4 Lessons To Learn From The NVIDIA GTC 20229 min readReading Time: 6 minutes
The NVIDIA GTC (GPU Technology Conference) is an annual global AI conference that brings together developers, engineers, researchers, inventors, and IT professionals to celebrate the innovation in AI and Deep Learning, Computer Graphics, Data Science, Machine Learning, and Autonomous Machines. Every year, the much-awaited event attracts 1000s of attendees, notable speakers, and influential thought experts from across the world who come together to discuss the advancements in computing that are transforming every industry, cutting-edge innovation across hardware and software, breakthroughs in emerging technologies, the evaluation of generalizable intelligence and so much more. We recently attended the NVIDIA GTC 2022 to understand the scope for AI disruption across industries in the coming years.
Here’s what we learned:
1. AI can fix processes and make them more efficient
“AI is about taking data and converting that into intelligence that can be used to improve the business.” Manuvir Das, Head of Enterprise Computing, NVIDIA at the NVIDIA GTC 2022.
Today AI has become ubiquitous across functions in every business. Das says, “No matter what industry you’re in, every organization has business functions and processes that are better off with AI.”
For instance, with AI-powered workflow automation, businesses can accomplish a wide range of tasks across functions. This allows them to create exponential value from their resources.
Studies show that employees spend 69 days a year on mundane, repetitive tasks. By delegating repetitive tasks to AI, employees and businesses have the time to provide meaningful services to their customers. Additionally, automation also enables the simplification and standardization of processes. This, in turn, enables greater productivity and smoother workflows that scale effortlessly. With AI, businesses can rapidly accelerate their go-to-market timelines. And consequently, accelerate revenue generation.
So how can teams integrate AI into their processes and unlock success? Das recommends thinking about it in terms of the following:
- What’s the right hardware, infrastructure, and where do I find it?
- How do I manage this infrastructure?
- What are all the software tools that I can use?
2. It is critical to understand customers at a granular level
“Customers are becoming increasingly omnichannel and expect a deeply personalized experience.” Kannan Achan, VP – Personalization and Recommendations, Walmart Tech at the NVIDIA GTC 2022.
Consumers’ experiences with digital-native companies such as Amazon, Apple, and Google have led them to expect best-in-class digital experiences and a seamless journey that is focused on meeting their needs. With AI, businesses can build in-depth user-profiles and create a 360° view of their customers. Teams can understand individual user preferences at a granular level, identify affinities and understand long-term usage behavior. This allows them to curate compelling experiences that are most relevant to their users.
That’s not all.
“Within a single session, a customer could have multiple micro intents. This could come off as noise but the devil is in the details. Businesses need to be able to untangle the various intents in a particular session and cater to them,” says Kannan.
Customers’ preferences evolve over time. Thus it is essential for businesses to ensure that they capture real-time intent — from session-specific behavioral and transactional signals — to give them what they’re really looking for.
Here’s how retail giant – Walmart personalizes the repurchase journey:
What should organizations consider while deploying personalization at scale? Here’s what Achan recommends:
- Consumer preferences, seasonality, business plans, and inventory are continuously evolving. It’s important to be considerate of this and account for changes during modeling.
- When delivering deeply personalized experiences, precision is key.
- Customers prefer to carry out certain actions on certain channels. Thus, it is important to encode the channel and stitch sessions meaningfully.
3. Data is an organization’s biggest asset – fix it now
80% of organizations are waking up to the fact that at least 80% of their content is unstructured. Bad data slows down your organization and pipelines. Businesses collect over 75,000 data points for a single customer. Up to 30% of this data becomes inaccurate every year. As a result, data science teams spend hours on time-consuming legacy processes and manually executing mundane tasks. This is time that could be better spent on extracting clear insights that could determine critical business decisions. According to a survey by Run.ai — 61% of enterprises believe data is the #1 challenge that lies in the way of AI innovation.
The answer to this? AI.
“If your work includes dealing with any kind of textual information, you should care about NLP. With NLP, businesses can structure their data, turn pain into delight and improve productivity,” said Kari Briski, VP of Software Product Management for AI/HPC at the NVIDIA GTC 2022.
With AI, enterprises can not only better manage data but transform it into a strategic asset as well. Teams can extract and organize their data to build structured datasets that match their business goals with minimal effort.
4. The key to successful AI deployments is a scalable infrastructure
While enterprise use of AI has grown by 270% over the past several years, many still struggle to make their AI initiatives successful. Almost half never make it to production, according to recent Gartner research.
For many, the complexity of AI infrastructure is one of the most frequently named challenges. “After the first 1-2 implementations, AI starts to proliferate and the complexity and need for a scalable platform really increases,” says Anne Hecht, Senior Director of Product Marketing, NVIDIA at the NVIDIA GTC 2022.
“Companies are committing to AI, investing millions into infrastructure and likely millions more in highly-trained staff. But if most AI models never make it into production, the promise of AI is not being realized,” says Omri Geller, CEO, and co-founder of Run: AI.
Today, infrastructure challenges are causing resources to sit idle, data scientists are still requesting manual access to GPUs and the transition to the cloud is an ongoing process. However, it’s important for teams to know that restricting their deployments to simply being on-prem or in the cloud, can be detrimental. In order to truly make the most of their AI deployments, organizations will have to figure out how to put the right work in the right place.
For instance, teams can match the model and its complexity to how close it needs to be to your customer and what the roundtrip time needs to look like. They can balance the roundtrip time vs the time needed to compute results and place the application in the right areas.
“Companies that handle these challenges the most effectively will bring models to market and win the AI race,” says Geller.
As AI becomes more mainstream, teams will need to look at how they make sense of AI, how they can accelerate their use of AI, specifically in the IT infrastructure world, how to make sense of it all, how to find a way to deploy it for all the users that need it while doing it in a way that fits with their infrastructure.
What do the experts recommend? Here’s what Charlie Boyle, VP and GM DGX Systems, NVIDIA has to say:
There are 5 areas that are critical for a successful AI project: Software, Networking, Hybrid Deployment, Management & Partners.
Understanding the AI Software your users may need is important not only to plan a successful AI project but the overall AI infrastructure. This allows you to serve the needs of your users now and also keep up with the rapid evolution cycle of AI software.
It is not enough to invest in state-of-the-art servers and GPUS, organizations should focus on networking them correctly so that both the organization & the user can access what they need when they need it.
Teams need to figure out what they want to deploy on cloud, on-prem, on edge, what are the parameters for when things need to be moved back and forth, the strategy that drives this, and the tools to enable that.
An AI system is just like any other system and is typically managed similarly to any Linux-based system. However, the usage pattern and the way people run jobs might be very different from traditional IT jobs, so teams need to be cognizant of the differentiating factors of an AI implementation and formulate a blueprint on how to manage that.
Organizations that are just getting started with AI should leverage the expertise of various partners to help them design and build systems, deploy and manage the whole lifecycle of AI systems.
At Blox.ai, we’re constantly speaking to businesses heads and decision-makers in organizations across industries about the scope of AI for their business and all the ways it can empower their teams to deliver meaningful outcomes. Learn how we foster successful AI transformations and make businesses AI-Native, one use case at a time.