Posts in category: nVidia AI

Stuttgart’s supercomputer center has been cruising down the autobahn of high performance computing like a well-torqued coupe, and now it’s making a pitstop for some AI fuel.

Germany’s High-Performance Computing Center Stuttgart (HLRS), one of Europe’s largest supercomputing centers, has tripled the size of its staff and increased its revenues from industry collaborations 20x since Michael Resch became director in 2002. In the past year, much of the growth has come from interest in AI.

With demand for machine learning on the rise, HLRS signed a deal to add 192 NVIDIA Ampere architecture GPUs linked on NVIDIA Mellanox InfiniBand network to its Hawk supercomputer based on an Apollo system from Hewlett Packard Enterprise.

Hawk Flies to New Heights

The GPUs will propel what’s already ranked as the world’s 16th largest system to new heights. In preparation for the expansion, researchers are gearing up AI projects that range from predicting the path of the COVID-19 pandemic to the science behind building better cars and planes.

“Humans can create huge simulations, but we can’t always understand all the data — the big advantage of AI is it can work through the data and see its consequences,” said Resch, who also serves as a professor at the University of Stuttgart with a background in engineering, computer science and math.

The center made its first big leap into AI last year when it installed a Cray CS-Storm system with more than 60 NVIDIA GPUs. It is already running AI programs that analyze market data for Mercedes-Benz, investment portfolios for a large German bank and a music database for a local broadcaster.

“It turned out to be an extremely popular system because there’s a growing community of people who understand AI has a benefit for them,” Resch said of the system now running at near capacity. “By the middle of this year it was clear we had to expand to cover our growing AI requirements,” he added.

The New Math: HPC+AI

The future for the Stuttgart center, and the HPC community generally, is about hybrid computing where CPUs and GPUs work together, often to advance HPC simulations with AI.

“Combining the two is a golden bullet that propels us into a better future for understanding problems,” he said.

For example, one researcher at the University of Stuttgart will use data from as many as 2 billion simulations to train neural networks that can quickly and economically evaluate metal alloys. The AI model it spawns could run on a PC and help companies producing sheet metal choose the best alloys for, say, a car door.

“This is extremely helpful in situations where experimentation is difficult or costly,” he said.

And it’s an apropos app for the center situated in the same city that’s home to the headquarters of both Mercedes and Porsche.

In the Flow with Machine Learning

A separate project in fluid dynamics will take a similar approach.

A group from the university will train neural networks on data from highly accurate simulations to create an AI model that can improve analysis of turbulence. It’s a critical topic for companies such as Airbus that are collaborating with HLRS on efforts to mine the aerospace giant’s data on airflow.

The Stuttgart center also aims to use AI as part of a European research project to predict when hospital beds could fill up in intensive-care units amid the pandemic. The project started before the coronavirus hit, but it accelerated in the wake of COVID-19.

Tracking the Pandemic with AI

One of the project’s goals is to give policy makers a four-week window to respond before hospitals would reach their capacity.

“It’s a critical question with so many people dying — we’ve seen scenarios in places like Italy, New York and Wuhan where ICUs filled up in the first weeks of pandemic,” Resch said.

“So, we will conduct simulations and predictions of the outlook for the pandemic over the next weeks and months, and GPUs will be extremely helpful for that,” he added.

It’s perhaps the highest profile of many apps now in the pipeline for the GPU-enhanced engine that will propel Stuttgart’s researchers further down the road on their journey into AI.

It can take a puppy weeks to learn that certain kinds of behaviors will result in a yummy treat, extra cuddles or a belly rub — and that other behaviors won’t. With a system of positive reinforcement, a pet pooch will in time anticipate that chasing squirrels is less likely to be rewarded than staying by their human’s side.

Deep reinforcement learning, a technique used to train AI models for robotics and complex strategy problems, works off the same principle.

In reinforcement learning, a software agent interacts with a real or virtual environment, relying on feedback from rewards to learn the best way to achieve its goal. Like the brain of a puppy in training, a reinforcement learning model uses information it’s observed about the environment and its rewards, and determines which action the agent should take next.

To date, most researchers have relied on a combination of CPUs and GPUs to run reinforcement learning models. This means different parts of the computer tackle different steps of the process — including simulating the environment, calculating rewards, choosing what action to take next, actually taking action, and then learning from the experience.

But switching back and forth between CPU cores and powerful GPUs is by nature inefficient, requiring data to be transferred from one part of the system’s memory to another at multiple points during the reinforcement learning training process. It’s like a student who has to carry a tall stack of books and notes from classroom to classroom, plus the library, before grasping a new concept.

With Isaac Gym, NVIDIA developers have made it possible to instead run the entire reinforcement learning pipeline on GPUs — enabling significant speedups and reducing the hardware resources needed to develop these models.

Here’s what this breakthrough means for the deep reinforcement learning process, and how much acceleration it can bring developers.

Reinforcement Learning on GPUs: Simulation to Action 

When training a reinforcement learning model for a robotics task — like a humanoid robot that walks up and down stairs — it’s much faster, safer and easier to use a simulated environment than the physical world. In a simulation, developers can create a sea of virtual robots that can quickly rack up thousands of hours of experience at a task.

If tested solely in the real world, a robot in training could fall down, bump into or mishandle objects — causing potential damage to its own machinery, the object it’s interacting with or its surroundings. Testing in simulation provides the reinforcement learning model a space to practice and work out the kinks, giving it a head start when shifting to the real world.

In a typical system today, the NVIDIA PhysX simulation engine runs this experience-gathering phase of the reinforcement learning process on NVIDIA GPUs. But for other steps of the training application, developers have traditionally still used CPUs.

A key part of reinforcement learning training is conducting what’s known as the forward pass: First, the system simulates the environment, records a set of observations about the state of the world and calculates a reward for how well the agent did.

The recorded observations become the input to a deep learning “policy” network, which chooses an action for the agent to take. Both the observations and the rewards are stored for use later in the training cycle.

Finally, the action is sent back to the simulator so that the rest of the environment can be updated in response.

After several rounds of these forward passes, the reinforcement learning model takes a look back, evaluating whether the actions it chose were effective or not. This information is used to update the policy network, and the cycle begins again with the improved model.

GPU Acceleration with Isaac Gym 

To eliminate the overhead of transferring data back and forth from CPU to GPU during this reinforcement learning training cycle, NVIDIA researchers have developed an approach to run every step of the process on GPUs. This is Isaac Gym, an end-to-end training environment, which includes the PhysX simulation engine and a PyTorch tensor-based API.

Isaac Gym makes it possible for a developer to run tens of thousands of environments simultaneously on a single GPU. That means experiments that previously required a data center with thousands of CPU cores can in some cases be trained on a single workstation.

Decreasing the amount of hardware required makes reinforcement learning more accessible to individual researchers who don’t have access to large data center resources. It can also make the process a lot faster.

A simple reinforcement learning model tasked with getting a humanoid robot to walk can be trained in just a few minutes with Isaac Gym. But the impact of end-to-end GPU acceleration is most useful for more challenging tasks, like teaching a complex robot hand to manipulate a cube into a specific position.

This problem requires significant dexterity by the robot, and a simulation environment that involves domain randomization, a mechanism that allows the learned policy to more easily transfer to a real-world robot.

Research by OpenAI tackled this task with a cluster of more than 6,000 CPU cores plus multiple NVIDIA Tensor Core GPUs — and required about 30 hours of training for the reinforcement learning model to succeed at the task 20 times in a row using a feed-forward network model.

Using just one NVIDIA A100 GPU with Isaac Gym, NVIDIA developers were able to achieve the same level of success in around 10 hours — a single GPU outperforming an entire cluster by a factor of 3x.

To learn more about Isaac Gym, visit our developer news center.

Video above shows a cube manipulation task trained by Isaac Gym on a single NVIDIA A100 GPU and rendered in NVIDIA Omniverse.

Anima Anandkumar is setting a personal record this week with seven of her team’s research papers accepted to NeurIPS 2020.

The 34th annual Neural Information Processing Systems conference is taking place virtually from Dec. 6-12. The premier event on neural networks, NeurIPS draws thousands of the world’s best researchers every year.

Anandkumar, NVIDIA’s director of machine learning research and Bren professor at CalTech’s CMS Department, joined AI Podcast host Noah Kravitz to talk about what to expect at the conference, and to explain what she sees as the future of AI.

The papers that Anandkumar and her teams at both NVIDIA and CalTech will be presenting are focused on topics including how to design more robust priors that improve network perception and how to create useful benchmarks to evaluate where neural networks need to improve.

In terms of what Anandkumar is focused on going forward, she continues to work on the transition from supervised to unsupervised and self-supervised learning, which she views as the key to next-generation AI.

Key Points From This Episode:

• Anandkumar explains how her interest in AI grew from a love of math at a young age as well as influence from her family — her mother was an engineer and her grandfather a math teacher. Her family was also the first in their city to have a CNC machine — an automated machine, such as a drill or lathe, controlled by a computer — which sparked an interest in programming.
• Anandkumar was instrumental in spearheading the development of tensor algorithms, which are crucial in achieving massive parallelism in large-scale AI applications. That’s one reason for her enthusiasm for NeurIPS, which is not constrained by a particular domain but focused more on improving algorithm development.

Tweetables:

“How do we ensure that everybody in the community is able to get the best benefits from the current AI and can contribute in a meaningful way?” — Anima Anandkumar [2:44]

“Labs like NVIDIA Research are thinking about, ‘Okay, where do we go five to 10 years and beyond from here?’” — Anima Anandkumar [11:16]

“What I’m trying to do is bridge this gap [between academia and industry] so that my students and collaborators are getting the best of both worlds” — Anima Anandkumar [23:54]

You Might Also Like:

NVIDIA Research’s David Luebke on Intersection of Graphics, AI

There may be no better guest to talk AI and graphics than David Luebke, vice president of graphics research at NVIDIA. He co-founded NVIDIA Research in 2006, after eight years on the faculty of the University of Virginia.

NVIDIA’s Jonah Alben Talks AI

Imagine building an engine with 54 billion parts. Now imagine each piece is the size of a gnat’s eyelash. That gives an idea of the scale Jonah Alben works at. Alben is the co-leader of GPU engineering at NVIDIA.

Demystifying AI with NVIDIA’s Will Ramey

One of NVIDIA’s best explainers, Will Ramey, provides an introduction to the AI boom and the key concepts behind it. Ramey is the senior director and global head of developer programs at NVIDIA.

Tune in to the AI Podcast

Get the AI Podcast through iTunes, Google Podcasts, Google Play, Castbox, DoggCatcher, Overcast, PlayerFM, Pocket Casts, Podbay, PodBean, PodCruncher, PodKicker, Soundcloud, Spotify, Stitcher and TuneIn. If your favorite isn’t listed here, drop us a note.

Make the AI Podcast Better

Have a few minutes to spare? Fill out this listener survey. Your answers will help us make a better podcast.

Whether talking with banks, cell phone providers or insurance companies, customers often encounter AI-powered voice interfaces to direct their calls to the right department.

But these interfaces typically are limited to understanding certain keywords. Onvego, an Israel-based startup, is working to make these systems understand what you say, no matter how you say it.

Before starting Onvego, the company’s founders created a mobile speech apps platform to assist blind people. Now they’re creating pre-built AI models for such use cases as accepting or requesting payments, scheduling appointments or booking reservations.

Onvego is a member of NVIDIA Inception, a program that accelerates AI and data science startups with go-to-market support, expertise and technology.

The company’s AI enables enterprises to easily build their own conversational interfaces in 10 different languages, with more on the way. Its technology already powers Israel’s toll road payment systems, enabling drivers to pay with their voice.

“Say the customer said, ‘I want to pay my bill.’ The system has to understand what that means,” said Alon Buchnik, CTO and co-founder of Onvego. “Once it does, it sends that information back to the speech machine, where logic is applied.”

The system then walks the driver through the payment process. Onvego’s AI also powers two emergency road services providers in Israel, providing AI-powered answers to drivers in need.

“The speech machine understands exactly what the problem is,” said Buchnik. “It understands if it needs to send a tow truck or just a technician.”

In Search of Ubiquity

Road-related applications are just the tip of the iceberg for Onvego. The company envisions its technology being inside everything from coffee machines to elevators. Along those lines, it’s forged partnerships with GE, GM, Skoda, Amazon and numerous other companies.

For instance, Onvego’s AI is being incorporated into a line of elevators, enabling the manufacturer to provide a conversational voice interface for users.

With the COVID-19 pandemic raging around the globe, Buchnik believes the company’s no-touch technology, such as activating elevators by voice only, can deliver an added benefit by reducing transmission of the virus.

But Onvego’s most-ambitious undertaking may be its contact call center technology. The company has developed an application, powered by NVIDIA GPUs, that’s designed to do the work of an entire call center operation.

It runs as a cloud-based service as well as enterprise on-premises solution that would provide real-time natural language call center support for IoT devices at the network’s edge, even where there’s no internet connectivity.

GPUs at the Core

Buchnik said that while it would be possible for the Onvego call center application to answer 50 simultaneous calls without GPUs, “it would require a huge CPU” to do so. “For the GPU, it’s nothing,” he said.

Onvego also uses a CUDA decoder so developers can access decoding capabilities on the GPU.

Training of the company’s automatic speech recognition models occurs on NVIDIA GPU-powered instances from AWS or Azure, which Onvego acquired through NVIDIA Inception.

Aside from its efforts to expand the use of its technology, Onvego is focused on the standalone container for locations at the edge or completely independent from the network. They plan to run it on an NVIDIA Jetson Nano.

The idea of providing intelligent natural language interfaces to people wherever they’re needed is providing Buchnik and his team with all the motivation they need.

“This is our vision,” he said. “This is where we want to be.”

Buchnik credits the NVIDIA Inception program for providing the company access to the best AI experts, top technical resources and support, and access to a large marketing program with strong positioning in different market verticals.

By using the NVIDIA resources and platforms, Onvego is hoping to promote its intelligent voice solutions to markets and industries that it has not yet reached.

Bill Dally — one of the world’s foremost computer scientists and head of NVIDIA’s research efforts — will deliver the keynote address during GTC China, the latest event in the world’s premier conference series focused on AI, deep learning and high performance computing.

Registration is not required to view the keynote, which will take place on Dec. 14, at 6 p.m. Pacific time (Dec. 15, 10 a.m. China Standard time). GTC China is a free, online event, running Dec. 15-19.

Tens of thousands of attendees are expected to join the event, with thousands more tuning in to hear Dally speak on the latest innovations in AI, graphics, HPC, healthcare, edge computing and autonomous machines. He will also share new research in the areas of AI inference, silicon photonics, and GPU cluster acceleration.

In a career spanning nearly four decades, Dally has pioneered many of the fundamental technologies underlying today’s supercomputer and networking architectures. As head of NVIDIA Research, he leads a team of more than 200 around the globe who are inventing technologies for a wide variety of applications, including AI, HPC, graphics and networking.

Prior to joining NVIDIA as chief scientist and senior vice president of research in 2009, he chaired Stanford University’s computer science department.

Dally is a member of the National Academy of Engineering and a fellow of the American Academy of Arts & Sciences, the Institute of Electrical and Electronics Engineers and the Association for Computing Machinery. He’s written four textbooks, published more than 250 papers and holds over 120 patents, and has received the IEEE Seymour Cray Award, ACM Eckert-Mauchly Award and ACM Maurice Wilkes Award.

Following Dally’s keynote, four senior NVIDIA executives will describe how the company’s latest breakthroughs in AI, data science and healthcare are being adopted in China. The panel discussion will take place on Monday, Dec. 14, at 7:10 p.m. Pacific (Dec. 15 at 11:10 a.m. CST).

GTC China Highlights

GTC is the premier conference for developers to strengthen their skills on a wide range of technologies. It will include 220+ live and on-demand sessions and enable attendees to ask questions and interact with experts.

Many leading organizations will participate, including Alibaba, AWS, Baidu, ByteDance, China Telecom, Dell Technologies, Didi, Hewlett Packard Enterprise, Inspur, Kuaishou, Lenovo, Microsoft, Ping An, Tencent, Tsinghua University and Xiaomi.

Certified instructors will provide virtual training for hundreds of participants in the NVIDIA Deep Learning Institute. DLI seats are currently sold out.

NVIDIA Inception, an acceleration program for AI and data science startups, will host 12 leading Chinese startups in the NVIDIA Inception Startup Showcase. Attendees will have the opportunity to see presentations from the 12 CXOs, whose companies were selected by winning a vote among more than 40 participating startups.

For more details and to register for GTC China at no charge, visit www.nvidia.cn/gtc/.

AI is the largest technology force of our time, with the most potential to transform industries. It will bring new intelligence to healthcare, education, automotive, retail and finance, creating trillions of dollars in a new AI economy.

As businesses look ahead to 2021 priorities, now’s a great time to look back at where the world stands on global AI adoption.

Retailers like Walmart and Tesco are mining new AI opportunities for product forecasting, supply chain management, intelligent store installations and predicting consumer buying trends. Healthcare players in the age of COVID are trying to speed scientific research and vaccine development.

Meantime, educators are employing AI to train a data-savvy workforce. And legions of businesses are examining how AI can help them adapt to remote work and distance collaboration.

Yet mainstream adoption of AI continues to skew toward big tech companies, automotive and retail, which are attempting to scale across their organizations instead of investing in skunkwork projects, according to a 2019 McKinsey global survey of about 2,000 organizations.

We asked some of the top experts at NVIDIA where they see the next big things in AI happening as companies parse big data and look for new revenue opportunities. NVIDIA works with thousands of AI-focused startups, ISVs, hardware vendors and cloud companies, as well as companies and research organizations around the world. These broad collaborations offer a bird’s eye view into what’s happening and where.

Here’s what our executives had to say:

CLEMENT FARABET
Vice President, NVIDIA AI Infrastructure

AI as a Compiler: As AI training algorithms get faster, more robust and with richer tooling, AI will become equivalent to a compiler — developers will organize their datasets as code, and use AI to compile them into models.The end state of this is a large ecosystem of tooling/platforms (just like today’s tools for regular software) to enable more and more non-experts to “program” AIs. We’re partially there, but I think the end state will look very different than where we are today — think compilation in seconds to minutes instead of days of training. And we’ll have very efficient tools to organize data, like we do for code via git today.

AI as a Driver: AI will be assisting most vehicles to move around the physical world and continuously learning from their environments and co-pilots (human drivers) to improve, on their way to becoming fully independent drivers. The value of this is there today and will only grow larger. The end state is commoditized level 4 autonomous vehicles, relying on cheap enough sensor platforms.

BRYAN CATANZARO
Vice President, NVIDIA Applied Deep Learning Research

Conversational AI: Chabots might seem like so-last-decade when it comes to video games designed to take advantage of powerful PC graphics cards and CPUs in today’s computers. AI for some time has been used to generate responsive, adaptive or intelligent behaviors primarily in non-player characters. Conversational AI will take gameplay further by allowing real-time interaction via voice to flesh out character-driven approaches. When your in-game enemies start to talk and think like you, watch out.

Multimodal Synthesis: Can a virtual actor win an Academy Award? Advances in multimodal synthesis — the AI-driven art of creating speech and facial expressions from data — will be able to create characters that look and sound as real as a Meryl Streep or Dwayne Johnson.

Remote Work: AI solutions will make working from home easier and more reliable (and perhaps more pleasant) through better videoconferencing, audio quality and auto-transcription capabilities.

ANIMA ANANDKUMAR
Director of ML Research, NVIDIA, and Bren Professor at Caltech

Embodied AI: The mind and body will start coming together. We will see greater adaptivity and agility in our robots as we train them to do more complex and diverse tasks.The role of high fidelity simulations is critical here to overcome the dearth of real data.

AI4Science: AI will continue to get integrated into scientific applications at scale. Traditional solvers and pipelines will be ultimately completely replaced with AI to achieve as high as a 1000x increase in speed. This will require combining deep learning with domain-specific knowledge and constraints.

ALISON LOWNDES
Artificial Intelligence, NVIDIA Developer Relations

Democratized AI: The more people who have access to the dataset, and who are trained in how to mine it, the more innovations that will emerge. Nations will begin to solidify AI strategies, while universities and colleges will work in partnership with private industry to create more end-user mobile applications and scientific breakthroughs.

Simulation AI: “What does (insert AI persona here) think? The AI-based simulation increasingly will mimic human intelligence, with the ability to reason, problem solve and make decisions. You’ll see increased use here for both AI research and design and engineering.

AI for Earth Observation (AI4EO): It may be a small world after all, but there’s still a lot we don’t know about Mother Earth. A global AI framework would process satellite data in orbit and on the ground for rapid, if not real-time, actionable knowledge. It could create new monitoring solutions, especially for climate change, disaster response and biodiversity loss.

KIMBERLY POWELL
Vice President & General Manager, NVIDIA Healthcare

Federated Learning: The clinical community will increase their use of federated learning approaches to build robust AI models across various institutions, geographies, patient demographics and medical scanners. The sensitivity and selectivity of these models are outperforming AI models built at a single institution, even when there is copious data to train with. As an added bonus, researchers can collaborate on AI model creation without sharing confidential patient information. Federated learning is also beneficial for building AI models for areas where data is scarce, such as for pediatrics and rare diseases.

AI-Driven Drug Discovery: The COVID-19 pandemic has put a spotlight on drug discovery, which encompasses microscopic viewing of molecules and proteins, sorting through millions of chemical structures, in-silico methods for screening, protein-ligand interactions, genomic analysis, and assimilating data from structured and unstructured sources. Drug development typically takes over 10 years, however, in the wake of COVID, pharmaceutical companies, biotechs and researchers realize that acceleration of traditional methods is paramount. Newly created AI-powered discovery labs with GPU-accelerated instruments and AI models will expedite time to insight — creating a computing time machine.

Smart Hospitals: The need for smart hospitals has never been more urgent. Similar to the experience at home, smart speakers and smart cameras help automate and inform activities. The technology, when used in hospitals, will help scale the work of nurses on the front lines, increase operational efficiency and provide virtual patient monitoring to predict and prevent adverse patient events.

CHARLIE BOYLE
Vice President & General Manager, NVIDIA DGX Systems

Shadow AI: Managing AI across an organization will be a hot-button internal issue if data science teams implement their own AI platforms and infrastructure without IT involvement. Avoiding shadow AI requires a centralized enterprise IT approach to infrastructure, tools and workflow, which ultimately enables faster, more successful deployments of AI applications.

AI Center of Excellence: Companies have scrambled over the past 10 years to snap up highly paid data scientists, yet their productivity has been lower than expected because of a lack of supportive infrastructure. More organizations will speed the investment return on AI by building centralized, shared infrastructure at supercomputing scale. This will facilitate the grooming and scaling of data science talent, the sharing of best practices and accelerate the solving of complex AI problems.

Hybrid Infrastructure: The Internet of Things will lead decision-makers to adopt a mixed AI approach, using the public cloud (AWS, Azure, Oracle Cloud,Google Cloud) and private clouds (on-premises servers) to deliver applications faster (with lower latency, in industry parlance) to customers and partners while maintaining security by limiting the amount of sensitive data shared across networks. Hybrid approaches will also become more popular as governments adopt strict data protection laws governing the use of personal information.

KEVIN DEIERLING
Senior Vice President, NVIDIA Networking

Accelerating Change in the Data Center: Security and management will be offloaded from CPUs into GPUs, SmartNICs and programmable data processing units to deliver expanded application acceleration to all enterprise workloads and provide an extra layer of security. Virtualization and scalability will be faster, while CPUs will run apps faster and offer accelerated services.

AI as a Service: Companies that are reluctant to spend time and resources investing in AI, whether for financial reasons or otherwise, will begin turning to third-party providers for experimentation. AI platform companies and startups will become key partners by providing access to software, infrastructure and potential partners.

Transformational 5G: Companies will begin defining what “the edge” is. Autonomous driving is essentially a data center in the car, allowing the AI to make instantaneous decisions, while also being able to report back for training. You’ll see the same thing with robots in the warehouse and the workplace, where there will be inference learning at the edge and training at the core. Just like 4G spawned transformational change in transportation with Lyft and Uber, 5G will bring transformational deals and capabilities. It won’t happen all at once, but you’ll start to see the beginnings of companies seeking to take advantage of the confluence of AI, 5G and new computing platforms.

SANJA FIDLER
Director AI, NVIDIA and Professor Vector Institute for Artificial Intelligence

AI for 3D Content Creation: AI will revolutionize the content creation process, offering smart tools to reduce mundane work and to empower creativity. In particular, creating 3D content for architecture, gaming, films and VR/AR has been very laborious: games like Call of Duty take at least a year to make, even with hundreds of people involved and millions budgeted.

With AI, one will be able to build virtual cities by describing them in words, and see virtual characters come to life to converse and behave in desired ways without needing to hard code the behavior. Creating a 3D asset will  become as easy as snapping a photo, and modernizing and restyling old games will happen with the click of a button.

AI for Robotics Simulation: Testing robots in simulated environments is key for safety-critical applications such as self-driving cars or operating robots. Deep learning will bring simulation to the next level, by learning to mimic the world from data, both in terms of creating 3D environments, simulating diverse behaviors, simulating and re-simulating new or observed road scenarios, and simulating the sensors in ways that are closer to reality.

An Opportunity for Reinvention

To accomplish any or all of these tasks, organizations will have to move more quickly for internal alignment. For example, 72 percent of big AI adopters in the McKinsey survey say their companies’ AI strategy aligns with their corporate strategy, compared with 29 percent of respondents from other companies. Similarly, 65 percent of the high performers report having a clear data strategy that supports and enables AI, compared with 20 percent from other companies.

Even as the global pandemic creates uncertainty around the world, 2021 will be a time of reinvention as players large and small leverage AI to improve on their business models. More companies will operationalize AI as early results prove promising enough to commit more resources to their efforts.

What are the fundamental laws that govern our universe? How did the matter in the universe today get there? What exactly is dark matter?

The questions may be eternal, but no human scientist has an eternity to answer them.

Now, thanks to NVIDIA technology and cutting-edge AI, the more than 1,000 collaborators from 26 countries working on the Belle II particle physics experiment are able to learn more about these big questions, faster.

The Belle II detector, based just north of Tokyo, reproduces the particles created during the early universe by smashing high-energy electrons and anti-electrons together.

These collisions generate a serious amount of data. Researchers will make high-precision recordings of hundreds of billions of collisions over the experiment’s lifetime. Sifting through all this data, without sacrificing the detailed information needed for high-precision measurements, is a daunting task.

To reconstruct the way individual particles, detected at Belle II, decayed from larger groups of particles, researchers turned to AI, says James Kahn from the Karlsruhe Institute of Technology, or KIT, a Belle II researcher and AI consultant with Helmholtz AI, a German public research platform for applied AI.

“Given the successes of AI and its ability to learn by example on large volumes of data, this is the perfect place to apply it,” Kahn said.

And to accelerate that AI, they’re using the NVIDIA Ampere architecture’s multi-instance GPU technology, built into the NVIDIA A100 GPU.

Physics Meets the A100

Kahn’s team was able to get early access to the “fresh out of the oven” NVIDIA DGX A100, a compact system packing 5 petaflops of AI computing power.

It’s among the first in Europe, and the first connected via InfiniBand high-speed interconnect technology. It was installed at KIT thanks to the high-performance computing operations team at the Steinbuch Center for Computing.

This close connection among the AI consultant team, international scientists and the HPC operations team will be a benefit for future research.

“We are really happy to see that only a few hours after we had the DGX A100 up and running, scientific analyses were already being performed,” said Jennifer Buchmüller, HPC core facility leader at KIT.

There’s more to come: HoreKa, the next supercomputer at KIT, will be equipped with more than 740 NVIDIA A100 GPUs.

A New View on Particle Decays

All of this helps Kahn and his team accelerate a new approach developed at KIT in collaboration with researchers from the nearby University of Strasbourg.

By designing a new representation of particle decays, or how unstable subatomic particles fall apart, Kahn’s team has been able to use a specialized neural network, known as a graph neural network, to automate the reconstruction of the particle decays from the individual particles detected by Belle II.

“We realized we could re-express particle decays in terms of the detected particles’ relations alone,” said Kahn. “This was the key ingredient to enable a full, end-to-end AI solution.”

The team has already demonstrated this technique’s success on a selection of specially designed simulations of particle decays, and recently scaled up to simulations of the interactions occurring at Belle II.

Scaling up, however, required resources that could handle both the volume of data and the large neural networks trained on it.

To do so they split up the GPUs using the multi-instance GPU technology — which allows a single GPU to perform multiple tasks simultaneously — to perform a spread-and-search of the network hyperparameters.

“Architecture searches which took days could now be completed in a matter of hours,” Kahn said.

The result: more time for more science, and for more of those eternal questions to be asked, and answered.

NVIDIA Research’s latest AI model is a prodigy among generative adversarial networks. Using a fraction of the study material needed by a typical GAN, it can learn skills as complex as emulating renowned painters and recreating images of cancer tissue.

By applying a breakthrough neural network training technique to the popular NVIDIA StyleGAN2 model, NVIDIA researchers reimagined artwork based on fewer than 1,500 images from the Metropolitan Museum of Art. Using NVIDIA DGX systems to accelerate training, they generated new AI art inspired by the historical portraits.

The technique — called adaptive discriminator augmentation, or ADA — reduces the number of training images by 10-20x while still getting great results. The same method could someday have a significant impact in healthcare, for example by creating cancer histology images to help train other AI models.

“These results mean people can use GANs to tackle problems where vast quantities of data are too time-consuming or difficult to obtain,” said David Luebke, vice president of graphics research at NVIDIA. “I can’t wait to see what artists, medical experts and researchers use it for.”

The research paper behind this project is being presented this week at the annual Conference on Neural Information Processing Systems, known as NeurIPS. It’s one of a record 28 NVIDIA Research papers accepted to the prestigious conference.

This new method is the latest in a legacy of GAN innovation by NVIDIA researchers, who’ve developed groundbreaking GAN-based models for the AI painting app GauGAN, the game engine mimicker GameGAN, and the pet photo transformer GANimal. All are available on the NVIDIA AI Playground.

The Training Data Dilemma

Like most neural networks, GANs have long followed a basic principle: the more training data, the better the model. That’s because each GAN consists of two cooperating networks — a generator, which creates synthetic images, and a discriminator, which learns what realistic images should look like based on training data.

The discriminator coaches the generator, giving pixel-by-pixel feedback to help it improve the realism of its synthetic images. But with limited training data to learn from, a discriminator won’t be able to help the generator reach its full potential — like a rookie coach who’s experienced far fewer games than a seasoned expert.

It typically takes 50,000 to 100,000 training images to train a high-quality GAN. But in many cases, researchers simply don’t have tens or hundreds of thousands of sample images at their disposal.

With just a couple thousand images for training, many GANs would falter at producing realistic results. This problem, called overfitting, occurs when the discriminator simply memorizes the training images and fails to provide useful feedback to the generator.

In image classification tasks, researchers get around overfitting with data augmentation, a technique that expands smaller datasets using copies of existing images that are randomly distorted by processes like rotating, cropping or flipping — forcing the model to generalize better.

But previous attempts to apply augmentation to GAN training images resulted in a generator that learned to mimic those distortions, rather than creating believable synthetic images.

A GAN on a Mission

NVIDIA Research’s ADA method applies data augmentations adaptively, meaning the amount of data augmentation is adjusted at different points in the training process to avoid overfitting. This enables models like StyleGAN2 to achieve equally amazing results using an order of magnitude fewer training images.

As a result, researchers can apply GANs to previously impractical applications where examples are too scarce, too hard to obtain or too time-consuming to gather into a large dataset.

Different editions of StyleGAN have been used by artists to create stunning exhibits and produce a new manga based on the style of legendary illustrator Osamu Tezuka. It’s even been adopted by Adobe to power Photoshop’s new AI tool, Neural Filters.

With less training data required to get started, StyleGAN2 with ADA could be applied to rare art, such as the work by Paris-based AI art collective Obvious on African Kota masks.

Another promising application lies in healthcare, where medical images of rare diseases can be few and far between because most tests come back normal. Amassing a useful dataset of abnormal pathology slides would require many hours of painstaking labeling by medical experts.

Synthetic images created with a GAN using ADA could fill that gap, generating training data for another AI model that helps pathologists or radiologists spot rare conditions on pathology images or MRI studies. An added bonus: With AI-generated data, there are no patient data or privacy concerns, making it easier for healthcare institutions to share datasets.

NVIDIA Research at NeurIPS

The NVIDIA Research team consists of more than 200 scientists around the globe, focusing on areas including AI, computer vision, self-driving cars, robotics and graphics. Over two dozen papers authored by NVIDIA researchers will be highlighted at NeurIPS, the year’s largest AI research conference, taking place virtually from Dec. 6-12.

Check out the full lineup of NVIDIA Research papers at NeurIPS.

Main images generated by StyleGAN2 with ADA, trained on a dataset of fewer than 1,500 images from the Metropolitan Museum of Art Collection API.

Academic researchers are developing AI to solve challenging problems with everything from agricultural robotics to autonomous flying machines.

To help AI research like this make the leap from academia to commercial or government deployment, NVIDIA today announced the Applied Research Accelerator Program. The program supports applied research on NVIDIA platforms for GPU-accelerated application deployments.

The program will initially focus on robotics and autonomous machines. Worldwide spending on robotics systems and drones is forecast to reach $241 billion by 2023, an 88 percent increase from the $128.7 billion in spending expected for 2020, according to IDC. The program will also extend to other domains such as Data Science, NLP, Speech and Conversational AI in the months ahead.

The new program will support researchers and the organizations they work with in rolling out the next generation of applications developed on NVIDIA AI platforms, including the Jetson developer kits and SDKs like DeepStream and Isaac.

Researchers working with sponsoring organizations will also gain support from NVIDIA through technical guidance, hardware grants, funding, grant application support, AI training programs, not to mention networking and marketing opportunities.

NVIDIA is now accepting applications to the program from researchers working to apply robotics and AI for automation in collaboration with enterprises seeking to deploy new technologies in the market.

Accelerating and Deploying AI Research

The NVIDIA Applied Research Accelerator Program’s first group of participants have already demonstrated AI capabilities meriting further development for agriculture, logistics and healthcare.

• The University of Florida is developing AI applications for smart sprayers used in agriculture, and working with Chemical Containers Inc. to deploy AI on machines running NVIDIA Jetson to reduce the amount of plant protection products applied to tree crops.
• The Institute for Factory Automation and Production Systems at Friedrich-Alexander-University Erlangen-Nuremberg, based in Germany, is working with materials handling company KION and the intralogistics research association IFL to design drones for warehouse autonomy using NVIDIA Jetson.
• The Massachusetts Institute of Technology is developing AI applications for disinfecting surfaces with UV-C light using NVIDIA Jetson. It’s also working with Ava Robotics to deploy autonomous disinfection on robots to minimize human supervision and additional risk of exposure to COVID-19.

Applied Research Accelerator Program Benefits  

NVIDIA offers hardware grants along with funding in some cases for academic researchers who can demonstrate AI feasibility in practical applications. The program also provides letters of support for third-party grant applications submitted by researchers.

Members will also have access to technical guidance on using NVIDIA platforms, including Jetson, as well as Isaac and DeepStream.

Membership in the new program includes access to training courses via the Deep Learning Institute to help researchers master a wide range of AI technologies.

NVIDIA also offers researchers opportunities to present and network at the GPU Technology Conferences.

Interested researchers can apply today for the Applied Research Accelerator Program.

Swedish startup Einride has rejigged the big rig for highways around the world.

The autonomous truck maker launched the next generation of its cab-less autonomous truck, known as the Pod, with new, advanced functionality and pricing. The AI vehicles, which will be commercially available worldwide, will be powered by the latest in high-performance, energy-efficient compute — NVIDIA DRIVE AGX Orin.

These scalable self-driving haulers will begin to hit the road in 2023, with a variety of models available to customers around the world.

Autonomous trucks are always learning, taking in vast amounts of data to navigate the unpredictability of the real world, from highways to crowded ports. This rapid processing requires centralized, high-performance AI compute.

With the power of AI, these vehicles can easily rise to the demands of the trucking industry. The vehicles can operate 24 hours a day, improving delivery times. And, with increased efficiency, they can slash the annual cost of logistics in the U.S. by 45 percent, according to experts at McKinsey.

Einride’s autonomous pods and trucks are built for every type of route. They can automate short, routine trips like the loading and unloading of containers on cargo ships and managing port operations, as well as autonomously drive on the highway, dramatically streamlining the shipping and logistics.

A New Pod Joins the Squad

The latest Einride Pod features a refined design that balances sleek features with the practical requirements of wide-scale production.

Its rounded edges give it an aerodynamic shape for greater efficiency and performance, without sacrificing cargo space. The Pod’s lighting system — which includes headlights, tail lights and indicators — provides a signature look while improving visibility for road users.

The cab-less truck comes in a range of variations, depending on use case. The AET 1 (Autonomous Electric Transport) model is purpose-built for closed facilities with dedicated routes — such as a port or loading bay. The AET 2 can handle fenced-in areas as well as short-distance public roads between destinations.

The AET 3 and AET 4 vehicles are designed for fully autonomous operation on backroads and highways, with speeds of up to 45 km per hour.

Einride is currently accepting reservations for AET 1 and AET 2, with others set to ship starting in 2022.

Trucking Ahead with Orin

The Einride Pod is able to achieve its scalability and autonomous functionality by leveraging the next generation in AI compute.

NVIDIA Orin is a system-on-a-chip born out of the data center, consisting of 17 billion transistors and the result of four years of R&D investment. It achieves 200 TOPS — nearly 7x the performance of the previous generation SoC Xavier — and is designed to handle the large number of applications and deep neural networks that run simultaneously in autonomous trucks, while achieving systematic safety standards such as ISO 26262 ASIL-D.

This massive compute capability ensures the Einride Pod is continuously learning, expanding the environments and situations in which it can operate autonomously.

These next-generation electric, self-driving freight transport vehicles built on NVIDIA DRIVE are primed to safely increase productivity, improve utilization, reduce emissions and decrease the world’s dependence on fossil fuels.