In June of last year, five researchers at Facebook's Artificial Intelligence Research unit published an article showing how bots can simulate negotiation-like conversations. While for the most part the bots were able to maintain coherent dialogue, the researchers found that the software agents would occasionally generate strange sentences like: "Balls have zero to me to me to me to me to me to me to me to." On seeing these results, the team realized that they had failed to include a constraint that limited the bots to generating sentences within the parameters of spoken English, meaning that they developed a type of machine-English patois to communicate between themselves. These findings were considered to be fairly interesting by other experts in the field, but not totally surprising or groundbreaking. A month after this initial research was released, Fast Company published an article entitled AI Is Inventing Language Humans Can't Understand.

The great strength of CNN architectures is their capability to automatically learn a hierarchy of feature detectors in order to solve some task. What it has been observed is that regardless of the architecture, the dataset and the target semantic space (and of course the initialization assuming it's random), the first layers seem to always converge to specific kinds of feature detectors: the Gabor Filters. This is actually a very interesting and important phenomenon as it seems to suggest the Gabor Filters are the most efficient way to start the semantic extraction process from an image. It would mean Gabor Filters block is a sort of "generic building block" which could be used to design NN aimed at solving computer vision problems This is one of the main goal of Transfer Learning: finding "building blocks" which can be composed to build a NN and fine tuned, instead of trained from scratch, on the Dataset Being able to properly understand how the CNN specializes while training is important to get to Transfer Learning: "transferring" the network "capability of solving a problem" which means basically adapting its weights properly, to another similar problem in a Data Efficient Way The data efficiency is in fact one of the most important aspects of transfer learning: it is well known that Supervised Learning is an effective way to make a certain, typically big, NN become able to solve a problem but it scales badly in terms of data as it typically requires A LOT OF supervision signal which, in case of manual annotation, is expensive to collect as it relies on humans to provide it. Furthermore the more the task is difficult, the more the annotations need to be provided by human experts instead of normal people and the former one's time is more expensive than the latter ones

The Mystic is designed to give the ultimate aerial video and photography experience, creating breathtaking imagery without the need to learn complicated film techniques. The Mystic automatically detects objects and avoids obstacles using the cutting-edge motion intelligence similarly found in the self-driving car. With gesture interaction, you can take stunning aerial selfies, using poses to control the drone. The Mystic recognizes each pose as a specific command and will follow your instructions, moving forward and backward, side to side, and taking photos. The Mystic is the first drone to support up to 6 different gestures, all of which can be customized to your personal preference.

Recently, Talmor and Berant (2018) introduced ComplexWebQuestions - a dataset focused on answering complex questions by decomposing them into a sequence of simpler questions and extracting the answer from retrieved web snippets. In their work the authors used a pre-trained reading comprehension (RC) model (Salant and Berant, 2018) to extract the answer from the web snippets. In this short note we show that training a RC model directly on the training data of ComplexWebQuestions reveals a leakage from the training set to the test set that allows to obtain unreasonably high performance. As a solution, we construct a new partitioning of ComplexWebQuestions that does not suffer from this leakage and publicly release it. We also perform an empirical evaluation on these two datasets and show that training a RC model on the training data substantially improves state-of-the-art performance.

Modern consumers expect more from brands than ever. As a result of digital proliferation and advances in data science, consumers expect brands to deliver consistent, personalized, high-quality experiences across a growing set of relevant channels. Of course, this presents quite a challenge for brand marketers, who must develop enough on-brand content to address their various audiences -- doing so quickly and at scale, all while measuring the effectiveness of each channel, piece of content and customer interaction. This crunch is making it more difficult for brands to continuously create the content that consumers want to see to make their purchasing decisions. To beat that content crunch, many brands are producing a mix of user-, influencer- and brand-generated content. Of those three, user-generated content, or UGC, is the one that brands have the least control over -- both in terms of creation and curation, but delivers the highest impact.

On the heels of United States President Donald Trump's historic de-nuclearization summit with North Korean leader Kim Jong-un, non-proliferation is once again a timely topic. Since the dawn of the nuclear age, keeping tabs on who has military-grade nuclear capabilities and materials has been a vital – and difficult – task. Thankfully, it's also one that may be getting easier, thanks to leaps forward in fields like data analysis, machine learning and artificial intelligence. Last month, Thomson Reuters Labs was invited to present at a workshop called "Applications of Innovative Tools and Technologies for Nonproliferation and Disarmament" held in Krems, Austria, for diplomats representing their countries at the International Atomic Energy Agency (IAEA) and other international organizations. The diplomatic workshop was preceded by a day-long session for technical participants at the Vienna Center for Disarmament and Non-Proliferation.

There aren't many gadgets that I'm still using five years after I buy them, except for maybe a laptop. Even then, that's getting quite long in the tooth given how quickly upgrades arrive these days. Chromecast and Google Cast are still things that I use multiple times a day, every day. When Google first introduced the Chromecast in 2013, the company promised to make any TV with an HDMI port a smart display with the combination of a thumbdrive-like dongle and your home WiFi. That it did, but in the months that followed, Google expanded the tech undergirding its TV accessory well beyond that $35 device.

The use of natural wood, hand-knitted fabrics, and leather lend a rustic hand-made vibe to the Maverick.Cavalier Cavalier Audio is a New Jersey-based designer and manufacturer of voice-controlled speakers. The company's latest product is Maverick portable Bluetooth and Wi-Fi speaker with Amazon Alexa voice control. Maverick blends a hand-crafted finish and immersive sound to a market that's positively crowded out with'me-too' products produced from mass-produced plastics and so-so design. The Maverick has been designed to stand out from the crowd by using authentic materials, stylish design while incorporating the latest audio technology. Maverick was dreamed up by a team of musicians and engineers with the aim of offering world-class acoustics to life via a 20W stereo speaker system featuring two active drivers and dual passive radiators that the designers claim will produce true room-filling experience.

It's a well-known fact that humans are prone to be racist, but as artificial intelligence evolves, it has become evident that technology is racist too. Recent studies have found flaws in technological "advancements" that disadvantage people of a darker hue. For example, a ProPublica study found that software used to determine prison sentences for criminals (ostensibly to eradicate human bias) is biased against Black people. Similarly, an MIT Media Lab study revealed that certain facial recognition software can't seem to identify women of color. Technology like that and other artificial intelligence innovations rely on algorithms.

According to a Forrester report, robots will eliminate 6 percent of all jobs in the U.S. by 2021. McKinsey's assessment is even more expansive -- they believe that by 2030 one-third of American jobs could become automated. This, however, doesn't mean that life will soon be like "The Jetsons." As technological developments have done in the past, the next generation of robots utilizing artificial intelligence and automation to streamline processes currently handled with the assistance of human workers will significantly alter the job market. This idea represents a form of disruptive innovation, a term that refers to when an emerging technology can utilize fewer resources, thus competing better against those without it.