DeepMind just released a Generative model able to outperform widely-used nowcasting methods in 89% of situations for its accuracy and usefulness assessed by more than 50 expert meteorologists! Their model focuses on predicting precipitations in the next 2 hours and achieves that surprisingly well. It is a generative model, which means that it will generate the forecasts instead of simply predicting them. It basically takes radar data from the past to create future radar data. So using both time and spatial components from the past, they can generate what it will look like in the near future.

Building usable models to run AI algorithms requires not just adequate data to train systems, but also the right hardware subsequently to run them. But because the theoretical and practical are often not the same thing, there is often a gap between what data scientists may hope to do and what they practically do. Today, a startup called Deci that has built a deep learning platform to help bridge that gap -- by building models that can work with the data and hardware that are available to use -- is announcing some funding after finding strong traction for its products with Fortune 500 tech companies running mass-market, AI-based products based on video and other computer vision-based services. The Tel Aviv-based startup has picked up a Series A of $21 million, money that it will be using to continue expanding its product and customer base. Insight Partners is leading the round, with previous backers Square Peg, Emerge and Jibe Ventures, alongside some new backers: Samsung Next, Vintage Investment Partners, and Fort Ross Ventures.

A deep model was trained on historical crash data, road maps, satellite imagery, and GPS to enable high-resolution crash maps that could lead to safer roads. Today's world is one big maze, connected by layers of concrete and asphalt that afford us the luxury of navigation by vehicle. For many of our road-related advancements -- GPS lets us fire fewer neurons thanks to map apps, cameras alert us to potentially costly scrapes and scratches, and electric autonomous cars have lower fuel costs -- our safety measures haven't quite caught up. We still rely on a steady diet of traffic signals, trust, and the steel surrounding us to safely get from point A to point B. To get ahead of the uncertainty inherent to crashes, scientists from MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL) and the Qatar Center for Artificial Intelligence developed a deep learning model that predicts very high-resolution crash risk maps. Fed on a combination of historical crash data, road maps, satellite imagery, and GPS traces, the risk maps describe the expected number of crashes over a period of time in the future, to identify high-risk areas and predict future crashes.

The manufacturing process of both large-scale sectors such as the automobile industry to small-scale sectors such as device or router manufacturing is vulnerable to defects. These defects could be very minute or microscopic flaws but are capable of determining the final product quality. If not properly identified this can result in a mass recall of products that could significantly alter the brand reputation and cause millions of dollars in losses. Not only for the product manufacturers, but the defective product can also affect the buyers as well, resulting in a catastrophic accident. For instance, If a nuclear industry purchases these undetected defective items, it can lead to fatal and devastating outcomes for employees as well as the environment.

CEO at Reface, an AI/ML startup shifting from the face-swapping app to the platform for creating personalized content. Last year, 64.2 zettabytes of data were created globally -- enough to fill about 1 trillion 64GB flash drives -- according to IDC. It might be hard to believe, but the total amount of digital data created over the next five years will double the amount of information developed since the birth of digital storage. The percentage of information generated synthetically may be negligible for now, but by 2030 (registration required), synthetic data is expected to completely overshadow real data in AI models. What role will synthetic media play, and what trends are exploding on the market of deep learning products?

Artificial intelligence is at the top of many lists of the most important skills in today's job market. In the last decade or so we have seen a dramatic transition from the "AI winter" (where AI has not lived up to its hype) to an "AI spring" (where machines can now outperform humans in a wide range of tasks). Having spent the last 25 years as an AI researcher and practitioner, I'm often asked about the implications of this technology on the workforce. I'm quite often disheartened by the amount of disinformation there is on the internet on this topic, so I've decided to share some of my own thoughts. The difference between what I am about to write, and what you may have read before elsewhere is due to an inherent bias. Rather than being a pure "AI" practitioner, my PhD and background is in Cognitive Science - the scientific study of how the mind works, spanning such areas as psychology, neuroscience, philosophy, and artificial intelligence. My area of research has been to look explicitly at how the human mind works, and to reverse engineer these processes in the development of artificial intelligence platforms.

The recent hype surrounding Generative Pre-trained Transformer 3 (GPT-3), the new artificial intelligence (AI) based natural language processing (NLP) model, is worth observing, particularly from the enterprise front. Both keen observation and casual look-see applied to this latest language model that generates human-like written content are worth your time and effort. It can also show you that the hype is real. However, like every technological innovation, GPT-3 has its shortcomings, yet it is a great leap for AI. In May 2020, OpenAI, an AI research lab founded by Elon Musk, launched the latest version of an AI-based Natural Language Processing system named GPT-3 that can mimic human language.

When people hear "artificial intelligence," many envision "big data." There's a reason for that: some of the most prominent AI breakthroughs in the past decade have relied on enormous data sets. Image classification made enormous strides in the 2010s thanks to the development of ImageNet, a data set containing millions of images hand sorted into thousands of categories. More recently GPT-3, a language model that uses deep learning to produce humanlike text, benefited from training on hundreds of billions of words of online text. So it is not surprising to see AI being tightly connected with "big data" in the popular imagination.

Reasoning is an essential part of human intelligence and thus has been a long-standing goal in artificial intelligence research. With the recent success of deep learning, incorporating reasoning with deep learning systems, i.e., neuro-symbolic AI has become a major field of interest. We propose the Neuro-Symbolic Forward Reasoner (NSFR), a new approach for reasoning tasks taking advantage of differentiable forward-chaining using first-order logic. The key idea is to combine differentiable forward-chaining reasoning with object-centric (deep) learning. Differentiable forward-chaining reasoning computes logical entailments smoothly, i.e., it deduces new facts from given facts and rules in a differentiable manner.

Convolutional Neural Networks (CNNs) for visual tasks are believed to learn both the low-level textures and high-level object attributes, throughout the network depth. This paper further investigates the `texture bias' in CNNs. To this end, we regenerate multiple instances of training examples from each original image, through a process we call `repainting'. The repainted examples preserve the shape and structure of the regions and objects within the scenes, but diversify their texture and color. Our method can regenerate a same image at different daylight, season, or weather conditions, can have colorization or de-colorization effects, or even bring back some texture information from blacked-out areas. The in-place repaint allows us to further use these repainted examples for improving the generalization of CNNs. Through an extensive set of experiments, we demonstrate the usefulness of the repainted examples in training, for the tasks of image classification (ImageNet) and object detection (COCO), over several state-of-the-art network architectures at different capacities, and across different data availability regimes.