Ian Goodfellow is an author of the popular textbook on deep learning (simply titled "Deep Learning"). He invented Generative Adversarial Networks (GANs) and with his 2014 paper is responsible for launching the incredible growth of research on GANs. He got his BS and MS at Stanford, his PhD at University of Montreal with Yoshua Bengio and Aaron Courville. He held several research positions including at OpenAI, Google Brain, and now at Apple as director of machine learning. This recording happened while Ian was still at Google Brain.

More than 30% said their companies have allocated $50 million or more to smart automation projects, and more than half have already spent at least $10 million; the initiatives include various combinations of robotic process automation (RPA), artificial intelligence, machine learning, cognitive computing and analytics; highest expenditure levels were for the finance and accounting category, marked by 23% of respondents as receiving investment of slightly more than US$50 million; the technology that organizations are experimenting with or piloting the most is AI (36); 30% of companies are opting not to invest or are unsure of their plans for smart automation (KPMG Easing the Pressure Points). Lawyers surveyed think AI will be valuable for tracking billable time (53% of US layers, 49% of UK lawyers), conflicts clearance (43% and 41%), and compliance with client billing documents (34% for both US and UK lawyers) (Intapp survey reveals lawyers' attitudes toward technology). The portion of auto companies not using or testing AI rose to 39% in 2019 from 26% in 2017 (Capgemini). "The accelerated growth of RPA is being driven by high levels of efficiency and productivity that can now be achieved from intelligent automation, which combines advanced RPA, artificial intelligence and embedded analytics. The demand for RPA solutions has surged as legacy companies are now competing with'digital native' companies like Amazon and Uber, in which nearly every part of the business is completely automated"--Mihir Shukla, CEO of Automation Anywhere Inc., an RPA maker that expects to deploy three million software robots at organizations worldwide by 2020, a 200% increase from today (Wall Street Journal).

Toyota announced creation of a $100 million venture fund to invest in autonomous driving and robotic technology start-ups as automakers increasingly push into the self-driving market. Toyota AI Ventures, a Silicon Valley-based subsidiary of Toyota, plans to invest the said amount into early-stage startups that are developing "disruptive" technologies in those fields, the company said. Jim Adler, managing director of Toyota AI Ventures, said in a statement, "Auto manufacturers must participate in the startup ecosystem to stay ahead of the rapid shift in the auto industry." The company added that the fund is part of Toyota's mission is a futuristic "discover what's next" phase. Toyota's AI venture fund has already invested in 19 different start-ups over the last two years, bringing its total funding commitment to autonomous driving technology to $200 million, the company further stated. Toyota AI Ventures looks for early-stage startups, across a range of industries, that are applying AI, data, and cloud technologies to tackle important problems and create new market opportunities.

In March, a gunman walked into two mosques in Christchurch, New Zealand, opened fire, and killed dozens of worshippers. According to a police official, the suspected gunman was arrested 36 minutes after police were called to the scene. Now, a tech company believes its smart security cameras can prevent attacks like the tragedy in Christchurch, and says it plans to install its AI-powered systems in mosques around the world. Athena Security, the tech company behind the security system, and Al-Ameri International Trading announced the Keep Mosques Safe initiative last week. Al-Ameri International Trading, along with several Islamic non-profit groups, will fund the Keep Mosques Safe effort.

Today's intelligent assistants are full of skills. They can check the weather, traffic and sports scores. They can play music, translate words and send text messages. They can even do math, tell jokes and read stories. But, when it comes to conversations that lead somewhere grander, the wheels fall off.

Bayesian Optimization using Gaussian Processes is a popular approach to deal with the optimization of expensive black-box functions. However, because of the a priori on the stationarity of the covariance matrix of classic Gaussian Processes, this method may not be adapted for non-stationary functions involved in the optimization problem. To overcome this issue, a new Bayesian Optimization approach is proposed. It is based on Deep Gaussian Processes as surrogate models instead of classic Gaussian Processes. This modeling technique increases the power of representation to capture the non-stationarity by simply considering a functional composition of stationary Gaussian Processes, providing a multiple layer structure. This paper proposes a new algorithm for Global Optimization by coupling Deep Gaussian Processes and Bayesian Optimization. The specificities of this optimization method are discussed and highlighted with academic test cases. The performance of the proposed algorithm is assessed on analytical test cases and an aerospace design optimization problem and compared to the state-of-the-art stationary and non-stationary Bayesian Optimization approaches.

ArCo is the Italian Cultural Heritage knowledge graph, consisting of a network of seven vocabularies and 169 million triples about 820 thousand cultural entities. It is distributed jointly with a SPARQL endpoint, a software for converting catalogue records to RDF, and a rich suite of documentation material (testing, evaluation, how-to, examples, etc.). ArCo is based on the official General Catalogue of the Italian Ministry of Cultural Heritage and Activities (MiBAC) - and its associated encoding regulations - which collects and validates the catalogue records of (ideally) all Italian Cultural Heritage properties (excluding libraries and archives), contributed by CH administrators from all over Italy.

While which applies statistical and mathematical methods to process PLSR focuses on calculating the linear projections that shows the data obtained through spectroscopic techniques, in maximum correlation with the output or target variable, thus order to derive information of interest. The need for chemometric estimating a linear regression model determined by the projected analysis comes from the development of analytical coordinates. Benoudjit et al. [10] proposed linear and instruments and techniques that are capable of producing nonlinear regression methodologies which are based upon an large amount of complex data. Data collection through spectroscopic incremental routine for feature selection and using a validation technique is based on interaction of light energy of set. In [11,12] different techniques have been introduced variable wavelength with samples under test [1]. The ability to improve the results of previous method by choosing the of a sample to absorb or transmit light energy is recorded in best feature set for initializing the routine and finding a feature terms of values throughout a selected bandwidth of electromagnetic selection strategy that depends entirely on the shared spectrum. Whether it be food, pharmaceutical or information between spectral data and target variable. An textile industry, concentrations of chemical components of interesting approach to the chemometrics problems has been interest in samples are estimated through chemometric analysis.

We present a novel high frequency residual learning framework, which leads to a highly efficient multi-scale network (MSNet) architecture for mobile and embedded vision problems. The architecture utilizes two networks: a low resolution network to efficiently approximate low frequency components and a high resolution network to learn high frequency residuals by reusing the upsampled low resolution features. With a classifier calibration module, MSNet can dynamically allocate computation resources during inference to achieve a better speed and accuracy trade-off. We evaluate our methods on the challenging ImageNet-1k dataset and observe consistent improvements over different base networks. On ResNet-18 and MobileNet with alpha=1.0, MSNet gains 1.5% accuracy over both architectures without increasing computations. On the more efficient MobileNet with alpha=0.25, our method gains 3.8% accuracy with the same amount of computations.

The recent success of brain-inspired deep neural networks (DNNs) in solving complex, high-level visual tasks has led to rising expectations for their potential to match the human visual system. However, DNNs exhibit idiosyncrasies that suggest their visual representation and processing might be substantially different from human vision. One limitation of DNNs is that they are vulnerable to adversarial examples, input images on which subtle, carefully designed noises are added to fool a machine classifier. The robustness of the human visual system against adversarial examples is potentially of great importance as it could uncover a key mechanistic feature that machine vision is yet to incorporate. In this study, we compare the visual representations of white- and black-box adversarial examples in DNNs and humans by leveraging functional magnetic resonance imaging (fMRI). We find a small but significant difference in representation patterns for different (i.e. white- versus black- box) types of adversarial examples for both humans and DNNs. However, human performance on categorical judgment is not degraded by noise regardless of the type unlike DNN. These results suggest that adversarial examples may be differentially represented in the human visual system, but unable to affect the perceptual experience.