The ability to integrate information in the brain is considered to be an essential property for cognition and consciousness. Integrated Information Theory (IIT) hypothesizes that the amount of integrated information ($\Phi$) in the brain is related to the level of consciousness. IIT proposes that to quantify information integration in a system as a whole, integrated information should be measured across the partition of the system at which information loss caused by partitioning is minimized, called the Minimum Information Partition (MIP). The computational cost for exhaustively searching for the MIP grows exponentially with system size, making it difficult to apply IIT to real neural data. It has been previously shown that if a measure of $\Phi$ satisfies a mathematical property, submodularity, the MIP can be found in a polynomial order by an optimization algorithm. However, although the first version of $\Phi$ is submodular, the later versions are not. In this study, we empirically explore to what extent the algorithm can be applied to the non-submodular measures of $\Phi$ by evaluating the accuracy of the algorithm in simulated data and real neural data. We find that the algorithm identifies the MIP in a nearly perfect manner even for the non-submodular measures. Our results show that the algorithm allows us to measure $\Phi$ in large systems within a practical amount of time.

In meta-learning an agent extracts knowledge from observed tasks, aiming to facilitate learning of novel future tasks. Under the assumption that future tasks are 'related' to previous tasks, representations should be learned in a way which captures the common structure across learned tasks, while allowing the learner sufficient flexibility to adapt to novel aspects of new tasks. We present a framework for meta-learning that is based on generalization error bounds, allowing us to extend various PAC-Bayes bounds to meta-learning. Learning takes place through the construction of a distribution over hypotheses based on the observed tasks, and its utilization for learning a new task. Thus, prior knowledge is incorporated through setting an experience-dependent prior for novel tasks. We develop a gradient-based algorithm which minimizes an objective function derived from the bounds and demonstrate its effectiveness numerically with deep neural networks. In addition to establishing the improved performance available through meta-learning, we demonstrate the intuitive way by which prior information is manifested at different levels of the network.

Imagine a situation where a group of adversaries is preparing an attack on the United States or U.S. interests. An intelligence analyst has observed some signals, but the situation is rapidly changing. The analyst faces the decision to alert a principal decision maker that an attack is imminent, or to wait until more is known about the situation. This warning decision is based on the analyst's observation and evaluation of signals, independent or correlated, and on her updating of the prior probabilities of possible scenarios and their outcomes. The warning decision also depends on the analyst's assessment of the crisis' dynamics and perception of the preferences of the principal decision maker, as well as the lead time needed for an appropriate response. This article presents a model to support this analyst's dynamic warning decision. As with most problems involving warning, the key is to manage the tradeoffs between false positives and false negatives given the probabilities and the consequences of intelligence failures of both types. The model is illustrated by revisiting the case of the attack on Pearl Harbor in December 1941. It shows that the radio silence of the Japanese fleet carried considerable information (Sir Arthur Conan Doyle's "dog in the night" problem), which was misinterpreted at the time. Even though the probabilities of different attacks were relatively low, their consequences were such that the Bayesian dynamic reasoning described here may have provided valuable information to key decision makers.

The world is getting older. By 2050, the global population of those age 65 and older is projected to nearly double to 1.6 billion. This global graying has given birth to a new phrase: "super-aging." A nation is said to be super-aged when more than 1 in 5 of its people are 65 and older. The United States isn't there yet, but the trend line points that way in decades to come.

For many, artificial intelligence (AI) and the human workforce are at odds. These people are concerned that intelligent machines powered by increasingly sophisticated AI will take over human jobs, leaving some people with no source of income. Even more frightening is the prospect of a complete AI labor takeover if/when we reach the technological singularity. According to Sebastian Thurn, co-founder of Google's secret X laboratory, they may be worrying over nothing. During a talk at the ongoing World Government Summit in Dubai, Thurn said he believes AI will make humans into "superhuman workers" capable of doing more with the help of technology than without it, reported CNBC.

The rise of artificial intelligence (AI) has been quite remarkable in the past few years. Many experts predict that, like the Internet, AI will bring huge changes to our lives and be part of almost everything we do. Vast sums of money have been put into AI start-ups in China, and many existing tech companies have opened new research labs. But amid the optimistic atmosphere, many also have concerns about the ethical questions that artificial intelligence raises. Get the entire 10-part series on Warren Buffett in PDF.

While Alpine skiers fought high winds at the Pyeongchang Games today, there were no such problems for robots competing in their own'Olympics' ski challenge. Robots of all shapes and sizes skied, and in some cases tumbled, down a course at the Welli Hilli ski resort, an hour's drive west of Pyeongchang. Eight robotics teams from universities, institutes and a private company competed for a $10,000 (£7,240) prize in the Ski Robot Challenge. 'I heard the Alpine skiing has been postponed again due to wind conditions. That's a pity,' said Lee Sok-min, a member of the winning TAEKWAN-V team.

A DRC-Hubo robot prepares to light the torch of bearer Dr. Dennis Hong, a professor at the University of California, Los Angeles, during the Olympic Torch Relay in Daejeon, South Korea. The 2018 Pyeongchang Winter Olympics will be held from Feb. 9 to 25, 2018, in South Korea. PYEONGCHANG, South Korea -- This country likes robots. There are robots that vacuum the floor in the press center. A robot carried the Olympic flame.

OVER the past few years, the media has focused a lot on how companies the likes of Uber, Grab and Airbnb have exploited the so-called'sharing economy,' highlighting the fact that they don't own assets but are still able to dominate the transportation and accommodation industries respectively. But as these companies hog the headlines insofar as their valuations and growth rates are concerned, unbeknownst to many, they are also highly-tuned, data intensive-driven companies. Being data driven means that they are able to pour resources into helping to solve one of society's most challenging problems – that of traffic congestions in the countries they operate in. Last April, Southeast Asian-based (SEA) Grab launched the OpenTraffic initiative in Malaysia, an effort aimed at providing traffic data from Grab's GPS data streams to address traffic congestion and improve road safety in major Malaysian cities. The initiative was done in collaboration with the Malaysia Digital Economy Corporation Sdn Bhd (MDEC) and the World Bank Group.

The first thing you notice about Sophia, a robot, is the sound she makes. Like a late-90s Pentium processor struggling to load a video, she makes a whirring squeal that fills the room. But what I'll always remember about my recent conversation with a robot is her stare. I recently talked with Sophia in a cramped room at the Venetian Hotel in Las Vegas. The artificially intelligent humanoid was there for CES.