It's very rare to see any travel video or brochure these days that doesn't have an image shot from overhead, on a drone. Life seems more dramatic up from above, right? The units themselves have gotten way easier to use, more affordable and the camera quality is pretty amazing. Imagine being able to throw a high-quality camera in the air that can get stunning overhead shots, smooth video even in wind and always somehow return home to sender. It's one of the major tech advancements of our time, but it can be a little confusing.

The developers of a speech recognition bot assigned to analyze the public statements of politicians hit a major stumbling block when it tried to make sense of Donald Trump. Built by a tech startup called FactSquared, the bot AI was assigned to go through more than 11 million words Trump has spoken or tweeted since 1976--in interviews, campaign speeches, media appearances, and social media posts. According to FactSquared's CEO Bill Frischling, the bot failed to understand Trump's speeches until he brought in a specialist to strip out all of the bot's grammar and syntax coding. The tech startup FactSquared created an AI bot to try and catalog and analyze Donald Trump's public appearances and interviews, but they were so incoherent and rambling the bot actually crashed. 'It was still trying to punctuate it like it was English, versus trying to punctuate it like it was Trump,' Frischling told The LA Times.

To me, the San Francisco streets seemed deserted. To my self-driving car, they were full of hazards. In mid-March, just as the coronavirus outbreak started to change the world as we knew it, I took a ride in an autonomous vehicle through the narrow and winding, topsy-turvy streets of downtown San Francisco -- from the hairpin turns of Lombard Street to the steep hills surrounding Coit Tower and the famed Embarcadero waterfront. Even with tens of thousands of workers staying put as the first work-from-home orders hit, in the back of a Toyota Highlander piloted by autonomous vehicle start-up Zoox, I started to become hyper-aware of the circus of hazards robocars encounter on a daily basis. There was a cyclist or skateboarder in the blind spot.

Current textual question answering models achieve strong performance on in-domain test sets, but often do so by fitting surface-level patterns in the data, so they fail to generalize to out-of-distribution settings. To make a more robust and understandable QA system, we model question answering as an alignment problem. We decompose both the question and context into smaller units based on off-the-shelf semantic representations (here, semantic roles), and align the question to a subgraph of the context in order to find the answer. We formulate our model as a structured SVM, with alignment scores computed via BERT, and we can train end-to-end despite using beam search for approximate inference. Our explicit use of alignments allows us to explore a set of constraints with which we can prohibit certain types of bad model behavior arising in cross-domain settings. Furthermore, by investigating differences in scores across different potential answers, we can seek to understand what particular aspects of the input lead the model to choose the answer without relying on post-hoc explanation techniques. We train our model on SQuAD v1.1 and test it on several adversarial and out-of-domain datasets. The results show that our model is more robust cross-domain than the standard BERT QA model, and constraints derived from alignment scores allow us to effectively trade off coverage and accuracy.

Reading comprehension models often overfit to nuances of training datasets and fail at adversarial evaluation. Training with adversarially augmented dataset improves robustness against those adversarial attacks but hurts generalization of the models. In this work, we present several effective adversaries and automated data augmentation policy search methods with the goal of making reading comprehension models more robust to adversarial evaluation, but also improving generalization to the source domain as well as new domains and languages. We first propose three new methods for generating QA adversaries, that introduce multiple points of confusion within the context, show dependence on insertion location of the distractor, and reveal the compounding effect of mixing adversarial strategies with syntactic and semantic paraphrasing methods. Next, we find that augmenting the training datasets with uniformly sampled adversaries improves robustness to the adversarial attacks but leads to decline in performance on the original unaugmented dataset. We address this issue via RL and more efficient Bayesian policy search methods for automatically learning the best augmentation policy combinations of the transformation probability for each adversary in a large search space. Using these learned policies, we show that adversarial training can lead to significant improvements in in-domain, out-of-domain, and cross-lingual (German, Russian, Turkish) generalization without any use of training data from the target domain or language.

Traffic prediction plays an essential role in intelligent transportation system. Accurate traffic prediction can assist route planing, guide vehicle dispatching, and mitigate traffic congestion. This problem is challenging due to the complicated and dynamic spatio-temporal dependencies between different regions in the road network. Recently, a significant amount of research efforts have been devoted to this area, greatly advancing traffic prediction abilities. The purpose of this paper is to provide a comprehensive survey for traffic prediction. Specifically, we first summarize the existing traffic prediction methods, and give a taxonomy of them. Second, we list the common applications of traffic prediction and the state-of-the-art in these applications. Third, we collect and organize widely used public datasets in the existing literature. Furthermore, we give an evaluation by conducting extensive experiments to compare the performance of methods related to traffic demand and speed prediction respectively on two datasets. Finally, we discuss potential future directions.

Apple has released data from its Maps app to track whether people are complying with coronavirus lockdowns. The data shows whether its users are still requesting directions from their iPhones, and so can be used to see how much people are travelling compared to before the lockdowns came into effect. It shows a dramatic reduction in the number of people using driving, walking and transport directions to get around, suggesting that stay-in-place orders and other rules to stop the spread of coronavirus are working. In the San Francisco Bay area, for instance, requests for driving directions are down 70 per cent, and people looking for transit directions have dropped 84 per cent. In New York, transit requests were down 89 per cent.

Microsoft Real Estate and Security (RE&S) is responsible for heating and cooling 115 buildings in the Puget Sound area. Microsoft Core Services and Engineering (CSEO) partnered with RE&S to improve the effectiveness of the schedules for their heating, ventilation, and air conditioning (HVAC) system to reduce costs and increase employee comfort. CSEO implemented machine learning to predict when employees will arrive into Microsoft buildings each morning and how long it will take for a building to reach its optimal comfort temperature. As a result, we were able to generate a dynamic HVAC schedule that resulted in significant cost savings and increased employee comfort for RE&S. We're continuing to implement machine learning in our buildings throughout the Puget Sound region and we're encouraging the rest of Microsoft to use machine learning to optimize operations and drive digital transformation.

Self-driving start-up backed by Toyota has transformed its robo-taxi fleet into a driverless grocery delivery service last week. The move came in response to the ongoing lockdown in California and the heightened need for essential goods amid the Coronavirus pandemic. The start-up, Pony.ai, is partnering with e-commerce site Yamibuy on the initiative and its fleet consists of 10 self-driving electric Hyundai Konas. The outfit had been participating in a robo-taxi pilot in the city of Irvine, California, since November 2019 but was forced to ground its fleet in March owing to the state's lockdown orders. Now the vehicles will run delivery trips from Yamibuy's distribution center to residences, condos and apartment complexes in Irvine.

In keeping with our mission to provide comprehensive advertising analysis, MediaRadar puts together a report of the most important mergers and acquisitions news each week. Stay in the loop, whether you sell advertising space or focus on business development. This week, Gilead takes out FortySeven, Apple acquires start-up Voysis and Infor is purchased by Koch Industries. The American biotechnology company, Gilead has completed an acquisition of Forty Seven, Inc. at a rate of $97.50 per share that equates to a lump sum of $4.9 billion in cash. The deal bolsters Gilead's portfolio of oncology drugs through Forty Seven Inc.'s blood cancer medicine, which is expected to be on the market within 2 years.