All is open source, though not "published" yet so excuse me if the repository is a bit hard to navigate / unclear. I've been building a Unity-esque 2D engine with pygame, which should be easy to plug in with OpenAI's gym. Goal here isn't to build the most optimal environments per se, but a way to implement games that are human and AI playable. Hopefully I'll get to release a slightly more convenient setup soon, with each (sub)project separated into their own repo:)

I'm writing my first paper, which I am looking to publish in an academic journal. I'm a little intimidated by some of the papers that I was reading in the journals that I am looking to submit to. While the quality in the papers is definitely great, I feel there's a lot just with regards to the authors' skill in presentation. In addition to that, I feel like the journal papers seem to write well for the academic audience. I don't have a natural understanding of the nuances of this audience. Personally, I feel like I have a more down-to-earth personality, so if I was implementing a 20-layer network, I would say I am implementing a deep learning network, whereas in one of the academic papers, someone may say they are implementing a 20-layer encoder-decoder convolutional network.

Microsoft Cognitive Services let you build apps with powerful algorithms using just a few lines of code. They work across devices and platforms such as iOS, Android, and Windows, keep improving, and are easy to set up.

A number of this week's milestones in the history of technology demonstrate society's reactions to new technologies over the years: A discussion of AI replacing and augmenting human intelligence, a warning about the abundance of misinformation on the internet, and government regulation of a mass communication platform, suppressing free speech in the name of the public interest. On February 20, 1947, Alan Turing gave a talk at the London Mathematical Society in which he declared that "what we want is a machine that can learn from experience." Anticipating today's enthusiasm about machine learning and deep learning, Turing declared that "It would be like a pupil who had learnt much from his master, but had added much more by his own work. When this happens, I feel that one is obliged to regard the machine as showing intelligence." Turing also anticipated the debate over the impact of artificial intelligence on jobs: Does it destroy jobs (automation) or does it help humans do their jobs better and do more interesting things (augmentation)? Turing speculated that digital computers will replace some of the calculation work done at the time by human computers.

Abstract: We present a simple method to improve learning long-term dependencies in recurrent neural networks (RNNs) by introducing unsupervised auxiliary losses. These auxiliary losses force RNNs to either remember distant past or predict future, enabling truncated backpropagation through time (BPTT) to work on very long sequences. We experimented on sequences up to 16000 tokens long and report faster training, more resource efficiency and better test performance than full BPTT baselines such as Long Short Term Memory (LSTM) networks or Transformer. TL;DR: Combining auxiliary losses and truncated backpropagation through time in RNNs improves resource efficiency, training speed and generalization in learning long term dependencies.

Mail resume to Airspace Systems, Inc., Attn: Guy Bar-Nahum, 1933 Davis St., #229, San Leandro, CA 94577.

This weekend's 60th Annual Grammy Awards will feature big names like Beyoncé, Rihanna, and Watson. The latter is IBM's famous artificial intelligence platform, which the Grammys are enlisting to curate the videos and photos being released to music fans following along with this year's awards show on social media in real time. IBM is partnering with Grammys organizer the Recording Academy to provide Watson's AI services to populate the event's social media feeds with automatically-generated content during the Grammy Awards ceremony, which airs this Sunday, Jan. 28, on CBS. IBM's Watson will get to work before the ceremony even starts, analyzing and sorting "hours of video and close to 125,000 photographs" taken during the Grammys' hours-long red carpet show ahead of Sunday's event, IBM said in its announcement. The platform will use features such as facial recognition, even analyzing stars' "facial emotion," to pick out the best images and videos to post for fans online.

Bayesian graphical models have been shown to be a powerful tool for discovering uncertainty and causal structure from real-world data in many application fields. Current inference methods primarily follow different kinds of trade-offs between computational complexity and predictive accuracy. At one end of the spectrum, variational inference approaches perform well in computational efficiency, while at the other end, Gibbs sampling approaches are known to be relatively accurate for prediction in practice. In this paper, we extend an existing Gibbs sampling method, and propose a new deterministic Heron inference (Heron) for a family of Bayesian graphical models. In addition to the support for nontrivial distributability, one more benefit of Heron is that it is able to not only allow us to easily assess the convergence status but also largely improve the running efficiency. We evaluate Heron against the standard collapsed Gibbs sampler and state-of-the-art state augmentation method in inference for well-known graphical models. Experimental results using publicly available real-life data have demonstrated that Heron significantly outperforms the baseline methods for inferring Bayesian graphical models.

In an increasingly polarized world, demagogues who reduce complexity down to simple arguments based on emotion are gaining in popularity. Are opinions and online discussions falling into demagoguery? In this work, we aim to provide computational tools to investigate this question and, by doing so, explore the nature and complexity of online discussions and their space of opinions, uncovering where each participant lies. More specifically, we present a modeling framework to construct latent representations of opinions in online discussions which are consistent with human judgements, as measured by online voting. If two opinions are close in the resulting latent space of opinions, it is because humans think they are similar. Our modeling framework is theoretically grounded and establishes a surprising connection between opinion and voting models and the sign-rank of a matrix. Moreover, it also provides a set of practical algorithms to both estimate the dimension of the latent space of opinions and infer where opinions expressed by the participants of an online discussion lie in this space. Experiments on a large dataset from Yahoo! News, Yahoo! Finance, Yahoo! Sports, and the Newsroom app suggest that unidimensional opinion models may be often unable to accurately represent online discussions, provide insights into human judgements and opinions, and show that our framework is able to circumvent language nuances such as sarcasm or humor by relying on human judgements instead of textual analysis.

In his Wall Street Journal column this weekend, Nobel laureate Frank Wilczek offers a fascinating theory as to why we haven't been able to find signs of intelligent life elsewhere in the universe. Maybe, he suggests, intelligent beings are fated to shrink as their intelligence expands. Once the singularity happens, AI implodes into invisibility. Wilczek notes that "effective computation must involve interactions and that the speed of light limits communication." To optimize its thinking, an AI would have no choice but to compress itself to minimize delays in the exchange of messages.