On a recent afternoon at the NVIDIA robotics research lab in Seattle's University District, researchers use a simulated kitchen to test robots' ability to perform simple tasks such as grabbing objects. A 5-feet 7-inch tall white robot, basically a spindly arm affixed with a claw of the sort customarily found in an arcade vending machine, glided around the kitchen on its two Segway wheels. Following the command of a research scientist sitting at a nearby computer, the robot grabbed a Cheez-It box on the counter and extended its limb to gently place the snacks inside a cabinet. "What's deceptive is that what's simple to us in the kitchen is challenging for a robot," said University of Washington Computer Science and Engineering Professor Dieter Fox, who also serves as the lab's senior director of robotics research. The Silicon Valley-based technology company opened the robotics lab last fall to harness the UW's talent in a sector where Seattle plays a central role. Still, paranoia around the capabilities of AI technology persist.

Visual 1st, the executive conference focused on promoting innovation and partnerships in the photo and video ecosystem, will bring AI, AR, computational photography, and the future of digital cameras to the center stage, Oct. 3-4, at the Golden Gate Club, San Francisco, Calif. AI is already everywhere in imaging, from recognition to enhancement to auto-editing – and of course, there's much more to come. In parallel, AR solutions are proliferating at a rapid pace, serving use cases ranging from having lots of fun to being highly productive. As these two technologies evolve in mutually reinforcing ways, we, as an industry, must take the imaging solutions they enable to the next level of value and profitability, while also keeping things safe, secure and private for our customers – but how? Alexander Schiffhauer recently left his role as Technical Advisor to Google's CEO Sundar Pichai to take product management responsibility for the company's computational photography teams. Under his leadership, these teams have pioneered innovation on Pixel Camera, leveraging AI and computer vision techniques to create photos unimaginable only a few years ago.

Turn AI cameras on your employees and you can measure their productivity. Fly them over the Pacific Ocean and you've got yourself an automated shark-warning system. What's happening: UC Santa Barbara, with the help of a few AI experts from Salesforce, is using drones to monitor sharks near California beaches in real time.

October 14, 2019 A BSTRACT Expressivity is one of the most significant issues in assessing neural networks. In this paper, we provide a quantitative analysis of the expressivity from dynamic models, where Hilbert space is employed to analyze its convergence and criticality. From the feature mapping of several widely used activation functions made by Hermite polynomials, We found sharp declines or even saddle points in the feature space, which stagnate the information transfer in deep neural networks, then present an activation function design based on the Hermite polynomials for better utilization of spatial representation. Moreover, we analyze the information transfer of deep neural networks, emphasizing the convergence problem caused by the mismatch between input and topological structure. We also study the effects of input perturbations and regularization operators on critical expressivity. Finally, we verified the proposed method by multivariate time series prediction. The results show that the optimized DeepESN provides higher predictive performance, especially for long-term prediction. Our theoretical analysis reveals that deep neural networks use spatial domains for information representation and evolve to the edge of chaos as depth increases. In actual training, whether a particular network can ultimately arrive that depends on its ability to overcome convergence and pass information to the required network depth. K eywords Deep neural networks; expressivity; criticality theory; convergence; activation function; Hilbert transform 1 Introduction Deep neural networks (DNNs) have achieved outstanding performance in many fields, from the automatic translation to speech and image recognition [1, 2].

We develop a novel approximate Bayesian computation (ABC) framework, ABCDP, that obeys the notion of differential privacy (DP). Under our framework, simply performing ABC inference with a mild modification yields differentially private posterior samples. We theoretically analyze the interplay between the ABC similarity threshold $\epsilon_{abc}$ (for comparing the similarity between real and simulated data) and the resulting privacy level $\epsilon_{dp}$ of the posterior samples, in two types of frequently-used ABC algorithms. We apply ABCDP to simulated data as well as privacy-sensitive real data. The results suggest that tuning the similarity threshold $\epsilon_{abc}$ helps us obtain better privacy and accuracy trade-off.

The Crew Interactive Mobile CompaniON (CIMON) mobile astronaut assistant, which is equipped with artificial intelligence (AI), returned to Earth on 27 August 2019. The SpaceX CRS-18 Dragon spacecraft carrying CIMON was undocked from the International Space Station (ISS) at 16:59 CEST; the capsule splashed down in the Pacific Ocean approximately 480 kilometres southwest of Los Angeles and was recovered at 22:21 CEST. "We expect CIMON to return to Germany at the end of October," reports Christian Karrasch, CIMON Project Manager at the German Aerospace Center (Deutsches Zentrum für Luft- und Raumfahrt; DLR) Space Administration. He looks back on the past few months: "CIMON is a technology demonstration that has completely met our expectations. During its initial operation in space – a 90-minute mission with the German ESA astronaut Alexander Gerst on the ISS in November 2018 – it showed that it functions well in microgravity conditions and can interact successfully with astronauts. We are very proud to have been the first to use AI on the Space Station and have been working for several months on an improved successor model. With CIMON, we were able to lay the foundations for human assistance systems in space to support astronauts in their tasks and perhaps, in the future, to take over some of their work."

For decades, sci-fi movies have predicted a future in which humans lose control of intelligent machines and chaos ensues. Those apocalyptic portrayals of artificial intelligence may seem like a distant or unrealistic future. But the seeds of a reality in which we lose control of the machines we build are being sown today. "What is the biggest software-related issue to impact the economy in Puget Sound in 2019?" "Software in the cockpit of an airplane, software that the pilots couldn't turn off," Smith said. Smith was referring to the multi-billion dollar fallout from Boeing's faulty 737 Max software that resulted in two crashes killing 346 people. Boeing's manufacturing center is based in Renton, Wash.

Matrix completion is a part of compressed sensing, and refers to determining an unknown low-rank matrix from a relatively small number of samples of the elements of the matrix. The problem has applications in recommendation engines, sensor localization, quantum tomography etc. In a companion paper (Part-1), the first and second author showed that it is possible to guarantee exact completion of an unknown low rank matrix, if the sample set corresponds to the edge set of a Ramanujan bigraph. In this paper, we present for the first time an infinite family of unbalanced Ramanujan bigraphs with explicitly constructed biadjacency matrices. In addition, we also show how to construct the adjacency matrices for the currently available families of Ramanujan graphs. In an attempt to determine how close the sufficient condition presented in Part-1 is to being necessary, we carried out numerical simulations of nuclear norm minimization on randomly generated low-rank matrices. The results revealed several noteworthy points, the most interesting of which is the existence of a phase transition. For square matrices, the maximum rank $\bar{r}$ for which nuclear norm minimization correctly completes all low-rank matrices is approximately $\bar{r} \approx d/3$, where $d$ is the degree of the Ramanujan graph. This upper limit appears to be independent of the specific family of Ramanujan graphs. The percentage of low-rank matrices that are recovered changes from 100% to 0% if the rank is increased by just two beyond $\bar{r}$. Again, this phenomenon appears to be independent of the specific family of Ramanujan graphs.

Waymo might just expand its self-driving car service to southern California... eventually. The Alphabet company's cars will start mapping some Los Angeles streets this week to explore the possibility of fitting autonomous vehicles into the city's "dynamic transportation environment." The firm told Engadget that its initial effort will be limited to three cars in the downtown area and the Miracle Mile, but that still raise the possibility of seeing a modified Pacifica cruising down the boulevard. The company stressed that this isn't a definitive sign that Waymo One or similar driverless services are coming to LA. Rather, this is to gauge the viability of introducing services "one day." While LA has been open to self-driving initiatives, it's not really a hub for them -- you have to travel to the San Francisco Bay Area if you want that in California.

In recent years, our fascination with the potential of AI has taken a more starry-eyed turn, as shown in the 2013 sci-fi drama "Her," where the main character falls in love with a virtual assistant. In reality, artificial intelligence (AI) technology is quickly permeating every aspect of our lives. From Amazon's voice-activated Alexa to writing technology that helps managers craft job postings, AI is in our hearts, homes and workplaces. And it's only going to become a bigger part of our lives: Experts call the rise of AI the driving force behind the fourth industrial revolution. On a recent afternoon at the NVIDIA robotics research lab in Seattle's University District, researchers use a simulated kitchen to test robots' ability to perform simple tasks such as grabbing objects.