Here's the deal, says Mark Rosekind. Around him, some 400 workers clack away on computers, or roll out yoga mats in the central "town hall" space, or tend to the startup's fleet of self-driving, golf-cart-on-steroids prototypes. The deal is that in spite of all this kind of work--work that has put autonomous vehicles on the streets of cities around the world--regulators don't know how to ensure the potentially life-saving technology won't instead make roads more dangerous. "A company might think it's OK because it checks some box," says Rosekind, whose job is to help Zoox solve this puzzle. Maybe its robo-car has amassed 50 million of miles of data, or has executed a perfect three-point turn, or reliably pulls over when a wailing police car appears behind it.

Deep Neural Networks are quite vulnerable to adversarial perturbations. Current state-of-the-art adversarial attack methods typically require very time consuming hyper-parameter tuning, or require many iterations to solve an optimization based adversarial attack. To address this problem, we present a new family of trust region based adversarial attacks, with the goal of computing adversarial perturbations efficiently. We propose several attacks based on variants of the trust region optimization method. We test the proposed methods on Cifar-10 and ImageNet datasets using several different models including AlexNet, ResNet-50, VGG-16, and DenseNet-121 models. Our methods achieve comparable results with the Carlini-Wagner (CW) attack, but with significant speed up of up to $37\times$, for the VGG-16 model on a Titan Xp GPU. For the case of ResNet-50 on ImageNet, we can bring down its classification accuracy to less than 0.1\% with at most $1.5\%$ relative $L_\infty$ (or $L_2$) perturbation requiring only $1.02$ seconds as compared to $27.04$ seconds for the CW attack. We have open sourced our method which can be accessed at [1].

Despite the tremendous success of deep neural networks in various learning problems, it has been observed that adding an intentionally designed adversarial perturbation to inputs of these architectures leads to erroneous classification with high confidence in the prediction. In this work, we propose a general framework based on the perturbation analysis of learning algorithms which consists of convex programming and is able to recover many current adversarial attacks as special cases. The framework can be used to propose novel attacks against learning algorithms for classification and regression tasks under various new constraints with closed form solutions in many instances. In particular we derive new attacks against classification algorithms which are shown to achieve comparable performances to notable existing attacks. The framework is then used to generate adversarial perturbations for regression tasks which include single pixel and single subset attacks. By applying this method to autoencoding and image colorization tasks, it is shown that adversarial perturbations can effectively perturb the output of regression tasks as well.

NASA has finally pinpointed the exact landing location of its new Mars explorer, thanks to a powerful camera on its Mars Reconnaissance Orbiter. While the space agency knew InSight had landed within an 81-mile-long (130 km) ellipse on the red planet, there was no way to determine exactly where it touched down within this region. Now, a series of images captured this week by MRO's HiRISE camera have confirmed that the lander, heat shield, and parachute all sit within 1,000 feet of each other on a lava plain called Elysium Planitia. NASA has finally pinpointed the exact landing location of its new Mars explorer, thanks to a powerful camera on its Mars Reconnaissance Orbiter. In the images released today by NASA, InSight and its parts appear as bright teal specks on rust-colored landscape.

The popular narrative around artificial intelligence research is that it's mainly a war between China and the United States. Not so fast, says Europe. New data released today (Dec. The data was assembled from Scopus, a citation database owned by scientific publishing company Elsevier. If the current trend continues, China will soon overtake Europe in the number of papers published.

Donald Trump is speaking Mandarin. This is happening in the city of Tianjin, about an hour's drive south of Beijing, within a gleaming office building that belongs to iFlytek, one of China's rapidly rising artificial-intelligence companies. Beyond guarded gates, inside a glitzy showroom, the US president is on a large TV screen heaping praise on the Chinese company. It's Trump's voice and face, but the recording is, of course, fake--a cheeky demonstration of the cutting-edge AI technology iFlytek is developing. Jiang Tao chuckles and leads the way to some other examples of iFlytek's technology.

BEIJING – China's policy of "reform and opening up" has brought monumental changes to the world's most populous country since its launch 40 years ago under leader Deng Xiaoping. Next week, China will mark the anniversary of the shift, agreed to at a Communist Party gathering on Dec. 18, 1978. Ou Banlan, 52, is a retired garment factory worker in Shenzhen, a former fishing village that was the testing ground for the reforms and morphed into a major manufacturing and high-tech hub. "My life is much better than that of my parents' generation," said the diminutive woman with short black hair, standing in front of the factory where she once toiled. She was born and raised in a village outside Shenzhen.

Yes--but it's far from enough to satisfy China hawks like U.S. Trade Representative Robert Lighthizer. Markets clearly recognize this: The S&P 500 ended up only 0.5% Wednesday after the news broke. Moreover, the China 2025 plan itself--despite all the attention it has received--may be less menacing than it seems. What's really needed to take negotiations to the next level, and assuage market concerns, is for China to enact a few big-bang reforms to convince foreigners that Xi Jinping's administration is committed to level dealing. Doing away entirely with most joint-venture requirements--instead of endless foot-dragging and qualifications--is one possibility.

For years tech companies such as Amazon, Alphabet and Uber have promised us delivery drones bringing goods to our doorsteps in a matter of minutes. So why are they taking so long to arrive? If our skies are to become as crowded as our streets, airspace rules need updating to prevent accidents, terrorist attacks, and related problems, such as noise pollution. But that's easier said than done. According to a recent study by Nasa, the noise made by road traffic was "systematically judged to be less annoying" than the high-pitched buzzing made by drones.

Apple said it would add more than 1,000 employees apiece in San Diego, Seattle and Culver City, Calif., areas where it has been increasing staff to support its development of custom chips, machine-learning systems and Hollywood programming. It also plans hundreds of additional jobs in cities where it already has offices, including New York, Boston and Portland, Ore. The Austin campus would have the capacity to eventually accommodate 15,000 employees, Apple said, and was expected to make the company the city's largest private employer. The announcement came weeks after Amazon.com Inc. and Alphabet Inc. GOOGL -0.02% said they would expand in regions where they already have a presence.