Neural Machine Translation (NMT) has become a popular technology in recent years, and beam search is its de facto decoding method due to the shrunk search space and reduced computational complexity. However, since it only searches for local optima at each time step through one-step forward looking, it usually cannot output the best target sentence. Inspired by the success and methodology of AlphaGo, in this paper we propose using a prediction network to improve beam search, which takes the source sentence $x$, the currently available decoding output $y_1,\cdots, y_{t-1}$ and a candidate word $w$ at step $t$ as inputs and predicts the long-term value (e.g., BLEU score) of the partial target sentence if it is completed by the NMT model. Following the practice in reinforcement learning, we call this prediction network \emph{value network}. Specifically, we propose a recurrent structure for the value network, and train its parameters from bilingual data. During the test time, when choosing a word $w$ for decoding, we consider both its conditional probability given by the NMT model and its long-term value predicted by the value network. Experiments show that such an approach can significantly improve the translation accuracy on several translation tasks.

Unsupervised image-to-image translation aims at learning a joint distribution of images in different domains by using images from the marginal distributions in individual domains. Since there exists an infinite set of joint distributions that can arrive the given marginal distributions, one could infer nothing about the joint distribution from the marginal distributions without additional assumptions. To address the problem, we make a shared-latent space assumption and propose an unsupervised image-to-image translation framework based on Coupled GANs. We compare the proposed framework with competing approaches and present high quality image translation results on various challenging unsupervised image translation tasks, including street scene image translation, animal image translation, and face image translation. We also apply the proposed framework to domain adaptation and achieve state-of-the-art performance on benchmark datasets.

Learning meaningful representations that maintain the content necessary for a particular task while filtering away detrimental variations is a problem of great interest in machine learning. In this paper, we tackle the problem of learning representations invariant to a specific factor or trait of data. The representation learning process is formulated as an adversarial minimax game. We analyze the optimal equilibrium of such a game and find that it amounts to maximizing the uncertainty of inferring the detrimental factor given the representation while maximizing the certainty of making task-specific predictions. On three benchmark tasks, namely fair and bias-free classification, language-independent generation, and lighting-independent image classification, we show that the proposed framework induces an invariant representation, and leads to better generalization evidenced by the improved performance.

Kyunghyun Cho, an Assistant Professor at NYU's Center for Data Science (CDS) and the Courant Institute of Mathematical Sciences, has been selected as a 2017 CIFAR Azrieli Global Scholar. CIFAR is a global research institute that connects leading scholars through interdisciplinary research programs like child and brain development, cosmology, genetic networks, learning machines and brains, and more. Founded in 1982, as many as 18 Nobel Laureates have since been associated with CIFAR, and its selected researchers and fellows are continually amongst the most highly cited scholars in their fields. Supported by the Azrieli Foundation, CIFAR's two year program offers the opportunity for researchers like Cho be mentored by other experts in his field, as well as exchange ideas with industry leaders outside of academia. He will also receive $100,000 to support his work, which has already significant contributions to fields like medicine and neural machine translation.

Neural networks, which learn to perform computational tasks by analyzing huge sets of training data, have been responsible for the most impressive recent advances in artificial intelligence, including speech-recognition and automatic-translation systems. During training, however, a neural net continually adjusts its internal settings in ways that even its creators can't interpret. Much recent work in computer science has focused on clever techniques for determining just how neural nets do what they do. In several recent papers, researchers from MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL) and the Qatar Computing Research Institute have used a recently developed interpretive technique, which had been applied in other areas, to analyze neural networks trained to do machine translation and speech recognition. They find empirical support for some common intuitions about how the networks probably work.

The Duke University professor was at a statistics conference last year when, she said, she witnessed Steven Scott, a senior artificial intelligence (AI) researcher at Google, make sexual advances on one of her female students. According to Heller, when she spoke to Scott later at an event dinner, he was defensive and told the professor that she should be nice to him considering that he had secured her a Google-funded faculty research award. Artificial Intelligence has various definitions, but in general it means a program that uses data to build a model of some aspect of the world. This model is then used to make informed decisions and predictions about future events. The technology is used widely, to provide speech and face recognition, language translation, and personal recommendations on music, film and shopping sites.

Amazon Translate is a new service announced at AWS re:Invent 2017. At the time of writing, it is available in preview. Please consider joining it and sending us feedback! Let's try it on a few examples. Please keep in mind that the service is still in preview and that it's constantly learning: imperfections will quickly be fixed thanks to customer feedback.

While EMMA is a relatively simple application, developers are thinking bigger as well: Today's cognitive technologies can track the course, speed, and destination of nearly 2,000 airliners at a time, allowing them to fly safely.4 Over time, AI will spawn massive changes in the public sector, transforming how government employees get work done. It's likely to eliminate some jobs, lead to the redesign of countless others, and create entirely new professions.5 In the near term, our analysis suggests, large government job losses are unlikely. But cognitive technologies will change the nature of many jobs--both what gets done and how workers go about doing it--freeing up to one quarter of many workers' time to focus on other activities.

We introduce AI rationalization, an approach for generating explanations of autonomous system behavior as if a human had performed the behavior. We describe a rationalization technique that uses neural machine translation to translate internal state-action representations of an autonomous agent into natural language. We evaluate our technique in the Frogger game environment, training an autonomous game playing agent to rationalize its action choices using natural language. A natural language training corpus is collected from human players thinking out loud as they play the game. We motivate the use of rationalization as an approach to explanation generation and show the results of two experiments evaluating the effectiveness of rationalization. Results of these evaluations show that neural machine translation is able to accurately generate rationalizations that describe agent behavior, and that rationalizations are more satisfying to humans than other alternative methods of explanation.

We describe Sockeye (version 1.12), an open-source sequence-to-sequence toolkit for Neural Machine Translation (NMT). Sockeye is a production-ready framework for training and applying models as well as an experimental platform for researchers. Written in Python and built on MXNet, the toolkit offers scalable training and inference for the three most prominent encoder-decoder architectures: attentional recurrent neural networks, self-attentional transformers, and fully convolutional networks. Sockeye also supports a wide range of optimizers, normalization and regularization techniques, and inference improvements from current NMT literature. Users can easily run standard training recipes, explore different model settings, and incorporate new ideas. In this paper, we highlight Sockeye's features and benchmark it against other NMT toolkits on two language arcs from the 2017 Conference on Machine Translation (WMT): English-German and Latvian-English. We report competitive BLEU scores across all three architectures, including an overall best score for Sockeye's transformer implementation. To facilitate further comparison, we release all system outputs and training scripts used in our experiments. The Sockeye toolkit is free software released under the Apache 2.0 license.