by Editorial Team
We often talk about how artificial intelligence (AI) enables agency transformation and is part of the push toward a digital government. According to Tony Paikeday, director of NVIDIA AI systems and data science platform, every one of us will be the beneficiary of AI innovation that makes our lives safer, healthier, and more efficient.
Research that has led to breakthroughs in healthcare, engineering, and other disciplines happens over decades, and AI has the ability to dramatically compress that timeline.
“AI and, more specifically, machine learning (ML) and deep learning (DL) are having such a transformative effect because of their ability to derive insights from data. Traditional programming approaches and tools stifle innovation. This takes time, requires analyzing vast amounts of data, and relies on specialized expertise and accumulated knowledge to deliver predictive capabilities that can benefit an organization,” he said. “But AI powered by GPU-accelerated computing, allows us to learn from data at an incredibly rapid pace and apply an algorithmic approach that can result in a timeline that would be impossible for humans to match by sifting through oceans of data and find meaningful patterns ourselves. In the world of AI, data is the new source code that powers innovation.”
AI holds a lot of promise in helping with the detection of cancer, for example. If you give an AI model enough data and allow it to “train” on that data, it will eventually become very good at looking for the patterns or anomalies, according to Paikeday. These patterns, which might tell an experienced clinician whether they’re seeing something that’s cancerous or benign, can be learned rapidly by an AI model that’s been fed a massive volume of images and associated outcomes. The model gets increasingly accurate as it trains on more data, improving its predictive accuracy with each iteration.
“The potential for AI to augment clinical workflow and enable healthcare providers to focus more of their time on challenging, more complex problems, has the potential to transform healthcare and deliver much better outcomes for patients.” Paikeday said this is just one example of the important frontiers where investments in AI innovation can impact every company and every citizen.
Another area where AI and, more specifically, machine learning are extremely valuable is in anomaly detection used by cybersecurity applications. These threat detection models can be used by analysts, alerting them to suspicious patterns found in data which might otherwise be missed. Not only can cyber teams improve accuracy and shorten their timelines, they are able to detect previously unseen attacks.
Another area where AI is gaining traction and proving incredibly value is in disaster relief. “The intersection of time and satellite information is essential to getting personnel and resources to those who need it most. Helping teams on the ground understand whether a street is still accessible and how much damage a building sustained can help agencies find survivors and bring relief quicker.” Deep learning models that are trained on images from environmental satellites, combined with aerial images, can also detect patterns of violence in conflict zones, helping those trying to intervene and investigate human rights abuses.
Many organizations are struggling with aging equipment and infrastructure that needs to be inspected for dangerous defects that can pose a safety or environmental hazard to both personnel and citizens. Whether that’s equipment or civil infrastructure, AI-powered robotic inspection technologies can reach places inaccessible to humans, and more efficiently find anomalies that might escape a time-based inspection window helping keep people out of harm’s way while saving taxpayers money.
Additionally, Paikeday points out that organizations must have infrastructure that can handle the size and speed needed to move away from disparate silos of AI innovation spread across an agency to create a cohesive infrastructure strategy. This means an organization must have massive computational power to successfully implement and leverage AI, ML, and DL. Paikeday also said they should have access to data science expertise and data sets on which models can be trained.
“For a few years, we’ve seen organizations build remarkable things with the power of accelerated computing and AI,” Paikeday shared. “A lot of this work was done by individual teams that had a mandate, but maybe lacked the IT infrastructure to support them. As a result, we’ve seen a kind of ‘shadow IT’ or ‘shadow AI’ springing up,” he said.
Organizations must have infrastructure that can handle the size and speed needed to move away from silos of AI innovation that are sprawling across an agency to an actual infrastructure strategy. That’s because when a team goes it alone and builds something from scratch, the systems integration, software engineering, and troubleshooting effort layered on top of rapidly churning open source code means a lot of time lost on activity that has nothing to do with data science.
“These are like silos of innovation that sprawl across an organization or agency, and many of them unaware of what others are doing. This is where government IT leaders can and should have a role – they have the opportunity to get ahead of this sprawl and develop an infrastructure strategy that can enable an organization’s AI success – at scale,” he said.
As a result, and because most teams just want results without having to take on the role of stack developer and systems integrator, agencies are looking to full stack options.
“ONTAP AI built on NVIDIA DGX and NetApp ONTAP has proven to be a faster way to deploy an IT-approved, validated infrastructure solution built on GPU-accelerated computing, that’s ready to go and can scale effortlessly,” he explained. “We’ve brought together an ecosystem of technology and services partners who make the whole effort turnkey for government agencies that want to get ahead in the race to innovate with AI.
“Building this kind of consolidated, shared infrastructure for AI delivers multiple benefits in terms of people, process and technology. Now agencies can build communities of experts who can share their knowledge, accessing a common platform, and create talent development pipelines to address the skills gap. It also results in standardization of process through best practice sharing, for going from AI ideation to proof-of-concept to production,” he said. “From a technology standpoint, centralized infrastructure allows government IT to drive down cost-per-experiment while eliminating stranded investments that ultimately can’t scale. In 2019, we’re already seeing government organizations take control of the AI agenda, with an AI infrastructure strategy built on ONTAP AI.”