Kategorie: KI Trends

When making financial decisions, it’s important to look at the big picture — say, one taken from a drone, satellite or AI-powered sensor.

The emerging field of spatial finance harnesses AI insights from remote sensors and aerial imagery to help banks, insurers, investment firms and businesses analyze risks and opportunities, enable new services and products, measure the environmental impact of their holdings, and assess damage after a crisis.

Spatial finance applications include monitoring assets, modeling energy efficiency, tracking emissions and pollution, detecting illegal mining and deforestation, and analyzing the risks of natural disasters. NVIDIA AI software and hardware can help the industry combine their business data with geospatial data to accelerate these applications.

By better understanding the environmental and social risks associated with an investment, the financial sector can choose to prioritize those that are more likely to support sustainable development — a framework known as environmental, social and governance (ESG).

Focus on sustainable investments is growing: A Bloomberg Intelligence analysis estimated that ESG assets will represent more than a third of total managed assets worldwide by 2025. And a report by the European Union Agency for the Space Programme predicts that the insurance and finance industry will become the top consumer of Earth observation data and services over the next decade — resulting in more than $1 billion in total revenue by 2031.

Several members of NVIDIA Inception, a global program that supports cutting-edge startups, are advancing these efforts with GPU-accelerated AI applications that can track water pollution near industrial plants, quantify the financial risk of wildfires, assess damage after storms and more.

Powerful Compute for Large-Scale Data

GPU-accelerated AI and data science can rapidly extract insights from complex, unstructured data — enabling banks and businesses to set up real-time streaming and analysis of data as it’s captured from satellites, drones, antennas and edge sensors.

By monitoring aerial imagery — available for free from public space agencies, or at higher granularity from private companies — analysts can get a clear view of how much water is being used from a reservoir over time, how many trees are being cut down for a construction project or how many homes were damaged by a tornado.

This capability can help audit investments by verifying the accuracy of written records such as government-mandated disclosures, environmental impact reports or even insurance claims.

For example, investors might track the supply chain of a company that reports it has achieved net zero in its production line, and discover that it actually relies on an overseas plant emitting coal ash visible in satellite images. Or, sensors that analyze heat emissions from buildings could help identify low-emitting businesses for a tax credit.

NVIDIA’s edge computing solutions, including the NVIDIA Jetson platform for autonomous machines and other embedded applications, are powering numerous AI initiatives in spatial finance.

In addition to using NVIDIA hardware to speed up their applications, developers are adopting software including the NVIDIA DeepStream software development kit for streaming analytics, part of the NVIDIA Metropolis platform for vision AI. They’re also using the NVIDIA Omniverse platform for building and operating metaverse applications for detailed, 3D visualizations of geospatial data.

Insuring Property — From Assessing Risks to Accelerating Claims

NVIDIA Inception members are developing GPU-accelerated applications that turn geospatial data into insights for insurance companies, reducing the number of expensive onsite visits needed to monitor the status of insured properties.

RSS-Hydro, based in Luxembourg, uses GPU computing on premises and in the cloud to train FloodSENS, a machine learning app that maps flood impact from satellite images. The company also uses NVIDIA Omniverse to animate FloodSENS in 3D, helping the team more effectively communicate flood risks and inform resource allocation planning during emergencies.

Toronto-based Ecopia AI uses deep learning-based mapping systems to mine geospatial data, helping to produce next-generation digital maps with highly accurate segmentation of buildings, roads, forests and more. These maps power diverse applications across the public and private sectors, including government climate resilience initiatives and insurance risk assessment. Ecopia uses NVIDIA GPUs to develop its AI models.

CrowdAI, based in the San Francisco Bay Area, uses deep learning tools to accelerate the insurance claims process by automatically analyzing aerial images and videos to detect assets that were damaged or destroyed in natural disasters. The company uses NVIDIA GPUs for both training and inference.

CrowdAI’s deep learning model detected buildings from this aerial image taken in the aftermath of Hurricane Michael in 2018. The AI also categorizes the level of damage – ranging from green representing no damage; to yellow and orange for minor and major damage, respectively; to purple for destroyed buildings. Image credit: CrowdAI, Inc., DigitalGlobe, NOAA, and Nearmap.

Predicting Risks and Opportunities for Businesses

Inception startups are also using geospatial data to help government groups and banks quantify the risks and opportunities of their investments — such as predicting crop yields, detecting industrial pollution and measuring the land and water use of an asset.

Switzerland-based Picterra is supporting sustainable finance with a geospatial MLOps platform that enables banks, insurance companies and financial consultancies to analyze ESG metrics. The company’s AI-driven insights can help the financial industry make investment decisions, model risk and quickly quantify vulnerabilities and opportunities in investment portfolios. The company uses NVIDIA Tensor Core GPUs and the NVIDIA CUDA Toolkit to develop its AI models, which process raw data from satellite, drone and aerial imagery.

London-based Satellite Vu, a startup applying satellite technology to address global challenges, will be able to monitor the temperature of any building on the planet in near real time using infrared camera data. These infrared images will provide its customers with insights about the economic activity, the energy efficiency of buildings, the urban heat island effect and more.

And Sourcenergy, based in Houston, uses geospatial data to power an energy supply chain intelligence platform that can help the financial services industry with market research. Its AI tools, developed using NVIDIA A100 GPUs, enable investors to independently create real-time models of energy companies’ well inventories and project costs, giving them insights even before the companies share data in their quarterly earnings reports.

Learn more about NVIDIA’s work in financial services, and read more on geospatial AI in investment management in chapter 10 of this handbook.

A trio of top scientists is helping lead one of the most ambitious efforts in the history of computing — building a digital twin of Earth.

Peter Bauer, Bjorn Stevens and Francisco “Paco” Doblas-Reyes agree that a digital twin of Earth needs to support resolutions down to a kilometer so a growing set of users can explore the risks of climate change and how to adapt to them. They say the work will require accelerated computing, AI and lots of collaboration.

Their Herculean efforts, some already using NVIDIA technologies, inspired Earth-2, NVIDIA’s contribution to the common cause.

“We will dedicate ourselves and our significant resources to direct NVIDIA’s scale and expertise in computational sciences, to join with the world’s climate science community,” Jensen Huang, founder and CEO of NVIDIA, had said when he announced the Earth-2 initiative in late 2021.

Collaborating on an Unprecedented Scale

Huang’s commitment signaled support for efforts like Destination Earth (DestinE), a pan-European project to create digital twins of the planet.

“No single computer may be enough to do it, so it needs a distributed, international effort,” said Bauer, a veteran with more than 20 years at Europe’s top weather forecasting center who now leads the project that aims to make planet-scale models available by 2030.

Last year, he co-authored a Nature article that said the work “requires collaboration on an unprecedented scale.”

Bauer calls for broad international cooperation on a new Earth information system.

In a March GTC talk, Bauer envisioned a federation that “mobilizes resources from many countries, including private players, and NVIDIA could be one that would be very interesting.”

Peter Bauer

Such resources would enable the enormous work of developing new numeric and machine-learning models, then running them in massive inference jobs to make predictions that stretch across multiple decades.

DestinE has its roots in a 2008 climate conference. It’s the fruit of a number of programs, including many Bauer led in his years with the European Centre for Medium-Range Weather Forecasts — based in Reading, England — which develops some of the most advanced weather forecast models in the world.

Consuming a Petabyte a Day

The collaboration is broad because the computing requirements are massive.

Francisco Doblas-Reyes

“We’re talking about producing petabytes of data a day that have to be delivered very quickly,” said Doblas-Reyes, director of the Earth sciences department at the Barcelona Supercomputing Center, a lead author at Intergovernmental Panel on Climate Change — a group that creates some of the most definitive reports on climate change — and a contributor to the DestinE program.

The digital twin effort will turn the traditional approach to weather and climate forecasting “upside down so users can be the drivers of the process,” he said in a March talk at GTC, NVIDIA’s developer conference. The goal is to “put the user at the helm of producing climate information that’s more useful for climate adaptation,” he said.

His talk described the new models, workflows and systems needed to capture in detail the chaotic nature of climate systems.

Articulating the Vision

The vision for a digital twin crystalized in a keynote at the SC20 supercomputing conference from Stevens, a director at the Max Planck Institute for Meteorology, in Hamburg. He leads work on one of the world’s top weather models for climate applications, as well as an effort to enable simulations at kilometer-level resolution, an order of magnitude finer than today’s best work.

“We need a new type of computing capability … for planetary information systems that let us work through the consequences of our actions and policies, so we can build a more sustainable future,” he said.

Stevens’ landmark talk at SC20 crystallized the vision of a digital twin of Earth.

Stevens described a digital twin that’s accurate and interactive. For example, he imagined people querying it to see how a warming climate could affect flooding in northern Europe or food security in Africa.

AI Enables Interactive Simulations

AI will play a lead role in giving users that level of interactivity, he said in a talk at GTC last year.

“We need AI to get to where we need to be,” he said, giving shout-outs to NVIDIA and colleagues, including Bauer and Doblas-Reyes. “Real steps forward come from people bringing their different perspectives together and rethinking how we work.”

Climate simulations pursue ultra-high resolution for greater accuracy.

Doblas-Reyes agreed in his GTC talk this year.

“In my opinion, AI is a necessary complement for the digital twin — it’s the only way to offer true interactivity to users and help provide a good trajectory of what’s to come in our climate,” he said.

On a Journey Together

All three scientists gave examples of how NVIDIA technologies have been used in a wide variety of projects addressing climate change.

In his GTC talk, Stevens took a characteristically playful turn. He showed a cartoon version of Huang, like Isaac Newton, struck with a falling apple and an insight for how to engage with the scientific effort.

“We need you Jensen, and you need us,” Stevens said.

Stevens playfully portrayed Huang as Isaac Newton in his GTC talk.

The MareNostrum 5 system coming to the Barcelona center provides one example. It’s expected to accelerate some of the DestinE work on NVIDIA H100 Tensor Core GPUs.

Building a digital twin of Earth is “an exciting opportunity to re-think the future of HPC with AI on top,” said Mike Pritchard, a veteran climate scientist who directs climate research at NVIDIA.

NVIDIA Omniverse for connecting 3D tools and developing metaverse applications, NVIDIA Modulus for physics-informed machine learning and NVIDIA Triton for AI inference all have roles to play in the broad effort, he said.

It’s a long and evolving collaboration, Bauer said in his GTC talk. “I sent my first email to NVIDIA on these issues 14 years ago, and NVIDIA has been with us on this journey ever since.”

To learn more, read the concept paper developed for the Berlin Summit for Earth Virtualization Engines, July 3-7, where Huang will deliver a keynote address.

It worked like magic. Computer vision algorithms running in a data center saw that a disease was about to infect a distant wheat field in India.

Sixteen days later, workers in the field found the first evidence of the outbreak.

It was the kind of wizardry people like Vinay Indraganti call digital transformation. He’s practiced it for 25 years, the last dozen of them at companies like Ingredion, a Fortune 500 food-ingredient producer.

The India project was the first big test of AGRi360 — a product suite for sustainable agriculture powered by NVIDIA Metropolis — from the startup that Indraganti co-founded, Blu Cocoon Digital.

Mobile App Taps Cloud Smarts

The pilot was both simple and effective.

Farm workers took pictures of the plants, time-stamped and geotagged by a mobile app. They sent them to the Microsoft Azure cloud, where Blu Cocoon’s custom models found patterns that enabled their uncanny prediction.

Thanks to his background in the industry, Indraganti knows the value of such timely intelligence. It can help farmers and their entire food chain of vendors reap a bumper harvest.

“It’s a vast area, that’s why we’ve made ‘AI for food’ our mantra at Blu Cocoon,” he said in an interview from the suburban Chicago office of the company headquartered in Kolkata.

A Third Eye on the Field

AGRi360 acts “like a third eye in the field,” said Pinaki Bhattacharya, a microbiologist who heads R&D at Blu Cocoon Digital.

AGRi360 puts a dashboard of AI-powered tool in farmers’ hands.

In the pilot, it gave farmers an early warning to apply a small amount of pesticide to arrest the disease. An agrochemical company got a heads up about conditions in the area, helping it manage its supply chain.

In the future, food producers that buy the crops will get key details about their microbiology. That helps in planning exactly how and when to process the crops into products to meet the regulatory requirements where they’ll be sold.

“AGRi360 captures all these insights thanks to AI fed by pictures from farmworkers taken while they’re doing their regular jobs,” Bhattacharya said.

Evaluating Seeds and Soils

The AI models got their start in research using computer vision to quickly assess soil conditions and the quality of seeds.

Those skills are now part of the AGRi360 product portfolio along with products that monitor plant health and best practices in farming. Today, AGRi360 is in use in two countries, improving the quantity and quality of crop yields.

One customer reports it’s on track to source 100% of its products sustainably by 2025. Another saw revenues for an insecticide rise, thanks to the service.

“Our sales of Cartap 50sp grew 70% in six months thanks to AGRi360’s ability to identify emerging crop infections early,” said Vandan Churiwal, a director at Krishi Rayasan, a leading agrochemical supplier based in Kolkata.

“As a result, we’re expanding our license with Blu Cocoon to bring AI-powered insights into every area of our business,” he said.

Faster Training and Inference

Initially, the startup used CPUs to train and run its AI models. Now it exclusively uses NVIDIA GPUs and the Metropolis framework for computer vision.

“It used to take us two months to train a single AI model on CPUs,” said Indraganti. “Now, with NVIDIA A10 Tensor Core GPUs, all four models in AGRi360 can be trained in a few hours — that’s a game changer.”

The time savings add up quickly because the models need to be retrained for new crops, variants and soil types.

GPUs reduced the time to complete inference jobs, too. Predictions that require 15-20 minutes on CPUs get generated in 2-3 seconds on NVIDIA T4 Tensor Core GPUs. The speed also enables Blu Cocoon to test its models on large and growing datasets.

From Shipyards to Snack Bars

Looking ahead, Blu Cocoon is extending its work in the food supply chain into managing containers in shipyards. It’s already testing computer vision models for a customer in India.

“We’ve figured out a way to optimize movement of containers, reducing their time in the yard and minimizing touch points to save time and money,” said Indraganti.

The startup is even helping food producers create recipes with AI. It’s already cooked up a gluten-free muffin for one packaged-foods client with plant-based cheeses, shakes and snack bars next on the menu.

One customer reports the AI-powered system helped reduce the time to create a new recipe by 80%.

“We named the company Blu Cocoon Digital because we look beyond the horizon and across the ocean for ways to nurture our customers’ aspirations with digital technology — and it all runs on the NVIDIA platform and Microsoft Azure,” he said.

Read about Monarch Tractor to learn other ways AI is advancing agriculture.

Leading users and industry-standard benchmarks agree: NVIDIA H100 Tensor Core GPUs deliver the best AI performance, especially on the large language models (LLMs) powering generative AI.

H100 GPUs set new records on all eight tests in the latest MLPerf training benchmarks released today, excelling on a new MLPerf test for generative AI. That excellence is delivered both per-accelerator and at-scale in massive servers.

For example, on a commercially available cluster of 3,584 H100 GPUs co-developed by startup Inflection AI and operated by CoreWeave, a cloud service provider specializing in GPU-accelerated workloads, the system completed the massive GPT-3-based training benchmark in less than eleven minutes.

“Our customers are building state-of-the-art generative AI and LLMs at scale today, thanks to our thousands of H100 GPUs on fast, low-latency InfiniBand networks,” said Brian Venturo, co-founder and CTO of CoreWeave. “Our joint MLPerf submission with NVIDIA clearly demonstrates the great performance our customers enjoy.”

Top Performance Available Today

Inflection AI harnessed that performance to build the advanced LLM behind its first personal AI, Pi, which stands for personal intelligence. The company will act as an AI studio, creating personal AIs users can interact with in simple, natural ways.

“Anyone can experience the power of a personal AI today based on our state-of-the-art large language model that was trained on CoreWeave’s powerful network of H100 GPUs,” said Mustafa Suleyman, CEO of Inflection AI.

Co-founded in early 2022 by Mustafa and Karén Simonyan of DeepMind and Reid Hoffman, Inflection AI aims to work with CoreWeave to build one of the largest computing clusters in the world using NVIDIA GPUs.

Tale of the Tape

These user experiences reflect the performance demonstrated in the MLPerf benchmarks announced today.

H100 GPUs delivered the highest performance on every benchmark, including large language models, recommenders, computer vision, medical imaging and speech recognition. They were the only chips to run all eight tests, demonstrating the versatility of the NVIDIA AI platform.

Excellence Running at Scale

Training is typically a job run at scale by many GPUs working in tandem. On every MLPerf test, H100 GPUs set new at-scale performance records for AI training.

Optimizations across the full technology stack enabled near linear performance scaling on the demanding LLM test as submissions scaled from hundreds to thousands of H100 GPUs.

In addition, CoreWeave delivered from the cloud similar performance to what NVIDIA achieved from an AI supercomputer running in a local data center. That’s a testament to the low-latency networking of the NVIDIA Quantum-2 InfiniBand networking CoreWeave uses.

In this round, MLPerf also updated its benchmark for recommendation systems.

The new test uses a larger data set and a more modern AI model to better reflect the challenges cloud service providers face. NVIDIA was the only company to submit results on the enhanced benchmark.

An Expanding NVIDIA AI Ecosystem

Nearly a dozen companies submitted results on the NVIDIA platform in this round. Their work shows NVIDIA AI is backed by the industry’s broadest ecosystem in machine learning.

Submissions came from major system makers that include ASUS, Dell Technologies, GIGABYTE, Lenovo, and QCT. More than 30 submissions ran on H100 GPUs.

This level of participation lets users know they can get great performance with NVIDIA AI both in the cloud and in servers running in their own data centers.

Performance Across All Workloads

NVIDIA ecosystem partners participate in MLPerf because they know it’s a valuable tool for customers evaluating AI platforms and vendors.

The benchmarks cover workloads users care about — computer vision, translation and reinforcement learning, in addition to generative AI and recommendation systems.

Users can rely on MLPerf results to make informed buying decisions, because the tests are transparent and objective. The benchmarks enjoy backing from a broad group that includes Arm, Baidu, Facebook AI, Google, Harvard, Intel, Microsoft, Stanford and the University of Toronto.

MLPerf results are available today on H100, L4 and NVIDIA Jetson platforms across AI training, inference and HPC benchmarks. We’ll be making submissions on NVIDIA Grace Hopper systems in future MLPerf rounds as well.

The Importance of Energy Efficiency

As AI’s performance requirements grow, it’s essential to expand the efficiency of how that performance is achieved. That’s what accelerated computing does.

Data centers accelerated with NVIDIA GPUs use fewer server nodes, so they use less rack space and energy. In addition, accelerated networking boosts efficiency and performance, and ongoing software optimizations bring x-factor gains on the same hardware.

Energy-efficient performance is good for the planet and business, too. Increased performance can speed time to market and let organizations build more advanced applications.

Energy efficiency also reduces costs because data centers accelerated with NVIDIA GPUs use fewer server nodes. Indeed, NVIDIA powers 22 of the top 30 supercomputers on the latest Green500 list.

Software Available to All

NVIDIA AI Enterprise, the software layer of the NVIDIA AI platform, enables optimized performance on leading accelerated computing infrastructure. The software comes with the enterprise-grade support, security and reliability required to run AI in the corporate data center.

All the software used for these tests is available from the MLPerf repository, so virtually anyone can get these world-class results.

Optimizations are continuously folded into containers available on NGC, NVIDIA’s catalog for GPU-accelerated software.

Read this technical blog for a deeper dive into the optimizations fueling NVIDIA’s MLPerf performance and efficiency.

Editor’s note: This post is a part of our Meet the Omnivore series, which features individual creators and developers who accelerate 3D workflows and create virtual worlds using NVIDIA Omniverse, a development platform built on Universal Scene Description, aka OpenUSD.

As augmented reality (AR) becomes more prominent and accessible across the globe, Kiryl Sidarchuk is helping to erase the border between the real and virtual worlds.

Kiryl Sidarchuk

Co-founder and CEO of AR-Generation, which is a member of the NVIDIA Inception program for cutting-edge startups, Sidarchuk with his company developed MagiScan, an AI-based 3D scanner app.

It lets users capture any object with their smartphone camera and quickly creates a high-quality, detailed 3D model of it for use in any AR or metaverse application.

AR-Generation now offers an extension that enables direct export of 3D models from MagiScan to NVIDIA Omniverse, a development platform for connecting and building 3D tools and metaverse applications.

It’s made possible with speed and ease by Universal Scene Description, aka OpenUSD, an extensible framework that serves as a common language between digital content-creation tools.

“Augmented reality will become an integral part of everyday life,” said Sidarchuk, who’s based in Nicosia, Cyprus. “We customized our app to allow export of 3D models based on real-world objects directly to Omniverse, enabling users to showcase the models in AR and integrate them into any metaverse or game.”

Omniverse extensions are core building blocks that let anyone create and extend functions of Omniverse apps using the popular Python or C++ programming languages.

It was simple and convenient for AR-Generation to build the extension, Sidarchuk said, thanks to easily accessible documentation, as well as technical guidance from NVIDIA teams, free AWS credits and networking opportunities with other AI-driven companies — all benefits of being a part of NVIDIA Inception.

Capture, Click and Create 3D Models From Real-World Objects 

Sidarchuk estimates that MagiScan can create 3D models from objects 10x faster and at up to 100x less cost than it would take a designer to do so manually.

This frees creators up to focus on fine-tuning their work and makes AR more accessible to all through a simple app.

AR-Generation chose to build an extension for Omniverse because the platform “provides a convenient environment that integrates all the tools for working with 3D and generative AI,” said Sidarchuk. “Plus, we can collaborate and exchange ideas with colleagues in real time.”

Export 3D models from MagiScan to Omniverse with OpenUSD.

Sidarchuk’s favorite feature of Omniverse is its OpenUSD compatibility, which enables seamless interchange of 3D data between creative applications. “OpenUSD is the format of the future,” he said.

Based on this framework, the MagiScan extension for Omniverse enables fast, affordable creation of high-quality 3D models for any object. MagiScan is available for download on iOS and Android devices.

“It can help everyone from individuals to large corporations save time and money in digitalization,” said Sidarchuk, who claims his first word as a toddler was “money.”

The business-oriented developer started his first company at age 16. It was a one-man endeavor, buying fresh fruits and vegetables from a small village and selling them in Minsk, the capital of Belarus. “That’s how I earned enough to buy my first car,” he mused.

More than a dozen years later, when he’s not working to “enhance human capabilities through augmented-reality technologies,” he said, Sidarchuk now spends his free time with his five-year-old daughter, Aurora.

Watch Sidarchuk discuss 3D modeling, AI and AR on a replay of his Omniverse livestream on demand, and learn more about the MagiScan extension for Omniverse.

Join In on the Creation

Anyone can build their own Omniverse extension or Connector to enhance their 3D workflows and tools. Creators and developers across the world can download NVIDIA Omniverse for free, and enterprise teams can use the platform for their 3D projects.

Check out artwork from other “Omnivores” and submit projects in the gallery. Connect your workflows to Omniverse with software from Adobe, Autodesk, Epic Games, Maxon, Reallusion and more.

Get started with NVIDIA Omniverse by downloading the standard license free, or learn how Omniverse Enterprise can connect your team. Developers can get started with Omniverse resources and learn about OpenUSD. Explore the growing ecosystem of 3D tools connected to Omniverse.

Stay up to date on the platform by subscribing to the newsletter, and follow NVIDIA Omniverse on Instagram, Medium and Twitter. For more, join the Omniverse community and check out the Omniverse forums, Discord server, Twitch and YouTube channels. 

While generative AI is a relatively new household term, drug discovery company Insilico Medicine has been using it for years to develop new therapies for debilitating diseases.

The company’s early bet on deep learning is bearing fruit — a drug candidate discovered using its AI platform is now entering Phase 2 clinical trials to treat idiopathic pulmonary fibrosis, a relatively rare respiratory disease that causes progressive decline in lung function.

Insilico used generative AI for each step of the preclinical drug discovery process: to identify a molecule that a drug compound could target, generate novel drug candidates, gauge how well these candidates would bind with the target, and even predict the outcome of clinical trials.

Doing this using traditional methods would have cost more than $400 million and taken up to six years. But with generative AI, Insilico accomplished them for one-tenth of the cost and one-third of the time — reaching the first phase of clinical trials just two and a half years after beginning the project.

“This first drug candidate that’s going to Phase 2 is a true highlight of our end-to-end approach to bridge biology and chemistry with deep learning,” said Alex Zhavoronkov, CEO of Insilico Medicine. “This is a significant milestone not only for us, but for everyone in the field of AI-accelerated drug discovery.”

Insilico is a premier member of NVIDIA Inception, a free program that provides cutting-edge startups with technical training, go-to-market support and AI platform guidance. The company uses NVIDIA Tensor Core GPUs in its generative AI drug design engine, Chemistry42, to generate novel molecular structures — and was one of the first adopters of an early precursor to NVIDIA DGX systems in 2015.

AI Enables End-to-End Preclinical Drug Discovery

Insilico’s Pharma.AI platform includes multiple AI models trained on millions of data samples for a range of tasks. One AI tool, PandaOmics, rapidly identifies and prioritizes targets that play a significant role in a disease’s effectiveness — like the infamous spike protein on the virus that causes COVID-19.

The Chemistry42 engine can design within days new potential drug compounds that target the protein identified using PandaOmics. The generative chemistry tool uses deep learning to come up with drug-like molecular structures from scratch.

“Typically, AI companies in drug discovery focus either on biology or on chemistry,” said Petrina Kamya, head of AI platforms at Insilico. “From the start, Insilico has been applying the same deep learning approach to both fields, using AI both to discover drug targets and generate chemical structures of small molecules.”

Over the years, the Insilico team has adopted different kinds of deep neural networks for drug discovery, including generative adversarial networks and transformer models. They’re now using NVIDIA BioNeMo to accelerate the early drug discovery process with generative AI.

Finding the Needle in the AI Stack

To develop its pulmonary fibrosis drug candidate, Insilico used Pharma.AI to design and synthesize about 80 molecules, achieving unprecedented success rates for preclinical drug candidates. The process — from identifying the target to nominating a promising drug candidate for trials — took under 18 months.

During Phase 2 clinical trials, Insilico’s pulmonary fibrosis drug will be tested in several hundred people with the condition in the U.S. and China. The process will take several months — but in parallel, the company has more than 30 programs in the pipeline to target other diseases, including a number of cancer drugs.

“When we first presented our results, people just did not believe that generative AI systems could achieve this level of diversity, novelty and accuracy,” said Zhavoronkov. “Now that we have an entire pipeline of promising drug candidates, people are realizing that this actually works.”

Learn more about Insilico Medicine’s Chemistry42 platform for AI-accelerated drug candidate screening in this talk from NVIDIA GTC.

Subscribe to NVIDIA healthcare news and generative AI news.

Researchers at Yamagata University in Japan have harnessed AI to uncover four previously unseen geoglyphs — images on the ground, some as wide as 1,200 feet, made using the land’s elements — in Nazca, a seven-hour drive south of Lima, Peru.

The geoglyphs — a humanoid, a pair of legs, a fish and a bird — were revealed using a deep learning model, making the discovery process significantly faster than traditional archaeological methods.

The team’s deep learning model training was executed on an IBM Power Systems server with an NVIDIA GPU.

Using open-source deep learning software, the researchers analyzed high-resolution aerial photographs, a technique that was part of a study that began in November 2019.

Published this month in the Journal of Archaeological Science, the study confirms the deep learning model’s findings through onsite surveys and highlights the potential of AI in accelerating archaeological discoveries.

The deep learning techniques that comprise the hallmark of modern AI are used for various archeological efforts, whether analyzing ancient scrolls discovered across the Mediterranean or categorizing pottery sherds from the American Southwest.

The Nazca lines, a series of ancient geoglyphs that date from 500 B.C. to 500 A.D. — primarily likely from 100 B.C. to 300 A.D. — were created by removing darker stones on the desert floor to reveal lighter-colored sand beneath.

The drawings — depicting animals, plants, geometric shapes and more — are thought to have had religious or astronomical significance to the Nazca people who created them.

The discovery of these new geoglyphs indicates the possibility of more undiscovered sites in the area.

And it underscores how technology like deep learning can enhance archaeological exploration, providing a more efficient approach to uncovering hidden archaeological sites.

Read the full paper.

Featured image courtesy of Wikimedia Commons.

Detecting delirium isn’t easy, but it can have a big payoff: speeding essential care to patients, leading to quicker and surer recovery.

Improved detection also reduces the need for long-term skilled care, enhancing the quality of life for patients while decreasing a major financial burden. In the U.S., caring for those suffering from delirium costs up to $64,000 a year per patient, according to the National Institutes of Health.

In a paper published last month in Nature, researchers describe how they used a deep learning model called Vision Transformer, accelerated by NVIDIA GPUs, alongside a rapid-response electroencephalogram, or EEG, device to detect delirium in critically ill older adults.

The paper, called “Supervised deep learning with vision transformer predicts delirium using limited lead EEG,” is authored by Malissa Mulkey of the University of South Carolina, Huyunting Huang of Purdue University, Thomas Albanese and Sunghan Kim of the University of East Carolina, and Baijian Yang of Purdue.

Their innovative approach achieved a testing accuracy rate of 97%, promising a potential breakthrough in forecasting dementia. And by harnessing AI and EEGs, the researchers could objectively evaluate prevention and treatment methods, leading to better care.

This impressive result is due in part to the accelerated performance of NVIDIA GPUs, enabling the researchers to accomplish their tasks in half the time compared to CPUs.

Delirium affects up to 80% of critically ill patients. Yet conventional clinical detection methods identify fewer than 40% of cases — representing a significant gap in patient care. Presently, screening ICU patients involves a subjective bedside assessment.

The introduction of handheld EEG devices could make screening more accurate and affordable, but the lack of skilled technicians and neurologists poses a challenge.

The use of AI, however, can eliminate the need for a neurologist to interpret findings and allow for the detection of changes associated with delirium roughly two days before symptom onset, when patients are more receptive to treatment. It also makes it possible to use EEGs with minimal training.

The researchers applied an AI model called ViT, initially created for natural language processing and accelerated by NVIDIA GPUs, to EEG data — offering a fresh approach to data interpretation.

The use of a handheld rapid-response EEG device, which doesn’t require large EEG machines or specialized technicians, was another noteworthy study finding.

This practical tool, combined with advanced AI models for interpreting the data they collect, could streamline delirium screenings in critical care units.

The research presents a promising method for delirium detection that could shorten hospital stays, increase discharge rates, decrease mortality rates and reduce the financial burden associated with delirium.

By integrating the power of NVIDIA GPUs with innovative deep learning models and practical medical devices, this study underlines the transformative potential of technology in enhancing patient care.

As AI grows and develops, medical professionals are increasingly likely to rely on it to forecast conditions like dementia and intervene early, revolutionizing the future of critical care.

Read the full paper.

Conquer the lands in Microsoft’s award-winning Age of Empires III: Definitive Edition. It leads 10 new games supported today on GeForce NOW.

At Your Command

Stream battles all from the cloud.

Age of Empires III: Definitive Edition is a remaster of one of the most beloved real-time strategy franchises featuring improved visuals, enhanced gameplay, cross-platform multiplayer and more. Command mighty civilizations from across Europe and the Americas or jump to the battlefields of Asia. Members can experience two new game modes: Historical Battles and The Art of War Challenge Missions. Two new nations also join this edition — Sweden and the Inca — each with advantages for conquering the New World.

Build an empire today and stream across devices in glorious 4K resolution with an Ultimate membership.

Conquer Your Games List

Master the art of siege tactics in “Conqueror’s Blade” this week.

The GeForce NOW library is always expanding. Take a look at the 10 newly supported games this week.

Aliens: Dark Descent (New release on Steam, June 20)
Trepang2 (New release on Steam, June 21)
Forever Skies (New release on Steam, June 22)
Age of Empires III: Definitive Edition (Steam)
A.V.A Global (Steam)
Bloons TD 6 (Steam)
Conqueror’s Blade (Steam)
Layers of Fear (Steam)
Park Beyond (Steam)
Tom Clancy’s Rainbow Six Extraction (Steam)

Before diving into the weekend, let us know your answer to our question of the week on Twitter or in the comments below. Happy streaming!

You’ve been chosen to build the greatest empire in history.

What time period are you choosing to build it in?

— NVIDIA GeForce NOW (@NVIDIAGFN) June 14, 2023

Editor’s note: This post is part of our weekly In the NVIDIA Studio series, which celebrates featured artists, offers creative tips and tricks, and demonstrates how NVIDIA Studio technology improves creative workflows. We’re also deep diving on new GeForce RTX 40 Series GPU features, technologies and resources, and how they dramatically accelerate content creation.

Amir Anbarestani, an accomplished 3D artist who goes by the moniker Kingsletter, had a “shell of a good time” creating his Space Turtle scene this week In the NVIDIA Studio.

Kingsletter has always harbored a fascination with 3D art, he said. As a child, he often enjoyed exploring and crafting within immersive environments. Whether it was playing with plasticine — putty-like modeling material — or creating pencil drawings, his innate inclination for self-expression always found resonance within the expansive domain of 3D.

Space Turtle with MSI creator Z17HX courtesy of @AustraliaMSI & @NVIDIAStudio

Unleash your creativity with NVIDIA Studio drivers!

Get yours at: https://t.co/idJlWgb8UX pic.twitter.com/Ff6Y6RfQp4

— King’s Letter (@TheKingsletter) April 28, 2023

Below, he shares his inspiration and creative process using ZBrush, Adobe Substance 3D Painter and Blender.

An NVIDIA DLSS 3 plug-in is now available in Unreal Engine 5, offering select benefits including AI upscaling for high frame rates, super resolution and more for GeForce RTX 40 Series owners.

And 3D creative app Marvelous Designer launches Into the Omniverse its NVIDIA Omniverse Connector this month. Learn how talented artists are using the Connector, along with the Universal Scene Description (“OpenUSD”) framework, to elevate their creative workflows.

NVIDIA DLSS 3 Plug-In Is Unreal — Engine 5

NVIDIA Studio released a DLSS 3 plug-in compatible with Unreal Engine 5. The Play in Editor tool is useful for game developers to quickly review gameplay in a level while editing — and DLSS 3 AI upscaling will unlock significantly higher frame rates on GeForce RTX 40 Series GPUs for even smoother previewing.

NVIDIA DLSS 3 plug-in unlocks incredible visual details with DLSS 3 in Unreal Engine 5.

Plus, select Unreal Engine viewports offer DLSS 2 Super Resolution and upscaling benefits in typical content-creation workflows like modeling, lighting, animation and more.

Download DLSS 3 for Unreal Engine 5.2, available now. Learn more about NVIDIA technologies supported by Unreal Engine 5.

Turtle Recall 

The process began with sketching and initial sculpting in the ZBrush tool, where the concept of a floating turtle in space took shape and evolved into a dynamic shot of the creature soaring toward the camera.

“It’s remarkable how something as simple as shaping an idea’s basic form can be so immensely gratifying,” said Kingsletter on the blockout phase. “There’s a unique joy in starting with a blank canvas and gradually bringing the essence of a concept to life.”

Sketching and initial sculpting in ZBrush.

After finalizing the model in ZBrush, Kingsletter used ZRemesher to retopologize it, or generate a low-poly version suitable for the intended scene. This is useful for removing artifacts and other mesh issues before animation and rigging.

The RIZOMUV UV mapping 3D software was then deployed for unwrapping the model, the process of opening a mesh to make a 2D texture that covers a 3D object. This is effective for adding textures to objects with precision, a common need for professional artists.

Next, Kingsletter applied surface details, from subtle dusting to extreme wear and tear, with materials mimicking real-world behaviors such as sheen, subsurface scattering and more in Adobe Substance 3D Painter. RTX-accelerated light and ambient occlusion enabled fully baked models in mere seconds.

Textures added and baked rapidly in Adobe Substance 3D Painter.

Kingsletter then moved to Blender to animate the scene, setting up simple rigs and curves to bring the turtle’s flapping limbs and flight to life. Harnessing the potential of his MSI Creator Z17 HX Studio A13V NVIDIA Studio laptop from MSI with GeForce RTX 4070 graphics turtle-ly exceeded the artist’s lofty expectations.

The MSI Creator Z17 HX Studio laptop with GeForce RTX 4070 graphics.

“As a digital creative professional, I always strive to work with the best creative tools available,” Kingsletter said. “Choosing the MSI Creator laptop allowed me to exceed my creative professional needs and indulge in my passionate gaming hobby.”

He enriched the cosmic environment using Blender’s particle system, which scattered random debris, asteroids and a small, rotating planet throughout the outer-space scene. AI-powered RTX-accelerated OptiX ray tracing in the viewport unlocked buttery-smooth interactive animations in the viewport.

Create magnificent worlds in Blender accelerated by GeForce RTX graphics.

“Simulating smoke proved to be the most challenging aspect,” said Kingsletter about his first foray into this form of animation. “Through numerous trials and errors, I persevered until I achieved a truly satisfactory result.”

Realistic smoke elevated the 3D animation.

His RTX 4070 GPU facilitated smoother, more efficient rendering of the final visuals with RTX-accelerated OptiX ray tracing in Blender Cycles, ensuring the fastest final frame render.

When asked what he’d advise his younger artist self, Kingsletter said, “I’d enhance my observation skills. By immersing myself in the intricacies of form and paying careful attention to the world around me, I would have laid a stronger foundation for my creative journey.”

Wise words for all creators.

Digital 3D artist Kingsletter.

Check out Kingsletter’s beautiful 3D creations on Instagram.

Follow NVIDIA Studio on Instagram, Twitter and Facebook. Access tutorials on the Studio YouTube channel and get updates directly in your inbox by subscribing to the Studio newsletter.