Obsessively playing violent video games such as Call of Duty for hours every day for several years does not necessarily reduce the amount of empathy people feel towards others in the long term, according to a new study. Researchers in Germany showed "emotionally provocative images" to 15 avid gamers while their brains were being monitored by an MRI machine. They found there was no difference between the response patterns of the gamers' minds compared to a control group. To avoid short-term effects, the gamers had not played games for at least three hours before the test. This may have been a challenge for some of them as, on average, they had played video games for four hours a day every day for four years.

Led by Fei-Fei Li, the director of the Stanford University artificial intelligence lab and a newly minted Google employee, a team of academics recently explored a new way of tracking socioeconomic trends across the US. Rather than knocking on doors and asking questions, they pulled more than 50 million photos from Google Street View and fed them into neural networks. Simply by identifying the make, model, and year of automobiles appearing in the photos, the researchers said, their tech could accurately estimate the income, race, education, and voting patterns of citizens in particular precincts. If the number of sedans on a short stretch of road exceeded the number pickup trucks, for instance, they found that a city was 88 percent likely to vote for a Democrat during the next presidential election. If pickups exceeded sedans, a city was 82 percent likely vote Republican.

As we all push forward with the development of artificial intelligence (AI) solutions, software developers and data scientists increasingly want to leverage deep learning frameworks. To back up a bit, deep learning is a type of machine learning that can enable more complex solutions based on evaluation of abstractions of data. Scaling through added layers and processing, deep learning can build in aggregate from user input and experiences, much the way people learn. Deep learning frameworks enable algorithms to continually improve their performance on complex tasks like speech and image recognition. To get on this path to a new generation of AI solutions, developers and data scientists need to find ways to reduce the steep learning curve that comes with the deployment and configuration of deep learning frameworks. Then, find ways to accelerate the development, training, and deployment of models.

Nano, a Nature Research solution, has just been given a boost by artificial intelligence (AI), increasing the information available to users by sourcing data from over 400,000 relevant research papers. Launched in June 2016, Nano offers highly-indexed and structured information on nanotechnology. It first provided over 200,000 summaries of nanomaterials, containing information on properties, synthesis and applications. These were and continue to be drawn from 30 high-impact journals from all publishers and curated by experts from across this multidisciplinary field. The scale-up, announced today, will significantly increase the breadth of data available.

Airbnb wants to mold its hosts into a powerful organizing force, akin to a union, to advocate on its behalf with local governments around the world and to serve as an ideological rebuke to the advances of AI at other tech firms. As part of that effort to increase engagement with hosts, CEO Brian Chesky announced today that he is embarking on a world tour, forming a host advisory board that will provide feedback to the company and sit in on one of its four annual board meetings, and do monthly check-ins with Airbnb users via Facebook Live. "I want to be held accountable to the community," Chesky, who is modifying his title to CEO and head of community, told a group of hosts gathered at Airbnb HQ. "It's incredibly important because when we sit in a room trying to make decisions, we want to make sure we're doing it for the community, not to the community." Chesky will visit London, New York, Cape Town, Delhi, and Beijing to meet with hosts over the next couple of weeks, and hinted that more changes are coming to improve customer service and host experience. Putting hosts front and center is part of Airbnb's business strategy -- after all, the company relies on people to list their homes for rent -- but it also hints at Airbnb's transition into political advocacy.

Germany and Japan, March 8, 2017 --- Deutsches Forschungszentrum für Künstliche Intelligenz (German Research Center for Artificial Intelligence, 'DFKI') and Hitachi, Ltd. (Hitachi) today announced the joint development of AI (artificial intelligence) technology for human activity recongnition of workers using wearable devices. The AI technology performs real-time recognition of workers' activities by integrating technology in eye-tracking glasses*1to recognize gazed objects with technology in armband devices to recognize action. The recognition ability of each activity is achieved by having the AI understand the tools or parts used at the production site as well as anticipated actions through Deep Learning*2. DFKI and Hitachi will use this newly developed AI technology to assist operations and prevent human error, to contribute to enhancing quality and efficiency on the front line of manufacturing. In line with initiatives such as Industry 4.0*3in Germany and Society 5.0*4in Japan, the manufacturing industry is accelerating steps towards innovating production using AI and robotics, and the automation of menial tasks.

First they figured out how to play checkers and backgammon. Then they mastered chess, Go, "Jeopardy!" and even a few Atari video games. Now computers can challenge humans at the poker table -- and win. DeepStack, a software program developed at the University of Alberta's Computer Poker Research Group, took on 33 professional poker players in more than 44,000 hands of Texas hold'em. Overall, the program won by a significantly higher margin than if it had simply folded in each round, according to a new study in Science.

We saw a transition in the financial ecosystem from the Keynesian "animal spirits" of desk traders and investors to "microbe spirits" by computer algorithms simpler than animals or humans -- but just much, much, much faster. To understand these new phenomena we need a complexity approach. Econophysicists and financial gurus in complexity are no more in the fringe. What our work suggests is that since this war is already under way, and because from our paper it seems to correlate very well with observable crashes in the markets over timescales such as months, then any government watchdog/regulator charged with controlling their financial system also ought to have its own'troops on the ground'.» Not that this is wrong, but you can see that this level of competition would be very ferocious -- which is what happens in the markets.»

For an autonomous agent, executing a poor policy may be costly or even dangerous. For such agents, it is desirable to determine confidence interval lower bounds on the performance of any given policy without executing said policy. Current methods for exact high confidence off-policy evaluation that use importance sampling require a substantial amount of data to achieve a tight lower bound. Existing model-based methods only address the problem in discrete state spaces. Since exact bounds are intractable for many domains we trade off strict guarantees of safety for more data-efficient approximate bounds. In this context, we propose two bootstrapping off-policy evaluation methods which use learned MDP transition models in order to estimate lower confidence bounds on policy performance with limited data in both continuous and discrete state spaces. Since direct use of a model may introduce bias, we derive a theoretical upper bound on model bias for when the model transition function is estimated with i.i.d. trajectories. This bound broadens our understanding of the conditions under which model-based methods have high bias. Finally, we empirically evaluate our proposed methods and analyze the settings in which different bootstrapping off-policy confidence interval methods succeed and fail.

Consider a regression problem where there is no labeled data and the only observations are the predictions $f_i(x_j)$ of $m$ experts $f_{i}$ over many samples $x_j$. With no knowledge on the accuracy of the experts, is it still possible to accurately estimate the unknown responses $y_{j}$? Can one still detect the least or most accurate experts? In this work we propose a framework to study these questions, based on the assumption that the $m$ experts have uncorrelated deviations from the optimal predictor. Assuming the first two moments of the response are known, we develop methods to detect the best and worst regressors, and derive U-PCR, a novel principal components approach for unsupervised ensemble regression. We provide theoretical support for U-PCR and illustrate its improved accuracy over the ensemble mean and median on a variety of regression problems.