outperform competing models by a substantial margin, achieving state-of-the-art results according to both a human evaluation study and automatic evaluation metrics. Furthermore, experiments show the proposed models generate more fluent, relevant and goal-oriented responses.

Cyber security organizations charged with protecting IT infrastructures face daunting complex challenges. These include a vast array of information sources of variable trustworthiness, overwhelming numbers of incident reports, quickly changing offensive and defensive tactics, and the ongoing shortage of skilled cybersecurity personnel. Cognitive systems offer a new approach to addressing these challenges. Using natural language processing and machine learning techniques, systems such as IBM Watsonä can incorporate an enormous amount of information each day, discover the entities and relationships described, and apply reasoning over that knowledge to understand questions from the Security Analyst and provide answers in their own language. In this paper we discuss the technologies and techniques we used to build a cognitive cyber security assistant.