On Wednesday 10 April, the first image ever taken of a black hole was released. The picture, which shows a black hole surrounded by a hazy red and yellow circle, provides an unprecedented peek at one of the most mysterious entities in the universe. One of the scientists involved in the development of the picture is Dr Katie Bouman. We'll tell you what's true. You can form your own view.

The black hole that starred in the first ever photo to be taken of its kind has been given a name. The now famous swirling void will be known as Powehi, a Hawaiian word which has been bestowed by a language professor. And the name's meaning, chosen by University of Hawaii-Hilo Hawaiian Professor Larry Kimura, is as fittingly dramatic as the picture and work that produced it. We'll tell you what's true. You can form your own view.

The music management system was once seen as the future of computing: it was used to control the iPod, and was home to the iTunes Store, both of which helped to revolutionise the way people buy and listen to music. But with time it has become bloated with additional features – from watching films to managing devices like phones – and its performance has dropped. That has led to it becoming largely despised within the tech community. We'll tell you what's true. You can form your own view.

Video Friday is your weekly selection of awesome robotics videos, collected by your Automaton bloggers. We'll also be posting a weekly calendar of upcoming robotics events for the next few months; here's what we have so far (send us your events!): Let us know if you have suggestions for next week, and enjoy today's videos. The 2019 IEEE International Conference on Soft Robotics (RoboSoft) takes place in Seoul, South Korea, next week, and the organizers put together this preview video stuffed full of--what else?--soft robots. Single-stream recycling is currently an extremely labor intensive process due to the need for manual object sorting.

A collaboration between researchers from China's Beihang University and Microsoft Research Asia has produced TableBank, a new image-based dataset for table detection and recognition built with novel weak supervision from Word and Latex documents on the Internet. Researchers built several strong baselines using SOTA models with deep neural networks, which will enable deployment of more deep learning methods to table detection and recognition tasks. TableBank has been open-sourced on Github. "Existing research for image-based table detection and recognition usually fine-tunes pre-trained models on out-of-domain data with a few thousands human labeled examples, which is difficult to generalize on real world applications. With TableBank that contains 417K high-quality labeled tables, we build several strong baselines using state-of-the-art models with deep neural networks."

The use of artificial intelligence (AI) and machine learning is a relatively new concept in the world of investment, but it is one that is increasingly gaining traction and its popularity is only set to growth along with technological advances. In a recent report, Thematic Investing: Transforming World, Bank of America Merrill Lynch identified big data and AI as one of the five key themes set to shape our world in the next five years, with areas such as technology, e-commerce and payments seeing particularly strong tailwinds. "The pace at which themes are transforming businesses is unprecedented today, but we believe this rate of technological change will be faster still over the next five years," the report said. "The exponential growth of data (doubling every two-three years), cheapening computing power …and rise of a connected world (Internet of Things, mobile devices, social media) will bring about the fastest transformation in human history." While some industries are already fully embracing artificial intelligence, we are only just beginning to see the impact it can have on financial services.

When artificial intelligence systems start getting creative, they can create great things – and scary ones. Take, for instance, an AI program that let web users compose music along with a virtual Johann Sebastian Bach by entering notes into a program that generates Bach-like harmonies to match them. Run by Google, the app drew great praise for being groundbreaking and fun to play with. It also attracted criticism, and raised concerns about AI's dangers. My study of how emerging technologies affect people's lives has taught me that the problems go beyond the admittedly large concern about whether algorithms can really create music or art in general.

Deep learning is at the core of recent spoken language understanding (SLU) related tasks. More precisely, deep neural networks (DNNs) drastically increased the performances of SLU systems, and numerous architectures have been proposed. In the real-life context of theme identification of telephone conversations, it is common to hold both a human, manual (TRS) and an automatically transcribed (ASR) versions of the conversations. Nonetheless, and due to production constraints, only the ASR transcripts are considered to build automatic classifiers. TRS transcripts are only used to measure the performances of ASR systems. Moreover, the recent performances in term of classification accuracy, obtained by DNN related systems are close to the performances reached by humans, and it becomes difficult to further increase the performances by only considering the ASR transcripts. This paper proposes to distillates the TRS knowledge available during the training phase within the ASR representation, by using a new generative adversarial network called M2H-GAN to generate a TRS-like version of an ASR document, to improve the theme identification performances.

The standard Gibbs sampler of Mixed Multinomial Logit (MMNL) models involves sampling from conditional densities of utility parameters using Metropolis-Hastings (MH) algorithm due to unavailability of conjugate prior for logit kernel. To address this non-conjugacy concern, we propose the application of P\'olygamma data augmentation (PG-DA) technique for the MMNL estimation. The posterior estimates of the augmented and the default Gibbs sampler are similar for two-alternative scenario (binary choice), but we encounter empirical identification issues in the case of more alternatives ($J \geq 3$).

We present a deep learning method for singing voice conversion. The proposed network is not conditioned on the text or on the notes, and it directly converts the audio of one singer to the voice of another. Training is performed without any form of supervision: no lyrics or any kind of phonetic features, no notes, and no matching samples between singers. The proposed network employs a single CNN encoder for all singers, a single WaveNet decoder, and a classifier that enforces the latent representation to be singer-agnostic. Each singer is represented by one embedding vector, which the decoder is conditioned on. In order to deal with relatively small datasets, we propose a new data augmentation scheme, as well as new training losses and protocols that are based on backtranslation. Our evaluation presents evidence that the conversion produces natural signing voices that are highly recognizable as the target singer.