Google has dropped out of the running for JEDI, the massive Defense Department cloud computing contract potentially worth $10 billion. In a statement to Bloomberg, Google said that it decided not to participate in the bidding process, which ends this week, because the contract may not align with the company's principles for how artificial intelligence should be used. In statement to Bloomberg, Google spokesperson said "We are not bidding on the JEDI contract because first, we couldn't be assured that it would align with our AI Principles. And second, we determined that there were portions of the contract that were out of scope with our current government certifications," adding that Google is still "working to support the U.S. government with our cloud in many ways." Officially called Joint Enterprise Defense Infrastructure, bidding for the initiative's contract began two months ago and closes this week.

According to mainstream video games, modern warfare is all about cyborg arms, laser shields and jarheads blowing up baddies under the guidance of recognisable character actors. However, the frenetic antics of the Call of Duty series and its ilk are behind the times. The drone pilot protagonist of 2012's free indie game Unmanned is a more accurate representation of a modern soldier: a man who plays video games with his son every weekend, and who has also killed countless foreigners from a grey-walled cubicle in Nevada. You play an American warrior, square of jaw and beefy of build, who works from an office out in the desert. A click of his mouse sends tons of missile plummeting from anonymous drone planes with an eerie blank space where you'd expect to see a cockpit.

Technology and innovation company Lockheed Martin Australia has become the first Foundation Partner with the University of Adelaide's new Australian Institute for Machine Learning. The strategic partnership will deliver world-leading machine learning research for national security, the space industry, business, and the broader community. Machine learning is a form of artificial intelligence that enables computers and machines to learn how to do complex tasks without being programmed by humans. This technology is driving what is known as the "fourth industrial revolution". The University's new Australian Institute for Machine Learning (AIML) – which builds on decades of expertise in artificial intelligence and computer vision – will be based in the South Australian Government's new innovation precinct at Lot Fourteen (the old Royal Adelaide Hospital site).

Bank of America Chief Operations and Technology Officer Cathy Bessant lent her voice to a drum beat in Washington last week, saying that blockchain and AI have years to go before they have significant benefits for consumers and investors. Blockchain and artificial intelligence aren't ready for primetime, financial leaders were told in Washington last week. At conferences for SIFMA, the securities industry trade group, and the Commodity Futures Trading Commission, speakers were pretty much in agreement that fintech's promises will take years to become everyday realities. "[Blockchain/distributed ledger technology] is not as impactful as the hype," asserted Bank of America Chief Operations and Technology Officer Cathy Bessant at the SIFMA annual conference. She pointed to DLT as a promising technology waiting for proven benefits for consumers, businesses and regulators to cash in.

The resulting collision avoidance strategy can be represented as a numeric table. This methodology has been used in the development of the Airborne Collision Avoidance System X (ACAS X) family of collision avoidance systems for manned and unmanned aircraft, but the high dimensionality of the state space leads to very large tables. To improve storage efficiency, a deep neural network is used to approximate the table. With the use of an asymmetric loss function and a gradient descent algorithm, the parameters for this network can be trained to provide accurate estimates of table values while preserving the relative preferences of the possible advisories for each state. By training multiple networks to represent subtables, the network also decreases the required runtime for computing the collision avoidance advisory. Simulation studies show that the network improves the safety and efficiency of the collision avoidance system. Because only the network parameters need to be stored, the required storage space is reduced by a factor of 1000, enabling the collision avoidance system to operate using current avionics systems.

Many deep learning models are vulnerable to the adversarial attack, i.e., imperceptible but intentionally-designed perturbations to the input can cause incorrect output of the networks. In this paper, using information geometry, we provide a reasonable explanation for the vulnerability of deep learning models. By considering the data space as a non-linear space with the Fisher information metric induced from a neural network, we first propose an adversarial attack algorithm termed one-step spectral attack (OSSA). The method is described by a constrained quadratic form of the Fisher information matrix, where the optimal adversarial perturbation is given by the first eigenvector, and the model vulnerability is reflected by the eigenvalues. The larger an eigenvalue is, the more vulnerable the model is to be attacked by the corresponding eigenvector. Taking advantage of the property, we also propose an adversarial detection method with the eigenvalues serving as characteristics. Both our attack and detection algorithms are numerically optimized to work efficiently on large datasets. Our evaluations show superior performance compared with other methods, implying that the Fisher information is a promising approach to investigate the adversarial attacks and defenses.

Teams of autonomous unmanned aircraft can be used to monitor wildfires, enabling firefighters to make informed decisions. However, controlling multiple autonomous fixed-wing aircraft to maximize forest fire coverage is a complex problem. The state space is high dimensional, the fire propagates stochastically, the sensor information is imperfect, and the aircraft must coordinate with each other to accomplish their mission. This work presents two deep reinforcement learning approaches for training decentralized controllers that accommodate the high dimensionality and uncertainty inherent in the problem. The first approach controls the aircraft using immediate observations of the individual aircraft. The second approach allows aircraft to collaborate on a map of the wildfire's state and maintain a time history of locations visited, which are used as inputs to the controller. Simulation results show that both approaches allow the aircraft to accurately track wildfire expansions and outperform an online receding horizon controller. Additional simulations demonstrate that the approach scales with different numbers of aircraft and generalizes to different wildfire shapes.

Abstract--With the rapid development of knowledge base, question answering based on knowledge base has been a hot research issue. In this paper, we focus on answering singlerelation factoid questions based on knowledge base. We build a question answering system and study the effect of context information on fact selection, such as entity's notable type, outdegree. Experimental results show that context information can improve the result of simple question answering. Question answering (QA) is a classic natural language processing task, which aims at building systems that automatically answer questions formulated in natural language [1]. In recent years, several large-scale general purpose knowledge bases (KBs) have been constructed, including Freebase [2], YAGO [3], DBpedia [4] and Wikidata [5] .

Google will shut down the consumer version of its social network Google after announcing data from up to 500,000 users may have been exposed to external developers by a bug that was present for more than two years in its systems. The company said in a blog that it had discovered and patched the leak in March of this year and had no evidence of misuse of user data or that any developer was aware or had exploited the vulnerability. Shares of its parent company Alphabet Inc, however, were down 1.5 per cent at $1150.75 (£878.71) in response to what was the latest in a run of privacy issues to hit the United States' big tech companies. Google said it had reviewed the issue, looking at the type of data involved, whether it could accurately identify the users to inform, whether there was any evidence of misuse, and whether there were any actions a developer or user could take. "None of these thresholds were met in this instance," it said.

The technology to create killer robots is already here and needs to be banned, a leading artificial intelligence scientist has warned. Stuart Russell, a professor of computer science at Berkeley University, California, said "allowing machines to choose to kill humans" would be "devastating" for world peace and security. The professor, who has worked in the field of artificial intelligence (AI) for more than 35 years, also warned that the window to ban lethal robots was "closing fast". His warning comes as campaigners are making the case at the United Nations (UN) this week for a global prohibition on lethal autonomous weapons systems. Yesterday the pressure group, the Campaign to Stop Killer Robots, showed a short film it produced to a meeting of countries participating in the Convention on Conventional Weapons, which painted a dystopian scenario based on existing technologies.