HARI SREENIVASAN: But first: Can robotics and artificial intelligence help improve that rush hour commute you're facing? Experts at Carnegie Mellon University think they can by monitoring traffic flow in real time. Jeffrey Brown has the story from Pittsburgh, part of our weekly series on the Leading Edge of science and technology. JEFFREY BROWN: You know the frustration. You're late for work or to pick up your child.

If you believe everything you read, you are probably quite worried about the prospect of a superintelligent, killer AI. The Guardian, a British newspaper, warned recently that "we're like children playing with a bomb," and a recent Newsweek headline reads, "Artificial Intelligence Is Coming, and It Could Wipe Us Out." Numerous such headlines, fueled by comments from the likes of Elon Musk and Stephen Hawking, are strongly influenced by the work of one man: professor Nick Bostrom, author of the philosophical treatise Superintelligence: Paths, Dangers, and Strategies. Bostrom is an Oxford philosopher, but quantitative assessment of risks is the province of actuarial science. He may be dubbed the world's first prominent "actuarial philosopher," though the term seems an oxymoron given that philosophy is an arena for conceptual arguments, and risk assessment is a data-driven statistical exercise. So what do the data say?

The last decade has seen a surge of interest in adaptive learning algorithms for data stream classification, with applications ranging from predicting ozone level peaks, learning stock market indicators, to detecting computer security violations. In addition, a number of methods have been developed to detect concept drifts in these streams. Consider a scenario where we have a number of classifiers with diverse learning styles and different drift detectors. Intuitively, the current 'best' (classifier, detector) pair is application dependent and may change as a result of the stream evolution. Our research builds on this observation. We introduce the $\mbox{Tornado}$ framework that implements a reservoir of diverse classifiers, together with a variety of drift detection algorithms. In our framework, all (classifier, detector) pairs proceed, in parallel, to construct models against the evolving data streams. At any point in time, we select the pair which currently yields the best performance. We further incorporate two novel stacking-based drift detection methods, namely the $\mbox{FHDDMS}$ and $\mbox{FHDDMS}_{add}$ approaches. The experimental evaluation confirms that the current 'best' (classifier, detector) pair is not only heavily dependent on the characteristics of the stream, but also that this selection evolves as the stream flows. Further, our $\mbox{FHDDMS}$ variants detect concept drifts accurately in a timely fashion while outperforming the state-of-the-art.

Obtaining enough labeled data to robustly train complex discriminative models is a major bottleneck in the machine learning pipeline. A popular solution is combining multiple sources of weak supervision using generative models. The structure of these models affects training label quality, but is difficult to learn without any ground truth labels. We instead rely on these weak supervision sources having some structure by virtue of being encoded programmatically. We present Coral, a paradigm that infers generative model structure by statically analyzing the code for these heuristics, thus reducing the data required to learn structure significantly. We prove that Coral's sample complexity scales quasilinearly with the number of heuristics and number of relations found, improving over the standard sample complexity, which is exponential in $n$ for identifying $n^{\textrm{th}}$ degree relations. Experimentally, Coral matches or outperforms traditional structure learning approaches by up to 3.81 F1 points. Using Coral to model dependencies instead of assuming independence results in better performance than a fully supervised model by 3.07 accuracy points when heuristics are used to label radiology data without ground truth labels.

Improving speech system performance in noisy environments remains a challenging task, and speech enhancement (SE) is one of the effective techniques to solve the problem. Motivated by the promising results of generative adversarial networks (GANs) in a variety of image processing tasks, we explore the potential of conditional GANs (cGANs) for SE, and in particular, we make use of the image processing framework proposed by Isola et al. [1] to learn a mapping from the spectrogram of noisy speech to an enhanced counterpart. The SE cGAN consists of two networks, trained in an adversarial manner: a generator that tries to enhance the input noisy spectrogram, and a discriminator that tries to distinguish between enhanced spectrograms provided by the generator and clean ones from the database using the noisy spectrogram as a condition. We evaluate the performance of the cGAN method in terms of perceptual evaluation of speech quality (PESQ), short-time objective intelligibility (STOI), and equal error rate (EER) of speaker verification (an example application). Experimental results show that the cGAN method overall outperforms the classical short-time spectral amplitude minimum mean square error (STSA-MMSE) SE algorithm, and is comparable to a deep neural network-based SE approach (DNN-SE).

The House just passed a bipartisan bill to encourage autonomous vehicles testing. On Wednesday, the House of Representatives did something that's woefully uncommon these days: It passed a bill with bipartisan support. The bill, called the SELF DRIVE Act, lays out a basic federal framework for autonomous vehicle regulation, signaling that federal lawmakers are finally ready to think seriously about self-driving cars and what they mean for the future of the country. "With this legislation, innovation can flourish without the heavy hand of government," said Representative Bob Latta, the Ohio Republican who heads up the Digital Commerce and Consumer Protection Subcommittee, in a floor speech just before the SELF DRIVE Act passed by a two-thirds majority. The Senate will need to pass its own bill before the legislative framework can become law.

Here's how The Donald Test works, an Artificial Intelligence has read through thousands of Hillary Clinton and Donald Trump tweets, learned how they sound ( stuff like tone, language and sentiment) and is able to detect if YOUR tweets sound more like Hillary or Donald. It also gives you a percentage of your closeness to either one them.

The House voted Wednesday to speed the introduction of self-driving cars by giving the federal government authority to exempt automakers from safety standards not applicable to the technology, and to permit deployment of up to 100,000 of the vehicles annually over the next several years. The bill was passed by a voice vote, and now goes to the Senate. State and local officials have raised concern that it limits their ability to protect the safety of their citizens by giving to the federal government sole authority to regulate the vehicles' design and performance. The House voted Wednesday, Sept. 6, 2017, to speed the introduction of self-driving cars by giving the federal government authority to exempt automakers from safety standards not applicable to the technology, and to permit deployment of up to 100,000 of the vehicles annually over the next several years. Members of the Senate Commerce committee are also working on self-driving car legislation, but a bill hasn't been introduced.

Derek Manky, global security strategist at Fortinet, said that the world is seeing more and more automation being built into black hat attackers' attack technology. What this means is, the time to respond to cyber-attack is shrinking drastically. Ten years ago, weeks or days to respond to a cyber-attack was adequate. Today, "we begin to measure in minutes (less than an hour)". "In the future, we will start measuring this in seconds. Humans cannot operate on this level, and therefore AI is crucial to respond at machine speed to the threat of cyber-attack," he said.