New Thoughts on Leveraging Cloud for Advanced AI Sponsored Content by Microsoft/NVIDIA
Artificial intelligence (AI) is becoming critical to many operations within companies. As the use and sophistication of AI grow, there is a new focus on the infrastructure requirements to produce results fast and efficiently. Many companies find that firing up cloud instances is not enough. Instead, companies must take a more strategic view of their cloud adoption to have the IT foundation required to fully use state-of-the-art AI. Doing so can deliver significant results across a wide variety of industries.
Specifically, AI requires an infrastructure that can meet the constantly increasing demands for high-performance compute and specialized needs of AI applications and workloads such as natural language processing, machine learning, and deep learning. To that point, a suitable infrastructure to support advanced AI must easily scale up and out.
Cloud infrastructure purpose-built for advanced AI
The recent Harvard Business Review whitepaper Analytic Services Rethinking Cloud Strategies for Advance AI noted the benefits of such an AI-first infrastructure and quantified how companies in different industries benefit from its use. According to the white paper, advanced AI applications must be supported by a cutting-edge infrastructure that provides the performance, flexibility, and scalability that these applications demand. But not just any cloud will do.
The diversity of cloud offerings gives organizations many options for their AI needs. That is particularly the case with generative AI. So, the question has shifted from whether to use the cloud for AI applications to which cloud provider best aligns with a company's strategic vision for AI. The selection will depend on the capabilities of the cloud vendor and the ecosystem of partners and vendors that is built around the vendor’s offerings.
These and other points were the subjects of a recent Harvard Business Review Analytic Services Webinar: Rethinking Cloud Strategies for Advanced AI. The webinar discussed cloud strategies to support advanced AI. (The webinar can be viewed on-demand here.) the speakers included IDC's Ritu Jyoti and Nidhi Chappell, General Manager of Azure HPC for AI, SAP and Confidential Computing, at Microsoft. Their talk examined how advanced AI creates unprecedented growth opportunities, the problems companies face related to cloud and AI technologies, and how to choose the right cloud platform for your AI goals.
Let's look at some examples from the leading companies in healthcare, automotive, fashion, and conservation featured in this Harvard Business Review Analytic Services whitepaper.
Innovative AI-led personalized cancer treatments
While radiology used to diagnose cancer has long embraced AI, Elekta, a Stockholm-based Swedish maker of precision radiation therapy solutions, focused on a related but more involved area: radiotherapy, which is used to treat cancer. Elekta found that many people worldwide do not have access to the needed personalized therapy, not because of a lack of technology but because of insufficient medical personnel from diverse disciplines that must collaborate to ensure the correct adjustments are made to treatment plans.
“We realized the tsunami of AI innovations that were happening in the computer vision and text recognition fields were eventually going to find their way into the medical field, as well,” said Rui Lopes, Director of New Technology Assessment at Elekta.
To address the problem, it embeds intelligence into devices to increase access to treatment for a larger swath of patients worldwide. “This provides not just personalization of care but democratization of a standard of care, allowing more advanced protocols to be deployed in regions of the world that lack the human capital to do so now,” said Lopes.
The models Elekta uses must easily scale. “You need to radically scale up the amount of data you use,” said Adam Moore, Director of Global Cloud Solutions for Elekta. “By training the models in the cloud, you can identify those problems earlier and build resilience into your compute infrastructure, so you avoid hardware failures.”
“We rely heavily on Azure cloud infrastructure. With Azure, we can create virtual machines on the fly with specific GPUs. If that’s not enough, we can cancel that virtual machine, create a new one, and then scale up as the project demands,” says Silvain Beriault, Lead Research Scientist at Elekta.
Developing a new generation of autonomous vehicles
Wayve, a London startup, is trying to bring deep learning and AI to the next generation of autonomous driving, something it calls AV2.0 (autonomous vehicles 2.0). In particular, the company wants to accelerate and scale autonomous vehicle development by using vision-based machine learning for rapid prototyping and quick iteration.
“Advanced AI, the latest and greatest, is absolutely pivotal to what we’re doing,” says Jamie Shotton, Chief Scientist at Wayve. “We have to train the algorithm on petabytes and potentially greater amounts of data that we’ve captured from our fleet of cars, which is a radically different approach to autonomous self-driving than anyone has done before.”
Moving to Azure infrastructure allows Wayve to rapidly improve its iteration speed and innovation rate for new autonomy features, which, in turn, helps cars drive better. Through its use of Azure Machine Learning, the company trains its AV2.0 models 90 percent faster compared to its previous data center environment.
“Using a managed platform gives us the ability to scale quickly and reliably. It also allows us to focus our efforts doing the research and solving problems around autonomous self-driving rather than building additional tools ourselves,” Shotton says.
Creating new fashions at the speed of the market
Fashion is one of the fastest-growing, most lucrative, and demanding industries, with high expectations of quick turnaround rates, creative designs, and a constant parade of new styles. As such, Portugal-based Fashable is trying to change the fashion industry with AI.
“In the near future, you will have a digital closet of clothing designs that you can ask a manufacturer to produce just for you,” says Orlando Ribas Fernandes, CEO and Co-founder of Fashable, ““We will use the metaverse to create physical goods that are exclusive to each person.”
Using Azure AI infrastructure, powered by NVIDIA GPUs for deep learning, Fashable built a generative AI application that can create dozens of original AI-generated clothing designs in minutes without the need for actual material. The algorithm ingests data from multiple sources to learn about trends, styles, and clothing types. Using social media to do A/B tests directly with customers lets designers gauge interest and forecast demand for their particular creations before going into production.
“We can share the collection with customers before they are produced, avoiding the problem of overstock,” says Orlando Ribas Fernandes, CEO and Co-founder of Fashable.
Protecting endangered species from wildlife crime
Wildlife Protection Solutions (WPS) use artificial intelligence on remote camera images for the conservation of endangered species and ecosystems. Its work helps recognize threats, classify species, and aids in anti-poaching to prevent human-wildlife conflict.
“Conservation is a huge challenge globally, and we’re not necessarily winning the war,” says Eric Schmidt, Executive Director of the Organization. To improve its odds in the fight, WPS is arming itself with AI models that search images from thousands of camera feeds, looking for humans and vehicles that may be engaged in suspicious activities or animals that may be encroaching on human populations.
For its AI needs, WPS uses Microsoft Azure's purpose-built AI infrastructure powered by NVIDIA GPUs. For example, the group's wpsWatch platform analyzes and monitors many inbound images from the remote cameras hosted in more than 100 sites across almost 20 counties. It is powered by Microsoft Azure VMs (virtual machines) with NVIDIA GPUs (graphics processing units) and was initially focused on the security and anti-poaching elements of the group’s mission.
A look to the future
These examples demonstrate the growing use of purpose-built infrastructure for AI. As companies increasingly adopt the latest AI technologies, like Generative AI, to transform their applications and derive business and economic value from AI, access to such an infrastructure will be critical for quickly getting value from AI economically.
Read the Harvard Business Review Analytic Services whitepaper “Rethinking Cloud Strategies for Advance AI“ and watch the webinar.
Visit the Microsoft and NVIDIA HPCwire Solution Channel for more articles and insights.