A hidden mechanism for achieving glides of hundreds of feet is revealed by computational modelling.
Scientists are currently thinking of ways to create robots resembling the gliding motion of flying snakes, according to a study published today (Dec .13) in Physics of Fluids.
‘Undulations’ encourage lift.
KaraGrubis/iStock.
The researchers anticipate their findings will improve our comprehension of gliding motion and result in a more effective design for future airborne snake robots.
Tel Aviv-based D-ID released today the first multimodal generative AI video platform to combine text, image and animation in one interface. The self-service video platform integrates D-ID’s proprietary generative AI technology with GPT-3 from Open AI and Stable Diffusion from Stability AI, allowing users to generate digital composite faces and speech in 119 languages based on their text prompts.
“This is a game changer for creators,” says Gil Perry, D-ID co-founder and CEO. “It’s the bleeding edge of generative AI,” he asserts, touting the startup’s expertise in deep learning and computer vision. When I talked to Perry last year, he said that the company’s long-term vision is “to lead the next disruption in the video entertainment space by creating AI-generated synthetic media in a responsible way.”
In the rapidly evolving generative AI space, “long-term” means “next year,” so now Perry talks about providing “digital humans” to enterprises, “transforming the way we communicate with machines and elevating our capabilities as humans.” He hopes that sometime next year, we could chat with the digital humans we will create with D-ID’s help.
Let’s talk about AI Art, Lensa, ChatGPT, and why it’s all deeper than you think.
Save money with Karma at https://shop.karmanow.com/marques_dec22 and get double the Karma cash for a limited time!
The sweater: http://shop.MKBHD.com.
Search LAION-5B: https://haveibeentrained.com.
Tech I’m using right now: https://www.amazon.com/shop/MKBHD?tag=lifeboatfound-20
Intro Track: http://youtube.com/20syl.
Ispace Inc. is a private Japanese company developing robotic landers and rovers for missions to the Moon. It aims to compete for both transportation and exploration mission contracts from space agencies and private industry. If successful, these spacecraft and the accompanying vehicles could enable clients to discover, map, and use the natural resources on Earth’s nearest neighbour.
In addition to its headquarters in Tokyo, the company has offices in the United States and Luxembourg, employing around 200 people. Although founded in 2010, its team of engineers had earlier competed in the Google Lunar X Prize.
Following more than a decade of research and development, ispace yesterday launched Hakuto-R Mission 1 – delivered into space on a partially reusable Falcon 9 Block 5 rocket. The spacecraft will now perform orbital manoeuvres, taking it as far as 1.5 million km (932,000 miles) from Earth, before arriving at the Moon sometime in April 2023.
Object detection means all the techniques and means for detecting, identifying, and classifying objects in an image. Recently, the field of artificial intelligence has seen many advances thanks to deep learning and image processing. It is now possible to recognize images or even find objects inside an image. With deep learning, object detection has become very popular with several families of models (R-CNN, YOLO, etc.). However, most of the existing methods in the literature adapt to the training database and fail to generalize when faced with images belonging to different domains.
Although most architectures are optimized for well-known benchmarks, significant results have been achieved using CNNs for tasks particular to a certain domain. However, these domain-specific solutions are often well-tuned for a specific target dataset, starting with carefully chosen architecture and training techniques. This method of training models has the drawback of unnecessarily adapting the approaches to a particular dataset. To address this issue, a research team from Intel offers a different strategy that also serves as the foundation of the Intel® Geti™ platform: a dataset-agnostic template for object detection training made up of carefully selected and pre-trained models and a reliable training pipeline for additional training.
The authors experimented with architectures in three categories: lightweight, extremely accurate, and medium, to develop a scope of the models used for the various object detection datasets regardless of complexity and object size. Pretrained weights are employed to reach model convergence quickly and begin with high accuracy. In addition, a data augmentation operation is performed to augment images with a random crop, horizontal flip, and brightness and color distortions. Multiscale training was applied for medium and accurate models to make them more robust. Additionally, to strike a balance between accuracy and complexity, the authors empirically selected particular resolutions for each model after conducting several trials. Early stopping and the adaptive ReduceOnPlateau scheduler are also used to end training if a few epochs of training do not further improve the outcome.
Achieving a long-sought goal of regenerative medicine.
Researchers from the University of California, San Francisco, (UCSF) engineered molecules that function as “cellular glue,” enabling them to precisely direct how cells bond with each other. This is a significant step toward regenerative medicine’s long-term goal of creating new tissues and organs, according to a press release.
Adhesive molecules are naturally present in the body and keep the tens of trillions of cells together in organized patterns. They build neural networks, develop structures, and direct immune cells to specific areas of the body. Adhesion also makes cell communication easier to maintain the body functioning as a self-regulating whole.
The science of meteorology has taken tremendous strides in the past two decades thanks to a confluence of several inputs: improved computing power; better modeling of data; more observational data points ranging from the device in your hand to the satellites orbiting earth; and advanced data science applications. As recently as two decades ago, providing an accurate forecast three to four days out was considered innovative. Today a five-day forecast is accurate about 80 percent of the time. Most weather experts are predicting even more extended accuracy by 2030 with the application of artificial intelligence for numerical weather prediction output. But beyond improving accuracy, here are a few other forecasting trends to watch in 2023.
Hyper-relevant Forecasting
Just like other sets of analytics have become more tailored, or localized to the user, weather intelligence is bringing forecast relevancy to an individual organization or entity. A business can determine which risks are most significant to their operations, such as wind gusts, lightning, heavy rains, and ice accretion, and then be alerted when those risk thresholds are met. While there’s growing use among utilities, municipalities and other infrastructure decision makers, hyper-relevant forecasting is growing in other sectors. For example, by combining weather data with purchasing trends and consumer demand data, one grocery chain learned that even a small change in temperature can result in a significant shift in what people buy. The store improved its revenues by modeling this impact and managing inventory accordingly. Even sports teams are applying hyper-relevant forecasting for everything from daily stadium operations to food and beverage decisions and strategic game plays.