AI startups experienced their best funding year ever, raising a record $9.33 billion, or nearly 10% of last year's total VC investments that reached $99.5 billion, an 18-year high since the dot-com era.Getty The Artificial Intelligence (AI) winter is definitely over. As venture capital (VC) funding nears record since the dot-com era, with U.S. companies raising $99.5 billion versus $119.6 billion in 2000 according to the latest PwC MoneyTree Report, AI startups also experienced their best year ever, raising a record $9.33 billion, or nearly 10% of last year's total VC investments. Since 2013, VC investments in AI startups had regularly increased over the following four years, with a compound annual growth rate (CAGR) of about 36%. However, AI-related funding significantly jumped last year, increasing 72% compared to 2017, despite a dip in deal activity, with 466 startups funded from 533 in 2017, and after increasing for four years. The report also reveals that seed-stage deal activity among AI-related companies rose to 28% in the fourth-quarter of 2018, compared to 24% in the three months prior, while expansion-stage deal activity jumped to 32%, from 23%.

In 1964, concerns about increasing automation led the federal government to establish the National Commission on Technology, Automation, and Economic Progress. The commission was tasked with studying the impact of technological and economic change. Even more than half a century ago, leaders foresaw a world where technology could lead to a new era of economic prosperity--but only if we met the challenge head on. Our predecessors did just that, and today we recognize the second half of the 20th century as a time of great innovation. No advance has captured our imagination more than artificial intelligence.

You may want to use some extra caution. A new notice released Tuesday by the Federal Trade Commission highlights a surge in'romance scams,' or scenarios where scammers trick love-lusting internet users into sending them money, only to later disappear. The scams cost victims an astonishing $143 million in 2018, up from $33 million the previous year and making it the most costly type of consumer fraud reported to the FTC. A new notice from the Federal Trade Commission highlights a surge in'romance scams,' or scenarios where scammers trick love-lusting internet users into sending them money Romance scams, where criminals create phony profiles to trick love-lusting victims into sending them money, are on the rise. To avoid falling prey, here's what you can do: These romance scams typically involve a user creating a phony profile and approaching someone via a dating app or website.

Teenagers who play violent video games are no more prone to real world aggressive behaviour than their peers, according to UK researchers who say their negative effects have been overstated. Fears that gory games like Grand Theft Auto and Call of Duty might make children think on-screen behaviours are acceptable have been a major concern for parents and policy makers for years. Last year President Donald Trump said violent games were "shaping young people's thoughts" in the wake of the Marjory Stoneman Douglas high school shooting in Parklands, Florida. But one of the most comprehensive studies to date, led by University of Oxford researchers, found no evidence of increased aggression among teens who had spent longer playing violent games in the past month. "The idea that violent video games drive real-world aggression is a popular one, but it hasn't tested very well over time," says lead researcher Professor Andrew Przybylski, director of research at the Oxford Internet Institute.

In the 1960s, the Department of Defense began shoveling money towards a small group of researchers with a then-fringe idea: making machines intelligent. Military money played a central role in establishing a new science--artificial intelligence. Sixty years later, the Pentagon believes AI has matured enough to become a central plank of America's national security. On Tuesday, the department released an unclassified version of its AI strategy, which calls for rapid adoption of AI in all aspects of the US military. The plan depends on the Pentagon working closely with the tech industry to source the algorithms and cloud computing power needed to run AI projects.

We build new test sets for the CIFAR-10 and ImageNet datasets. Both benchmarks have been the focus of intense research for almost a decade, raising the danger of overfitting to excessively re-used test sets. By closely following the original dataset creation processes, we test to what extent current classification models generalize to new data. We evaluate a broad range of models and find accuracy drops of 3% - 15% on CIFAR-10 and 11% - 14% on ImageNet. However, accuracy gains on the original test sets translate to larger gains on the new test sets. Our results suggest that the accuracy drops are not caused by adaptivity, but by the models' inability to generalize to slightly "harder" images than those found in the original test sets.

In a resource-constrained, contested environment, computing resources need to be aware of possible size, weight, and power (SWaP) restrictions. SWaP-aware computational efficiency depends upon optimization of computational resources and intelligent time versus efficiency tradeoffs in decision making. In this paper we address the complexity of various optimization strategies related to low SWaP computing. Due to these restrictions, only a small subset of less complicated and fast computable algorithms can be used for tactical, adaptive computing.

We study epidemic forecasting on real-world health data by a graph-structured recurrent neural network (GSRNN). We achieve state-of-the-art forecasting accuracy on the benchmark CDC dataset. To improve model efficiency, we sparsify the network weights via transformed-$\ell_1$ penalty and maintain prediction accuracy at the same level with 70% of the network weights being zero.

Cooperation information sharing is important to theories of human learning and has potential implications for machine learning. Prior work derived conditions for achieving optimal Cooperative Inference given strong, relatively restrictive assumptions. We relax these assumptions by demonstrating convergence for any discrete joint distribution, robustness through equivalence classes and stability under perturbation, and effectiveness by deriving bounds from structural properties of the original joint distribution. We provide geometric interpretations, connections to and implications for optimal transport, and connections to importance sampling, and conclude by outlining open questions and challenges to realizing the promise of Cooperative Inference.

Machine learning algorithms have exciting and wide-range potential. However, as the data frequently containsensitive personal information, there are real privacy concerns associated with the development and the deployment of this technology. Motivated by this observation, the line of work on differentially private learning (initiated by [23]) aims to construct learning algorithms that provide strong (mathematically proven) privacy protections for the training data. Both government agenciesand industrial companies have realized the importance of introducing strong privacy protection to statistical and machine learning tasks. A few recent examples include Google [20] and Apple [27] that are already using differentially private estimation algorithms that feed into machine learning algorithms, and the US Census Bureau announcement that they will use differentially privatedata publication techniques in the next decennial census [1]. Differential privacy is increasingly accepted as a standard for rigorous privacy. We refer the reader to the excellent surveys in [17] and [28]. The definition of differential privacy is, Definition 1.1 ([16]). Let A be a randomized algorithm whose input is a sample.