The last decade has seen a significant increase of interest in deep learning research, with many public successes that have demonstrated its potential. As such, these systems are now being incorporated into commercial products. With this comes an additional challenge: how can we build AI systems that solve tasks where there is not a crisp, well-defined specification? While multiple solutions have been proposed, in this competition we focus on one in particular: learning from human feedback. Rather than training AI systems using a predefined reward function or using a labeled dataset with a predefined set of categories, we instead train the AI system using a learning signal derived from some form of human feedback, which can evolve over time as the understanding of the task changes, or as the capabilities of the AI system improve. The MineRL BASALT competition aims to spur forward research on this important class of techniques. We design a suite of four tasks in Minecraft for which we expect it will be hard to write down hardcoded reward functions. These tasks are defined by a paragraph of natural language: for example, "create a waterfall and take a scenic picture of it", with additional clarifying details. Participants must train a separate agent for each task, using any method they want. Agents are then evaluated by humans who have read the task description. To help participants get started, we provide a dataset of human demonstrations on each of the four tasks, as well as an imitation learning baseline that leverages these demonstrations. Our hope is that this competition will improve our ability to build AI systems that do what their designers intend them to do, even when the intent cannot be easily formalized. Besides allowing AI to solve more tasks, this can also enable more effective regulation of AI systems, as well as making progress on the value alignment problem.

Pricing decisions are increasingly made by AI. Thanks to their ability to train with live market data while making decisions on the fly, deep reinforcement learning algorithms are especially effective in taking such pricing decisions. In e-commerce scenarios, multiple reinforcement learning agents can set prices based on their competitor's prices. Therefore, research states that agents might end up in a state of collusion in the long run. To further analyze this issue, we build a scenario that is based on a modified version of a prisoner's dilemma where three agents play the game of rock paper scissors. Our results indicate that the action selection can be dissected into specific stages, establishing the possibility to develop collusion prevention systems that are able to recognize situations which might lead to a collusion between competitors. We furthermore provide evidence for a situation where agents are capable of performing a tacit cooperation strategy without being explicitly trained to do so.

As pairwise ranking becomes broadly employed for elections, sports competitions, recommendations, and so on, attackers have strong motivation and incentives to manipulate the ranking list. They could inject malicious comparisons into the training data to fool the victim. Such a technique is called poisoning attack in regression and classification tasks. In this paper, to the best of our knowledge, we initiate the first systematic investigation of data poisoning attacks on pairwise ranking algorithms, which can be formalized as the dynamic and static games between the ranker and the attacker and can be modeled as certain kinds of integer programming problems. To break the computational hurdle of the underlying integer programming problems, we reformulate them into the distributionally robust optimization (DRO) problems, which are computationally tractable. Based on such DRO formulations, we propose two efficient poisoning attack algorithms and establish the associated theoretical guarantees. The effectiveness of the suggested poisoning attack strategies is demonstrated by a series of toy simulations and several real data experiments. These experimental results show that the proposed methods can significantly reduce the performance of the ranker in the sense that the correlation between the true ranking list and the aggregated results can be decreased dramatically.

Transfer-based adversarial attacks can effectively evaluate model robustness in the black-box setting. Though several methods have demonstrated impressive transferability of untargeted adversarial examples, targeted adversarial transferability is still challenging. The existing methods either have low targeted transferability or sacrifice computational efficiency. In this paper, we develop a simple yet practical framework to efficiently craft targeted transfer-based adversarial examples. Specifically, we propose a conditional generative attacking model, which can generate the adversarial examples targeted at different classes by simply altering the class embedding and share a single backbone. Extensive experiments demonstrate that our method improves the success rates of targeted black-box attacks by a significant margin over the existing methods -- it reaches an average success rate of 29.6\% against six diverse models based only on one substitute white-box model in the standard testing of NeurIPS 2017 competition, which outperforms the state-of-the-art gradient-based attack methods (with an average success rate of $<$2\%) by a large margin. Moreover, the proposed method is also more efficient beyond an order of magnitude than gradient-based methods.

The Foundation for the Defense of Democracies issues an alarming report about Beijing's expanding tentacles in international agencies; Eric Shawn has the Fox News exclusive. Two astronauts on Sunday made the first spacewalk outside China's new orbital station to set up cameras and other equipment using a 15-meter-long (50-foot-long) robotic arm. Liu Boming and Tang Hongbo were shown by state TV climbing out of the airlock as Earth rolled past below them. The third crew member, commander Nie Haisheng, stayed inside. Liu and Tang spent nearly seven hours outside the station, the Chinese space agency said.

The Defense Advanced Research Projects Agency (DARPA) has selected three companies to work on a research program that could dramatically lower the amount of bandwidth used by networks of cameras. The program, which is called Fast Event-based Neuromorphic Camera and Electronics (FENCE), aims to develop cameras that sense motion, but which are able also to determine what motion is important and represents a threat. According to the agency, the new technology could reduce latency and the usage of network capacity by transmitting only necessary information. Raytheon, BAE Systems and Northrup Grumman are the three defense contractors that have been chosen to work on the research project. Neuromorphic computer systems refers to silicon circuits that mimic brains.

Astronauts at China's new space station conducted their first spacewalk Sunday, state media reported, as Beijing presses on with its extraterrestrial ambitions. It was only the second time the country's astronauts have stepped out of their craft while in space. Three Chinese spacemen blasted off in June, docking at the Tiangong station where they are to remain for three months in China's longest crewed mission to date. On Sunday morning, two of them exited the core cabin, said state broadcaster CCTV. The first, Liu Boming, was transported via mechanical arm to a work site and the other, Tang Hongbo, moved by climbing on the outside of the cabin.

He is the celebrated deep-sea explorer who discovered the Titanic, as well as the German battleship Bismarck and other historic sunken vessels around the world. Now Dr Robert Ballard is pioneering cutting-edge technology – autonomous underwater vehicles that will "revolutionise" the search for more than three million shipwrecks that lie scattered across ocean floors, according to a Unesco estimate. Many will offer new insights into life on board at the time of sinking, hundreds or even thousands of years ago. "We're going to be finding them like crazy," Ballard told the Observer. "It's going to be rapid discovery because of this technology. New chapters of human history are to be read. "All the work I've done in the past in archaeology used vehicles that were connected to a ship.

Competition and Collaboration are parallel themes in international relations. However, the dominant paradigm of the 21st century is going to be competition. One of the major areas of this paradigm is going to be technology. National security structures in the west reflect the rise of emerging military technologies. The intensifying technology competition is also making inroads and avenues for purposeful collaboration among like-minded partners whether bilaterally or at multi-lateral institutions. Cheap computing power, easy availability of data and enabling accessible algorithms have made the rise of artificial intelligence possible.

The robot revolution is always allegedly just around the corner. In the utopian vision, technology emancipates human labor from repetitive, mundane tasks, freeing us to be more productive and take on more fulfilling work. In the dystopian vision, robots come for everyone's jobs, put millions and millions of people out of work, and throw the economy into chaos. Such a warning was at the crux of Andrew Yang's ill-fated presidential campaign, helping propel his case for universal basic income that he argued would become necessary when automation left so many workers out. It's the argument many corporate executives make whenever there's a suggestion they might have to raise wages: $15 an hour will just mean machines taking your order at McDonald's instead of people, they say. But we often spend so much time talking about the potential for robots to take our jobs that we fail to look at how they are already changing them -- sometimes for the better, but sometimes not.