Last week, the 2020 Tokyo Olympic Games organizing committee announced the launch of the "Tokyo 2020 Robot Project." The project will involve the deployment of an assortment of robots to do useful things for visitors at the games, but so far, we've just seen specific details about two: Toyota's Human Support Robot (HSR) and Delivery Support Robot (DSR). These robots are supposed to be part of a "practical real-life deployment helping people," and the idea is that HSR and DSR will work together to assist disabled visitors, showing them to their seats and fetching food or other items that can be ordered with a tablet. The Toyota HSR is a mobile manipulator, able to move around and pick stuff up. It can do all kinds of things, provided that you can program it to do all of those things, which is not easy, especially if it's supposed to operate autonomously in an Olympic venue rather than a robotics lab.

Nuro has partnered with Fry's Food Stores to utilize its autonomous vehicles to deliver groceries in Scottsdale. Supermarket giant Kroger said it soon will end a pilot program in which more than 2,000 grocery deliveries were made in self-driving vehicles from a store in Scottsdale, Arizona. The program, launched last August, featured deliveries in autonomous vehicles from robotics company Nuro from the Kroger-owned Fry's store at 7770 E. McDowell Road for customers in ZIP code 85257. The companies described it as the nation's first program featuring deliveries to the general public from fully unmanned vehicles. Wednesday will mark the final day of deliveries.

If you thought IBM using "quietly scraped" Flickr images to train facial recognition systems was bad, it gets worse. Our research, which will be reviewed for publication this summer, indicates that the U.S. government, researchers, and corporations have used images of immigrants, abused children, and dead people to test their facial recognition systems, all without consent. The very group the U.S. government has tasked with regulating the facial recognition industry is perhaps the worst offender when it comes to using images sourced without the knowledge of the people in the photographs. The National Institute of Standards and Technology, a part of the U.S. Department of Commerce, maintains the Facial Recognition Verification Testing program, the gold standard test for facial recognition technology. This program helps software companies, researchers, and designers evaluate the accuracy of their facial recognition programs by running their software through a series of challenges against large groups of images (data sets) that contain faces from various angles and in various lighting conditions.

According to a recent TNG survey, 73 percent of job seekers in Sweden believe they've been discriminated against during the job application process. By replacing the human recruiter with Tengai, TNG and Furhat believe they can make the screening process more fair while still providing a "human" touch. "I was quite sceptical at first before meeting Tengai, but after the meeting I was absolutely struck," healthcare recruiter Petra Elisson, who has been involved in the testing, told the BBC. "At first I really, really felt it was a robot, but when going more deeply into the interview I totally forgot that she's not human." As for ensuring that Tengai doesn't reflect the biases of its creators and training data -- a problem that has cropped up with other AIs -- Furhat's chief scientist, Gabriel Skantze, told the BBC the company is making it a point to conduct test interviews with a diverse mix of recruiters and volunteers before Tengai is ever in the position to actually decide an applicant's employment fate.

AstraZeneca is mounting a big push into digital technologies, which includes hiring a former Nasa artificial intelligence expert, as it seeks to accelerate drug discovery and show the value of its medicines in an increasingly tough pricing environment. The approach was mapped out at a meeting of about 200 senior leaders at its headquarters in Gothenburg, Sweden, late last month. It forms part of a blueprint for future growth that its chief executive, Pascal Soriot, has internally dubbed "AZ2025". The pharma industry is facing unprecedented levels of scrutiny over its pricing, particularly in the US. At the same time, the emergence of highly expensive potential cures for diseases such as cancer has put more pressure on drug companies to show their prices represent value for money.

Be it online, offline or omnichannel, today, retail is all about giving the best customer experience and technology is helping the industry to achieve its newly discovered goal. From frontend to backend, the new age technologies like artificial intelligence (AI) are not just making machines smarter but also business by helping it make optimum use of the allocated resources. After a lot of hits and misses, retailers today understand that they have to become a digitally savvy business to remain relevant in the future. If they fail to do so, their stubbornness will make them irrelevant. This awareness and how digitalization will revolutionize the industry is triggering the shift among retailers to adopt new technology. The company has scaled up from a hyperlocal reward program platform to a technology company helping retailers to maximize their marketing campaigns using AI.

It's not a perfect measure, but unit sales of industrial robots give some idea of a country's industrial might. The names of the top five buyers in 2017 – China, Japan, South Korea, the US and Germany – shouldn't be too surprising. The global average is 74 per 10,000. One factor in this is the small electronics and automotive sectors here, which are two major drivers of industrial robot investment. The high number of SME and micro-businesses in Australian manufacturing is another.

Recently, it has been shown that many functions on sets can be represented by sum decompositions. These decompositons easily lend themselves to neural approximations, extending the applicability of neural nets to set-valued inputs---Deep Set learning. This work investigates a core component of Deep Set architecture: aggregation functions. We suggest and examine alternatives to commonly used aggregation functions, including learnable recurrent aggregation functions. Empirically, we show that the Deep Set networks are highly sensitive to the choice of aggregation functions: beyond improved performance, we find that learnable aggregations lower hyper-parameter sensitivity and generalize better to out-of-distribution input size.

This work gives an in-depth derivation of the trainable evidence lower bound (ELBO) obtained from the marginal joint log-Likelihood with the goal of training a multi-modal variational Autoencoder (M²VAE). I. INTRODUCTION Variational auto encoder (VAE) combine neural networks with variational inference to allow unsupervised learning of complicated distributions according to the graphical model shown in Figure 1 (left). The specific objective of VAEs is the maximization of the marginal distribution p(a) p (a z)p(z) da. II This approach proposed by [1] is used in settings where only a single modality a is present in order to find a latent encoding z (c.f. Figure 1 (left)). This work gives an in-depth derivation of the trainable evidence lower bound (ELBO) obtained from the marginal joint log-Likelihood, that satisfies all plate models as depicted in Figure 1, we are with the goal of training a multi-modal variational Autoencoder (M²VAE).

The hyperbolic manifold is a smooth manifold of negative constant curvature. While the hyperbolic manifold is well-studied in the literature, it has gained interest in the machine learning and natural language processing communities lately due to its usefulness in modeling continuous hierarchies. Tasks with hierarchical structures are ubiquitous in those fields and there is a general interest to learning hyperbolic representations or embeddings of such tasks. Additionally, these embeddings of related tasks may also share a low-rank subspace. In this work, we propose to learn hyperbolic embeddings such that they also lie in a low-dimensional subspace. In particular, we consider the problem of learning a low-rank factorization of hyperbolic embeddings. We cast these problems as manifold optimization problems and propose computationally efficient algorithms. Empirical results illustrate the efficacy of the proposed approach.