Google Translate is getting brainier. The online translation tool recently started using a neural network to translate between some of its most popular languages – and the system is now so clever it can do this for language pairs on which it has not been explicitly trained. To do this, it seems to have created its own artificial language. Traditional machine-translation systems break sentences into words and phrases, and translate each individually. In September, Google Translate unveiled a new system that uses a neural network to work on entire sentences at once, giving it more context to figure out the best translation.

Twitter can sometimes feel like a language of its own, but one lecturer is using the social media site as a tool to teach Arabic. In Mahammed Bouabdallah's classes at the University of Westminster, London, students are set simple tasks using Twitter to complement their lessons. Bouabdallah publishes a photo or link and asks students to comment on it in Arabic, or runs a Twitter poll about events happening in Arabic-speaking countries and discusses the results in class. Sometimes, they will use Twitter's built-in translation tool and judge its accuracy. "They have to tweet outside the class, and we discuss it inside the class," says Bouabdallah. His own research suggests that Twitter is popular as a language learning tool, with 80 per cent of surveyed students responding positively to its use.

The 1st Workshop on Neural Machine Translation is a new annual workshop that will be co-located with ACL 2017 (Vancouver, July 30-August 4, 2017). Neural Machine Translation (NMT) is a simple new architecture for getting machines to learn to translate. Despite being relatively recent, NMT has demonstrated promising results and attracted much interest, achieving state-of-the-art results on a number of shared tasks. This workshop aims to cultivate research in neural machine translation and other aspects of machine translation and multilinguality that utilize neural models.

We present a memory augmented neural network for natural language understanding: Neural Semantic Encoders. NSE is equipped with a novel memory update rule and has a variable sized encoding memory that evolves over time and maintains the understanding of input sequences through read}, compose and write operations. NSE can also access multiple and shared memories. In this paper, we demonstrated the effectiveness and the flexibility of NSE on five different natural language tasks: natural language inference, question answering, sentence classification, document sentiment analysis and machine translation where NSE achieved state-of-the-art performance when evaluated on publically available benchmarks. For example, our shared-memory model showed an encouraging result on neural machine translation, improving an attention-based baseline by approximately 1.0 BLEU.

The year 2017 arrives and we humans are still in charge. The machines haven't taken over yet, but they are gaining on us. Google's DeepMind AlphaGo computer program recently beat the world champ at Go, a complex board game, while Japanese researchers plan to build the world's fastest supercomputer for use on artificial intelligence projects. It will do 130 quadrillion calculations per second, which is, um, really, really fast. She can explain it better than we can.

On Tuesday, the White House released a chilling report on AI and the economy. It began by positing that "it is to be expected that machines will continue to reach and exceed human performance on more and more tasks," and it warned of massive job losses. Yet to counter this threat, the government makes a recommendation that may sound absurd: we have to increase investment in AI. The risk to productivity and the US's competitive advantage is too high to do anything but double down on it. This approach not only makes sense, but also is the only approach that makes sense.

The thinking goes that machines are dumb, and for them to do cool shit, like translating, driving a car, or winning Go, we need to make them smart. And the only way we know how to do that efficiently is to teach them how to learn -- just like humans. If you had to sit and explain everything you ever learned to a computer, it would probably take hundreds of years. Yeah, *YOU* learned it in less, but you had a couple things going for you. Your brain, for example, has a somewhat intuitive understanding of physics.

While neural machine translation (NMT) is making good progress in the past two years, tens of millions of bilingual sentence pairs are needed for its training. However, human labeling is very costly. To tackle this training data bottleneck, we develop a dual-learning mechanism, which can enable an NMT system to automatically learn from unlabeled data through a dual-learning game. This mechanism is inspired by the following observation: any machine translation task has a dual task, e.g., English-to-French translation (primal) versus French-to-English translation (dual); the primal and dual tasks can form a closed loop, and generate informative feedback signals to train the translation models, even if without the involvement of a human labeler. In the dual-learning mechanism, we use one agent to represent the model for the primal task and the other agent to represent the model for the dual task, then ask them to teach each other through a reinforcement learning process. Based on the feedback signals generated during this process (e.g., the language-model likelihood of the output of a model, and the reconstruction error of the original sentence after the primal and dual translations), we can iteratively update the two models until convergence (e.g., using the policy gradient methods). We call the corresponding approach to neural machine translation \emph{dual-NMT}. Experiments show that dual-NMT works very well on English$\leftrightarrow$French translation; especially, by learning from monolingual data (with 10\% bilingual data for warm start), it achieves a comparable accuracy to NMT trained from the full bilingual data for the French-to-English translation task.

Several mechanisms to focus attention of a neural network on selected parts of its input or memory have been used successfully in deep learning models in recent years. Attention has improved image classification, image captioning, speech recognition, generative models, and learning algorithmic tasks, but it had probably the largest impact on neural machine translation. Recently, similar improvements have been obtained using alternative mechanisms that do not focus on a single part of a memory but operate on all of it in parallel, in a uniform way. Such mechanism, which we call active memory, improved over attention in algorithmic tasks, image processing, and in generative modelling. So far, however, active memory has not improved over attention for most natural language processing tasks, in particular for machine translation. We analyze this shortcoming in this paper and propose an extended model of active memory that matches existing attention models on neural machine translation and generalizes better to longer sentences. We investigate this model and explain why previous active memory models did not succeed. Finally, we discuss when active memory brings most benefits and where attention can be a better choice.

If you're fluent in more than you, already know that there's never any direct translation for a lot of phrases. Hell, I'll still use spanglish or other random blends to make sure I get the exact tone I'm looking for (And if my rather obvious humblebrag was a little off-putting, lo siento). But this is a big problem that computer scientists have been working on for years. And it's one that's almost impossible to solve without one of humanity's most recent (and terrifying) inventions: machine learning. The thinking goes that machines are dumb, and for them to do cool shit, like translating, driving a car, or winning Go, we need to make them smart.