In this paper, we propose a new learning technique named message-dropout to improve the performance for multi-agent deep reinforcement learning under two application scenarios: 1) classical multi-agent reinforcement learning with direct message communication among agents and 2) centralized training with decentralized execution. In the first application scenario of multi-agent systems in which direct message communication among agents is allowed, the message-dropout technique drops out the received messages from other agents in a block-wise manner with a certain probability in the training phase and compensates for this effect by multiplying the weights of the dropped-out block units with a correction probability. The applied message-dropout technique effectively handles the increased input dimension in multi-agent reinforcement learning with communication and makes learning robust against communication errors in the execution phase. In the second application scenario of centralized training with decentralized execution, we particularly consider the application of the proposed message-dropout to Multi-Agent Deep Deterministic Policy Gradient (MADDPG), which uses a centralized critic to train a decentralized actor for each agent. We evaluate the proposed message-dropout technique for several games, and numerical results show that the proposed message-dropout technique with proper dropout rate improves the reinforcement learning performance significantly in terms of the training speed and the steady-state performance in the execution phase.

A screen shows visitors being filmed by AI (Artificial Intelligence) security cameras with facial recognition technology at the 14th China International Exhibition on Public Safety and Security in Beijing. A Chinese surveillance firm using facial recognition technology left one of its databases exposed online for months, according to a prominent security researcher. A massive database for 2,565,724 people -- with names, ID card number, expiration date, home address, date of birth, nationality, gender, photograph, employer and GPS coordinates of locations -- was left online without authentication, according to a report from ZDNet. Security researcher Victor Gevers, who founded the database, told ZDNet that over a 24-hour period, a steady stream of nearly 6.7 million GPS coordinates was recorded, which means the database was actively tracking Uyghur Muslims as they moved around Xinjiang province in China. HOW AMAZON'S JEFF BEZOS AND THE NATIONAL ENQUIRER WENT TO WAR Human rights groups have said that China is keeping hundreds of thousands of Uyghur Muslims in internment camps, where they are indoctrinated, forced to perform labor and detained.

In the wake of recent customer data breaches, companies are recognizing the need for more protections and transparency around the collection and use of customer data. But few have paid equal attention to the issues arising from the collection and mining of workplace data. Companies have vast amounts of valuable data on work and their workforce, and executives recognize the opportunity to use this data to improve productivity and to motivate and engage people. We surveyed more than 10,000 workers, across all skill levels and generations, and 1400 C-level executives, in 13 countries and 13 industries. We found that more than 90% of the employees are willing to let their employers collect and use data on them and their work, but only if they benefit in some way.

On an isolated stretch of industrial flatland outside Knoxville, Tenn., a minibus is taking shape in a car factory unlike any other. The space is small, the size of a supermarket, and all but tool-free. Instead, perched in the center is the world's largest 3D printer, a gangly 10-by-40-foot behemoth with a steel-gray exterior, thick columnar footings, and derrick-like roof beams to true its frame. When the print heads are in motion, the equipment emits little more than a whisper, dexterously cutting sharp angles and rounded edges. Programmers on laptops and quality-control experts with tablets mill around, inputting design changes and fine-tuning the minibus's sensor instructions. Beyond the assembly room lies a kind of alchemist's playground, where young staffers with advanced degrees in materials science and mechanical engineering synthesize nanopolymers or test exotic particles for strength or thermal and electrical conductivity. The minibus, named Olli, is the latest offbeat product from Local Motors Inc., an 11-year-old startup.

Recently, the Victorian Government brought in new rules stating Victorian state schools will be banned from using facial recognition technology in classrooms unless they have the approval of parents, students and the Department of Education. Students may be justifiably horrified at the thought of being monitored as they move throughout the school during the day. But a roll marking system could be as simple as looking at a tablet or iPad once a day instead of being signed off on a paper roll. It simply depends on the implementation. Trials have already begun in independent schools in NSW and up to 100 campuses across Australia.

At the conceptual level the idea of killer robots is no longer a remote possibility. Already drones and other military machines can be piloted remotely, and some are equipped with missiles. Given the pace of development in relation of autonomous technology for vehicles, and the advances with artificial intelligence in general, the idea of a machine that can directs itself in battle is plausible and, some might argue, inevitable. Inevitable, that is, unless concerted action is taken by governments to ban the development of these types of devices. This would be something similar to the treaty that is in place prohibiting the use of chemical weapons (the 1997 Convention on the Prohibition of the Development, Production, Stockpiling and Use of Chemical Weapons and on their Destruction).

Robotics investments in January 2019 have crossed a minimum of $644 million worldwide, armed with a total of 25 robotics transactions. The $644 million raised in January is lower than the funding into this industry raised in December in tune of $652.7 million. One of the biggest investments in January that is $104 million Series A has been made into the Beijing Auto AI Technology Co. of China. Other notable investments in January 2019 into Robotics include the $100 million JV into Ekso Bionics Holdings Inc. and a $59.61 million Series B funding into China-based NASN Automotive Electronics Co. Here are the Top 10 Investments that ruled the Robotics Technologies space in January 2019.

A machine learning model was used by researchers from the University of Waikato, in New Zealand, to narrow down a massive 8 million tweets to a more manageable 1.2 million in order to look at how te reo Māori is being used in the genre. According to a recent press release, the team focused on 77 Māori loanwords, or te reo Māori words used in an English context, and used them as training data for their machine learning model. Machine learning allows data scientists to provide a computer with a large data set, and teach it to make predictions based on that data. The initial 8 million tweets contained a fair bit of distracting data'noise'. The irrelevant tweets are those that are not used in a New Zealand English context, or were otherwise unrelated.

Knowledge graph (KG) refinement mainly aims at KG completion and correction (i.e., error detection). However, most conventional KG embedding models only focus on KG completion with an unreasonable assumption that all facts in KG hold without noises, ignoring error detection which also should be significant and essential for KG refinement.In this paper, we propose a novel support-confidence-aware KG embedding framework (SCEF), which implements KG completion and correction simultaneously by learning knowledge representations with both triple support and triple confidence. Specifically, we build model energy function by incorporating conventional translation-based model with support and confidence. To make our triple support-confidence more sufficient and robust, we not only consider the internal structural information in KG, studying the approximate relation entailment as triple confidence constraints, but also the external textual evidence, proposing two kinds of triple supports with entity types and descriptions respectively.Through extensive experiments on real-world datasets, we demonstrate SCEF's effectiveness.

Sum-Product Networks (SPNs) are hierarchical probabilistic graphical models capable of fast and exact inference. Applications of SPNs to real-world data such as large image datasets has been fairly limited in previous literature. We introduce Convolutional Sum-Product Networks (ConvSPNs) which exploit the inherent structure of images in a way similar to deep convolutional neural networks, optionally with weight sharing. ConvSPNs encode spatial relationships through local products and local sum operations. ConvSPNs obtain state-of-the-art results compared to other SPN-based approaches on several visual datasets, including color images, for both generative as well as discriminative tasks. ConvSPNs are the first pure-SPN models applied to color images that do not depend on additional techniques for feature extraction. In addition, we introduce two novel methods for regularizing SPNs trained with hard EM. Both regularization methods have been motivated by observing an exponentially decreasing variance of log probabilities with respect to the depth of randomly structured SPNs. We show that our regularization provides substantial further improvements in generative visual tasks.