Recent progress in AI and Reinforcement learning has shown great success in solving complex problems with high dimensional state spaces. However, most of these successes have been primarily in simulated environments where failure is of little or no consequence. Most real-world applications, however, require training solutions that are safe to operate as catastrophic failures are inadmissible especially when there is human interaction involved. Currently, Safe RL systems use human oversight during training and exploration in order to make sure the RL agent does not go into a catastrophic state. These methods require a large amount of human labor and it is very difficult to scale up. We present a hybrid method for reducing the human intervention time by combining model-based approaches and training a supervised learner to improve sample efficiency while also ensuring safety. We evaluate these methods on various grid-world environments using both standard and visual representations and show that our approach achieves better performance in terms of sample efficiency, number of catastrophic states reached as well as overall task performance compared to traditional model-free approaches

Multi-task learning, as it is understood nowadays, consists of using one single model to carry out several similar tasks. From classifying hand-written characters of different alphabets to figuring out how to play several Atari games using reinforcement learning, multi-task models have been able to widen their performance range across different tasks, although these tasks are usually of a similar nature. In this work, we attempt to widen this range even further, by including heterogeneous tasks in a single learning procedure. To do so, we firstly formally define a multi-network model, identifying the necessary components and characteristics to allow different adaptations of said model depending on the tasks it is required to fulfill. Secondly, employing the formal definition as a starting point, we develop an illustrative model example consisting of three different tasks (classification, regression and data sampling). The performance of this model implementation is then analyzed, showing its capabilities. Motivated by the results of the analysis, we enumerate a set of open challenges and future research lines over which the full potential of the proposed model definition can be exploited.

In this paper, we present exploitability descent, a new algorithm to compute approximate equilibria in two-player zero-sum extensive-form games with imperfect information, by direct policy optimization against worst-case opponents. We prove that when following this optimization, the exploitability of a player's strategy converges asymptotically to zero, and hence when both players employ this optimization, the joint policies converge to a Nash equilibrium. Unlike fictitious play (XFP) and counterfactual regret minimization (CFR), our convergence result pertains to the policies being optimized rather than the average policies. Our experiments demonstrate convergence rates comparable to XFP and CFR in four benchmark games in the tabular case. Using function approximation, we find that our algorithm outperforms the tabular version in two of the games, which, to the best of our knowledge, is the first such result in imperfect information games among this class of algorithms.

Google has officially announced a major new effort in the video game world -- and it might just change the future of the roughly $135 billion industry. Speaking at the annual Games Development Conference (GDC) in San Francisco on Tuesday, Google CEO Sundar Pichai unveiled Stadia, a long-rumored cloud-based games streaming service. Unlike with a traditional video game console or computer, which processes a game locally as it's being played, Stadia games are processed in the cloud, with the action beamed instantaneously to players over the Internet. Google promises a lag-free experience, as long as you have a fast enough Internet connection. "We are starting our next big challenge: building a game platform for everyone," Pichai said.

That is, if you were fortunate enough to get an invite. While last year's invite-only conference, held in southern California's Palm Springs, produced striking images of Bezos strolling with a robotic dog designed by Boston Dynamics, the CEO this time took to the stage with a flying robo-dragonfly. Much of this year's buzz, however, has come straight from the stars; among the attendees is actor Mark Hamill, who portrayed'Star Wars' protagonists'Luke Skywalker' in the films' original trilogy. Bezos demonstrated a robotic dragon fly on stage that circled around his head. As Hamill, who recently revived Skywalker for the latest iteration of the Star Wars franchise, mingled with guests, the conference's other attendees showcased their newest and most exciting revelations in the fields of robotics, artificial intelligence, machine learning, and more.

Google has taken the wraps off of its new gaming service. Dubbed'Stadia,' the gaming platform operates entirely on the cloud and lets users'instantly' stream games on any device, without the need for pesky downloading. The service is slated to launch later this year in the U.S., U.K. and Canada, with more details about available game titles expected to come in the next few months. Stadia ditches the traditional console; instead, users can play games with their existing laptops, desktops, TVs, tablets or phones, as well as their own keyboard and mouse. No updates, no downloads,' Google said.

The'world's first' AI-powered hearing aid connects to your smartphone and can translate speech into 27 different languages, say its creators. Experts claim to have mastered near-real time translation of Arabic, Japanese and French, among others, by listening for the foreign language and relaying it to the phone. The device is akin to the fictitious alien Babel fish that performs instant translations in comedy science fiction series The Hitchhiker's Guide to the Galaxy. Creators Starkeys claim the technology reduces noisy environments by 50 per cent, while artificial intelligence optimises the hearing experience and its translation is compatible to Google Translate in terms of accuracy. The Livio AI, which is now on sale in the UK and costs £3,000 ($3,900), also has brain tracking technology and Alexa connectivity, and interfaces with the mobile app, Thrive Hearing.

Technology used to map the interior of Icelandic lava tubes could be used to explore deep caves on Mars and the moon, scientists say. 'Space drones', or LiDAR-equipped drones, which stands for Light Detection and Ranging, are currently used to examine the lava tube's shape and its ice formations. The team, from the SETI Institute and Astrobotic Technology, was investigating the Lofthellir Lava Tube Ice Cave in the remote Myvatn region of Iceland. Scientists are hoping to use the same technology used to map the interior of Icelandic lava tubes to explore deep caves on Mars. SETI's mission is to explore and examine the origin and nature of life in the universe.

Scientists have succeeded in creating simple cell-like robots that join together in large groups, move in a coordinated fashion and transport objects. They are able to coordinate their movements, transport objects and even respond to light. Scientists call them'particle robots' but even their creators admit they share similarities with the'grey goo' that prompted a famous warning from the Prince of Wales. Grey goo is a hypothetical end-of-the-world scenario involving molecular nanotechnology in which self-replicating robots consume all biomass on Earth. Scientists have succeeded in creating simple cell-like robots that join together in large groups, move in a coordinated fashion and transport objects.

Google announced its entry into the video game market with Google Stadia, a service that will allow players to stream video games to any screen – phone, tablet, TV or computer. Google announced Stadia at the Game Developers Conference in San Francisco on Tuesday. The cloud-powered service will allow users to log in from any screen using the Chrome browser, a Chromecast device or a Google Pixel phone or tablet and play the same games across all of them, with all the computational heavy-lifting done by Google's servers instead of a games console. It means that players won't have to purchase a box that sits under the TV in order to play, theoretically liberating video games from hardware altogether. Google did not announce pricing, but it is likely that the service will be subscription-based.