Monthly Archives: July 2023

A watershed moment on Nov. 22, 2022, was mostly virtual, yet it shook the foundations of nearly every industry on the planet.

On that day, OpenAI released ChatGPT, the most advanced artificial intelligence chatbot ever developed. This set off demand for generative AI applications that help businesses become more efficient, from providing consumers with answers to their questions to accelerating the work of researchers as they seek scientific breakthroughs, and much, much more.

Businesses that previously dabbled in AI are now rushing to adopt and deploy the latest applications. Generative AI — the ability of algorithms to create new text, images, sounds, animations, 3D models and even computer code — is moving at warp speed, transforming the way people work and play.

By employing large language models (LLMs) to handle queries, the technology can dramatically reduce the time people devote to manual tasks like searching for and compiling information.

The stakes are high. AI could contribute more than $15 trillion to the global economy by 2030, according to PwC. And the impact of AI adoption could be greater than the inventions of the internet, mobile broadband and the smartphone — combined.

The engine driving generative AI is accelerated computing. It uses GPUs, DPUs and networking along with CPUs to accelerate applications across science, analytics, engineering, as well as consumer and enterprise use cases.

Early adopters across industries — from drug discovery, financial services, retail and telecommunications to energy, higher education and the public sector — are combining accelerated computing with generative AI to transform business operations, service offerings and productivity.

Generative AI for Drug Discovery

Today, radiologists use AI to detect abnormalities in medical images, doctors use it to scan electronic health records to uncover patient insights, and researchers use it to accelerate the discovery of novel drugs.

Traditional drug discovery is a resource-intensive process that can require the synthesis of over 5,000 chemical compounds and yields an average success rate of just 10%. And it takes more than a decade for most new drug candidates to reach the market.

Researchers are now using generative AI models to read a protein’s amino acid sequence and accurately predict the structure of target proteins in seconds, rather than weeks or months.

Using NVIDIA BioNeMo models, Amgen, a global leader in biotechnology, has slashed the time it takes to customize models for molecule screening and optimization from three months to just a few weeks. This type of trainable foundation model enables scientists to create variants for research into specific diseases, allowing them to develop target treatments for rare conditions.

Whether predicting protein structures or securely training algorithms on large real-world and synthetic datasets, generative AI and accelerated computing are opening new areas of research that can help mitigate the spread of disease, enable personalized medical treatments and boost patient survival rates.

Generative AI for Financial Services

According to a recent NVIDIA survey, the top AI use cases in the financial services industry are customer services and deep analytics, where natural language processing and LLMs are used to better respond to customer inquiries and uncover investment insights. Another common application is in recommender systems that power personalized banking experiences, marketing optimization and investment guidance.

Advanced AI applications have the potential to help the industry better prevent fraud and transform every aspect of banking, from portfolio planning and risk management to compliance and automation.

Eighty percent of business-relevant information is in an unstructured format — primarily text — which makes it a prime candidate for generative AI. Bloomberg News produces 5,000 stories a day related to the financial and investment community. These stories represent a vast trove of unstructured market data that can be used to make timely investment decisions.

NVIDIA, Deutsche Bank, Bloomberg and others are creating LLMs trained on domain-specific and proprietary data to power finance applications.

Financial Transformers, or “FinFormers,” can learn context and understand the meaning of unstructured financial data. They can power Q&A chatbots, summarize and translate financial texts, provide early warning signs of counterparty risk, quickly retrieve data and identify data-quality issues.

These generative AI tools rely on frameworks that can integrate proprietary data into model training and fine-tuning, integrate data curation to prevent bias and use guardrails to keep conversations finance-specific.

Expect fintech startups and large international banks to expand their use of LLMs and generative AI to develop sophisticated virtual assistants to serve internal and external stakeholders, create hyper-personalized customer content, automate document summarization to reduce manual work, and analyze terabytes of public and private data to generate investment insights.

Generative AI for Retail

With 60% of all shopping journeys starting online and consumers more connected and knowledgeable than ever, AI has become a vital tool to help retailers match shifting expectations and differentiate from a rising tide of competition.

Retailers are using AI to improve customer experiences, power dynamic pricing, create customer segmentation, design personalized recommendations and perform visual search.

Generative AI can support customers and employees at every step through the buyer journey.

With AI models trained on specific brand and product data, they can generate robust product descriptions that improve search engine optimization rankings and help shoppers find the exact product they’re looking for. For example, generative AI can use metatags containing product attributes to generate more comprehensive product descriptions that include various terms like “low sugar” or “gluten free.”

AI virtual assistants can check enterprise resource planning systems and generate customer service messages to inform shoppers about which items are available and when orders will ship, and even assist customers with order change requests.

Fashable, a member of NVIDIA Inception’s global network of technology startups, is using generative AI to create virtual clothing designs, eliminating the need for physical fabric during product development. With the models trained on both proprietary and market data, this reduces the environmental impact of fashion design and helps retailers design clothes according to current market trends and tastes.

Expect retailers to use AI to capture and retain customer attention, deliver superior shopping experiences, and drive revenue by matching shoppers with the right products at the right time.

Generative AI for Telecommunications

In an NVIDIA survey covering the telecommunications industry, 95% of respondents reported that they were engaged with AI, while two-thirds believed that AI would be important to their company’s future success.

Whether improving customer service, streamlining network operations and design, supporting field technicians or creating new monetization opportunities, generative AI has the potential to reinvent the telecom industry.

Telcos can train diagnostic AI models with proprietary data on network equipment and services, performance, ticket issues, site surveys and more. These models can accelerate troubleshooting of technical performance issues, recommend network designs, check network configurations for compliance, predict equipment failures, and identify and respond to security threats.

Generative AI applications on handheld devices can support field technicians by scanning equipment and generating virtual tutorials to guide them through repairs. Virtual guides can then be enhanced with augmented reality, enabling technicians to analyze equipment in a 3D immersive environment or call on a remote expert for support.

New revenue opportunities will also open for telcos. With large edge infrastructure and access to vast datasets, telcos around the world are now offering generative AI as a service to enterprise and government customers.

As generative AI advances, expect telecommunications providers to use the technology to optimize network performance, improve customer support, detect security intrusions and enhance maintenance operations.

Generative AI for Energy

In the energy industry, AI is powering predictive maintenance and asset optimization, smart grid management, renewable energy forecasting, grid security and more.

To meet growing data needs across aging infrastructure and new government compliance regulations, energy operators are looking to generative AI.

In the U.S., electric utility companies spend billions of dollars every year to inspect, maintain and upgrade power generation and transmission infrastructure.

Until recently, using vision AI to support inspection required algorithms to be trained on thousands of manually collected and tagged photos of grid assets, with training data constantly updated for new components. Now, generative AI can do the heavy lifting.

With a small set of image training data, algorithms can generate thousands of physically accurate images to train computer vision models that help field technicians identify grid equipment corrosion, breakage, obstructions and even detect wildfires. This type of proactive maintenance enhances grid reliability and resiliency by reducing downtime, while diminishing the need to dispatch teams to the field.

Generative AI can also reduce the need for manual research and analysis. According to McKinsey, employees spend up to 1.8 hours per day searching for information — nearly 20% of the work week. To increase productivity, energy companies can train LLMs on proprietary data, including meeting notes, SAP records, emails, field best practices and public data such as standard material data sheets.

With this type of knowledge repository connected to an AI chatbot, engineers and data scientists can get instant answers to highly technical questions. For example, a maintenance engineer troubleshooting pitch control issues on a turbine’s hydraulic system could ask a bot: “How should I adjust the hydraulic pressure or flow to rectify pitch control issues on a model turbine from company X?” A properly trained model would deliver specific instructions to the user, who wouldn’t have to look through a bulky manual to find answers.

With AI applications for new system design, customer service and automation, expect generative AI to enhance safety and energy efficiency, as well as reduce operational expenses in the energy industry.

Generative AI for Higher Education and Research

From intelligent tutoring systems to automated essay grading, AI has been employed in education for decades. As universities use AI to improve teacher and student experiences, they’re increasingly dedicating resources to build AI-focused research initiatives.

For example, researchers at the University of Florida have access to one of the world’s fastest supercomputers in academia. They’ve used it to develop GatorTron — a natural language processing model that enables computers to read and interpret medical language in clinical notes that are stored in electronic health records. With a model that understands medical context, AI developers can create numerous medical applications, such as speech-to-text apps that support doctors with automated medical charting.

In Europe, an industry-university collaboration involving the Technical University of Munich is demonstrating that LLMs trained on genomics data can generalize across a plethora of genomic tasks, unlike previous approaches that required specialized models. The genomics LLM is expected to help scientists understand the dynamics of how DNA is translated into RNA and proteins, unlocking new clinical applications that will benefit drug discovery and health.

To conduct this type of groundbreaking research and attract the most motivated students and qualified academic professionals, higher education institutes should consider a whole-university approach to pool budget, plan AI initiatives, and distribute AI resources and benefits across disciplines.

Generative AI for the Public Sector

Today, the biggest opportunity for AI in the public sector is helping public servants to perform their jobs more efficiently and save resources.

The U.S. federal government employs over 2 million civilian employees — two-thirds of whom work in professional and administrative jobs.

These administrative roles often involve time-consuming manual tasks, including drafting, editing and summarizing documents, updating databases, recording expenditures for auditing and compliance, and responding to citizen inquiries.

To control costs and bring greater efficiency to routine job functions, government agencies can use generative AI.

Generative AI’s ability to summarize documents has great potential to boost the productivity of policymakers and staffers, civil servants, procurement officers and contractors. Consider a 756-page report recently released by the National Security Commission on Artificial Intelligence. With reports and legislation often spanning hundreds of pages of dense academic or legal text, AI-powered summaries generated in seconds can quickly break down complex content into plain language, saving the human resources otherwise needed to complete the task.

AI virtual assistants and chatbots powered by LLMs can instantly deliver relevant information to people online, taking the burden off of overstretched staff who work phone banks at agencies like the Treasury Department, IRS and DMV.

With simple text inputs, AI content generation can help public servants create and distribute publications, email correspondence, reports, press releases and public service announcements.

The analytical capabilities of AI can also help process documents to speed the delivery of vital services provided by organizations like Medicare, Medicaid, Veterans Affairs, USPS and the State Department.

Generative AI could be a pivotal tool to help government bodies work within budget constraints, deliver government services more quickly and achieve positive public sentiment.

Generative AI – A Key Ingredient for Business Success 

Across every field, organizations are transforming employee productivity, improving products and delivering higher-quality services with generative AI.

To put generative AI into practice, businesses need expansive amounts of data, deep AI expertise and sufficient compute power to deploy and maintain models quickly. Enterprises can fast-track adoption with the NeMo generative AI framework, part of NVIDIA AI Enterprise software, running on DGX Cloud. NVIDIA’s pretrained foundation models offer a simplified approach to building and running customized generative AI solutions for unique business use cases.

Learn more about powerful generative AI tools to help your business increase productivity, automate tasks, and unlock new opportunities for employees and customers. 

Arise, members! Capcom’s legendary role-playing game Dragon’s Dogma: Dark Arisen joins the GeForce NOW library today.

The RPG and THQ Nordic’s Jagged Alliance 3 are newly supported on GeForce NOW, playable on nearly any device.

From Dusk Till Pawn

Become the Arisen and take up the challenge in Capcom’s critically acclaimed RPG. Set in a huge open world, Dragon’s Dogma: Dark Arisen brings players on an epic adventure filled with challenging battles and action.

But there’s no need to go it alone: Adventure with up to three Pawns. These customizable AI companions fight independently, demonstrating prowess and ability they’ve developed based on traits learned from each player.

Players can share their Pawns online and reap rewards of treasures, tips and strategy hints for taking down terrifying enemies. Pawns can also be borrowed when specific skills are needed to complete various challenging quests.

Revisit Gransys or experience Dragon’s Dogma for the first time. Members can play the real Steam version of this RPG classic with support for stunning visuals and high-resolution graphics, even on devices like Macs, mobile devices and smart TVs. Priority members can adventure at up to 1080p 60 frames per second, or upgrade to an Ultimate membership for gameplay at up to 4K 120 fps, longer streaming sessions and RTX ON for supported games.

Game On

Another week means new games.

THQ Nordic’s tactical RPG Jagged Alliance 3 joins the cloud this week. Chaos reigns when the elected president of Grand Chien — a nation of rich natural resources and deep political divides — goes missing and a paramilitary force known as “The Legion” seizes control of the countryside. Recruit from a large cast of unique mercenaries and make choices to impact the country’s fate.

Members can look forward to the following this week:

• Jagged Alliance 3 (New release on Steam, July 14)
• Dragon’s Dogma: Dark Arisen (Steam)

On top of that, in collaboration with EE, the U.K.’s biggest and fastest mobile network, GeForce NOW launched new cloud gaming bundles featuring Priority and Ultimate memberships. To celebrate, check out how streamer Leah ‘Leahviathan’ Alexandra showcased GeForce NOW in action at the U.K.’s highest-altitude gaming den on the slopes of Ben Nevis, 1,500 feet above sea level in the clouds of the Scottish Highlands.

What are you planning to play this weekend? Let us know on Twitter or in the comments below.

Who’s an NPC you’d want to be friends with IRL?

— NVIDIA GeForce NOW (@NVIDIAGFN) July 12, 2023

A crack NVIDIA team of five machine learning experts spread across four continents won all three tasks in a hotly contested, prestigious competition to build state-of-the-art recommendation systems.

The results reflect the group’s savvy applying the NVIDIA AI platform to real-world challenges for these engines of the digital economy. Recommenders serve up trillions of search results, ads, products, music and news stories to billions of people daily.

More than 450 teams of data scientists competed in the Amazon KDD Cup ‘23. The three-month challenge had its share of twists and turns and a nail-biter of a finish.

Shifting Into High Gear

In the first 10 weeks of the competition, the team had a comfortable lead. But in the final phase, organizers switched to new test datasets and other teams surged ahead.

The NVIDIANs shifted into high gear, working nights and weekends to catch up. They left a trail of round-the-clock Slack messages from team members living in cities from Berlin to Tokyo.

“We were working nonstop, it was pretty exciting,” said Chris Deotte, a team member in San Diego.

A Product by Any Other Name

The last of the three tasks was the hardest.

Participants had to predict which products users would buy based on data from their browsing sessions. But the training data didn’t include brand names of many possible choices.

“I knew from the beginning, this would be a very, very difficult test,” said Gilberto “Giba” Titericz.

KGMON to the Rescue

Based in Curitaba, Brazil, Titericz was one of four team members ranked as grandmasters in Kaggle competitions, the online Olympics of data science. They’re part of a team of machine learning ninjas who’ve won dozens of competitions. NVIDIA founder and CEO Jensen Huang calls them KGMON (Kaggle Grandmasters of NVIDIA), a playful takeoff on Pokémon.

In dozens of experiments, Titericz used large language models (LLMs) to build generative AIs to predict product names, but none worked.

In a creative flash, the team discovered a work-around. Predictions using their new hybrid ranking/classifier model were spot on.

Down to the Wire

In the last hours of the competition, the team raced to package all their models together for a few final submissions. They’d been running overnight experiments across as many as 40 computers.

Kazuki Onodera, a KGMON in Tokyo, was feeling jittery. “I really didn’t know if our actual scores would match what we were estimating,” he said.

Deotte, also a KGMON, remembered it as “something like 100 different models all working together to produce a single output … we submitted it to the leaderboard, and POW!”

The team inched ahead of its closest rival in the AI equivalent of a photo finish.

The Power of Transfer Learning

In another task, the team had to take lessons learned from large datasets in English, German and Japanese and apply them to meager datasets a tenth the size in French, Italian and Spanish. It’s the kind of real-world challenge many companies face as they expand their digital presence around the globe.

Jean-Francois Puget, a three-time Kaggle grandmaster based outside Paris, knew an effective approach to transfer learning. He used a pretrained multilingual model to encode product names, then fine-tuned the encodings.

“Using transfer learning improved the leaderboard scores enormously,” he said.

Blending Savvy and Smart Software

The KGMON efforts show the field known as recsys is sometimes more art than science, a practice that combines intuition and iteration.

It’s expertise that’s encoded into software products like NVIDIA Merlin, a framework to help users quickly build their own recommendation systems.

Benedikt Schifferer, a Berlin-based teammate who helps design Merlin, used the software to train transformer models that crushed the competition’s classic recsys task.

“Merlin provides great results right out of the box, and the flexible design lets me customize models for the specific challenge,” he said.

Riding the RAPIDS

Like his teammates, he also used RAPIDS, a set of open-source libraries for accelerating data science on GPUs.

For example, Deotte accessed code from NGC, NVIDIA’s hub for accelerated software. Called DASK XGBoost, the code helped spread a large, complex task across eight GPUs and their memory.

For his part, Titericz used a RAPIDS library called cuML to search through millions of product comparisons in seconds.

The team focused on session-based recommenders that don’t require data from multiple user visits. It’s a best practice these days when many users want to protect their privacy.

To learn more:

• Watch a GTC session on building session-based recommenders with Merlin.
• Take a recsys course from the NVIDIA Deep Learning Institute.
• Check out the next-item prediction workflow that’s part of NVIDIA AI Enterprise, a complete software suite with the security and support businesses require.
• And watch the video below.

Startup MosaicML is on a mission to help the AI community improve prediction accuracy, decrease costs and save time by providing tools for easy training and deployment of large AI models.

In this episode of NVIDIA’s AI Podcast, host Noah Kravitz speaks with MosaicML CEO and co-founder Naveen Rao about how the company aims to democratize access to large language models.

MosaicML, a member of NVIDIA’s Inception program, has identified two key barriers to widespread adoption: the difficulty of coordinating a large number of GPUs to train a model and the costs associated with this process.

MosaicML was in the news earlier this month when Databricks announced an agreement to acquire MosaicML for $1.3 billion.

Making training of models accessible is key for many companies that need control over model behavior, respect data privacy and iterate fast to develop new products based on AI.

You Might Also Like

Jules Anh Tuan Nguyen Explains How AI Lets Amputee Control Prosthetic Hand, Video Games

A postdoctoral researcher at the University of Minnesota discusses his efforts to allow amputees to control their prosthetic limb — right down to the finger motions — with their minds.

Overjet’s Ai Wardah Inam on Bringing AI to Dentistry

Overjet, a member of NVIDIA Inception, is moving fast to bring AI to dentists’ offices. Dr. Wardah Inam, CEO of the company, discusses using AI to improve patient care.

Immunai CTO and Co-Founder Luis Voloch on Using Deep Learning to Develop New Drugs

Luis Voloch, co-founder and chief technology officer of Immunai, talks about tackling the challenges of the immune system with a machine learning and data science mindset.

Subscribe to the AI Podcast

The AI Podcast is now available through Amazon Music. Additionally, you can also 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.

Make the AI Podcast better. Have a few minutes to spare? Fill out this listener survey.