It is a stunning scene – the sun setting over what Steve Jobs had called'the best office building in the world'. The latest flyover drone video of Apple Park has made its way to the web, which shows just how massive the spaceship has become just days before its first residents are expected to move in. The clip has captured the curved glass panels that are being installed around the 2.8 million-square-foot main building and the large courtyard area that will boast thousands of drought-resistant trees. It is a stunning scene – the sun setting over what Steve Jobs had called'the best office building in the world'. Duncan Sinfield, a YouTuber and independent drone pilot, shared the 4 minute and 41 second video to his channel this week, 9to5Mac reported.

When Bakul Patel started as a policy advisor in the US Food and Drug Administration in 2008, he could pretty much pinpoint when a product was going to land in front of the reviewers in his division. Back when medical devices were heavy on the hardware--your pacemakers and your IUDs--it would take manufacturers years to get them ready for regulatory approval. FDA reviewers could keep up pretty well. But as computer code took on more complex tasks, like spotting specious moles and quantifying blood flow, their duties began to accelerate. Software developers needed months, not years, to make it to the market.

You might not have to register your drones through the Federal Aviation Administration anymore, but if your flyer of choice is a DJI model, you'll have to activate it online, or take a hit in flight range and functionality. The company is adding a new application activation process, tied to a firmware update for its family of drones. It'll come into effect at the end of this week, and is aimed at ensuring pilots use "the correct set of geospatial information and flight functions for your aircraft", determined by your location and user profile. The activation process will require users to connect to the internet through DJI's app, to verify your account and activate the update. Your drone won't be able to access the geospatial info and flight functions, camera streaming will be disabled and flight range will be curtailed to a 164-foot radius, up to 98 feet high. These rules will apply to all of DJI's "aircraft" running the latest firmware, whether you own a Phantom 4 or one of its cheaper drones.

A multiplexer-demultiplexer (MDM) data relay box aboard the International Space Station failed Saturday, and will be replaced Tuesday during a spacewalk by NASA astronauts Peggy Whitson and Jack Fischer, the space agency announced Sunday. The two-hour "contingency spacewalk" has been given the go-ahead by program managers of the ISS. The MDM data relay box is housed in the S0 truss, and is one of the two such systems, both of which are redundant -- which is to say, both boxes are independent and one can take over in the event the other fails. MDM-1, the failed data relay box, apparently malfunctioned shortly after noon Saturday. Crewmembers tried to restore power to the component multiple times, but were not successful.

Dropout is used as a practical tool to obtain uncertainty estimates in large vision models and reinforcement learning (RL) tasks. But to obtain well-calibrated uncertainty estimates, a grid-search over the dropout probabilities is necessary-- a prohibitive operation with large models, and an impossible one with RL. We propose a new dropout variant which gives improved performance and better calibrated uncertainties. Relying on recent developments in Bayesian deep learning, we use a continuous relaxation of dropout's discrete masks. Together with a principled optimisation objective, this allows for automatic tuning of the dropout probability in large models, and as a result faster experimentation cycles. In RL this allows the agent to adapt its uncertainty dynamically as more data is observed. We analyse the proposed variant extensively on a range of tasks, and give insights into common practice in the field where larger dropout probabilities are often used in deeper model layers.

Adversarial training has been shown to regularize deep neural networks in addition to increasing their robustness to adversarial examples. However, its impact on very deep state of the art networks has not been fully investigated. In this paper, we present an efficient approach to perform adversarial training by perturbing intermediate layer activations and study the use of such perturbations as a regularizer during training. We use these perturbations to train very deep models such as ResNets and show improvement in performance both on adversarial and original test data. Our experiments highlight the benefits of perturbing intermediate layer activations compared to perturbing only the inputs. The results on CIFAR-10 and CIFAR-100 datasets show the merits of the proposed adversarial training approach. Additional results on WideResNets show that our approach provides significant improvement in classification accuracy for a given base model, outperforming dropout and other base models of larger size.

Artificial intelligence has become as meaningless a description of technology as "all natural" is when it refers to fresh eggs. At least, that's the conclusion reached by Devin Coldewey, a Tech Crunch contributor. AI is also often mentioned as a potential cybersecurity technology. At the recent RSA conference in San Francisco, RSA CTO Zulfikar Ramzan advised potential users to consider AI-based solutions carefully, in particular machine learning-based solutions, according to an article on CIO. AI-based tools are not as new or productive as some vendors claim, he cautioned, explaining that machine learning-based cybersecurity has been available for over a decade via spam filters, antivirus software and online fraud detection systems.

Janzamin, H. Sedghi and A. Anandkumar, Beating the Perils of Non-Convexity: Guaranteed Training of Neural Networks using Tensor Methods, 2015 The AI Winter refers to the period in the mid-1980s in which interest in AI began to drop as the field failed to yield significant practical gains. Skynet is the name of the fictional, murderous artificial intelligence in the Terminator movie series. See also Harry Surden, Technological Opacity, Predictability, and Self-Driving Cars, 38 Cardozo L. Rev. 121, 162–63 (2016). See also In the Matter of the American Medical Association, et al., 94 F.T.C. 701 (1979) (doctors); In the Matter of Connecticut Chiropractic Ass'n, 114 F.T.C. 708 (1991) (chiropractors); In the Matter of Nat'l Soc'y of Prof'l Eng'rs, 116 F.T.C. 787 (1993) (engineers).

The days of claiming artificial intelligence as a feature that set one startup or company apart from the others are over. These days, one would be hard-pressed to find any technology company attracting venture funding or partnerships that doesn't posit to use some form of machine learning. But for companies trying to innovate in healthcare using artificial intelligence, the stakes are considerably higher, meaning the hype surrounding the buzzword can be deflated far more quickly than in some other industry, where a mistaken algorithm doesn't mean the difference between life and death. Over the past five years, the number of digital health companies employing some form of artificial intelligence has dramatically increased. CB Insights tracked 100 AI-focused healthcare companies just this year, and noted 50 had raised their first equity rounds since January 2015.

A Japanese soccer fan after Japan loses its 2014 World Cup soccer match against Colombia. It's midnight in Tokyo and Takehiro Onuki has just left the office, 16 hours after his shift began. Onuki, a 31-year-old salesman, is headed to the train station to catch the 12:24 a.m. The train will quickly fill up with other professional working men. At about 1:30 a.m., after having made a pit stop at a convenience store to grab a sandwich, Onuki arrives home. When he opens the bedroom door, he accidentally wakes his wife, Yoshiko, who just recently fell asleep after working an 11-hour day.