The Role of the Cloud in Advanced AI

Dave Vennergrund

Although AI has existed for as long as computing has been in existence, it is the cloud that is making AI the incredibly hot and highly in-demand technology that it is today. That’s why we’ve added this topic to our cloud series—we couldn’t miss an opportunity to talk about a technology capable of generating incredible benefits for agencies by helping them increase automation and drastically improve their operations. So, let’s discuss AI and look at why the government is so interested in it today.

Doing more with less

AI encompasses a broad set of capabilities that—in some manner—emulate human intelligence. Examples include facial recognition, speech recognition, robotics, language translation, and machine learning. If a machine is performing a function that used to require a human, you might call it AI. And all of the AI capabilities that I just listed are being used today in the federal government to help specific missions.

Today, AI tools are being used across organizations to automate many tasks, from automating call centers and help desks, to identifying risks, to preventing cyber attacks, to predicting network/system maintenance. Cyber analytics and machine learning models built with deep learning neural networks are being used to identify fraudulent transactions, assess risk for travelers, verify biometric identity, adjudicate benefits claims, manage disease, assess genomic expression data, screen airport luggage, and more.

AI even helps us build applications and analyze data more rapidly. Ultimately, AI is automating repetitive decision-making and large-scale pattern recognition at scales, speeds, and accuracies unachievable by humans. But most of that would not be happening so rapidly without the cloud.

What AI owes to the cloud

The cloud offers three critical ingredients needed for successful AI.

First, the cloud offers economical, elastic infrastructure to run machine learning algorithms. Cloud infrastructure includes virtual servers loaded with very large amounts of memory for machine learning algorithms; GPU servers ideal for parallel processing needed for deep learning; servers optimized with high network throughput for rapid application of models to real-time data; servers with high performance computing (HPC) features to support extensive modeling and simulation; and additional hardware configurations like Field Programmable Gate Arrays that are also optimized for parallel processing algorithms.

Second, the cloud offers an ever-increasing set of AI algorithms and tools for building AI solutions—and it offers them through services that cloud service providers are continuously updating and improving. This includes services for data ingestion, data transformation, data visualization, AI, machine learning and deep learning. In addition, the leading cloud providers offer data science virtual machine images pre-loaded with the powerful, open source data science and machine learning tools and frameworks used by experienced data scientists to innovate. Recently, cloud service providers have even expanded their tools to include AI services (machine learning as-a-service, predictive modeling as-a-service) and self-service tools that allow traditional developers and data analysts to innovate with AI methods without years of AI and Data Science experience.

And third, there’s data—the lifeblood for machine learning algorithms. Not all data comes from the cloud, but a lot is stored in the cloud—easily accessible in curated datasets ideal for training algorithms. Cloud providers host very large public data sets, many of them with labels ideal for training and comparing AI algorithms.

Without the cloud, the advanced AI solutions entering the marketplace today simply wouldn’t be possible. And the tools that cloud providers are offering are opening the door for more innovation across the federal government, and its industry partners.

What we’re doing with AI and the cloud

Companies are using the AI services and Data Science tools available in the cloud to develop pilots that illustrate the value of automation across the civil, defense, and intelligence areas.

We’ve used predictive modeling tools like the Amazon Machine Learning Service to predict which public water systems would violate safe drinking rules on chloroform in the wake of precipitation events. We’ve used Azure Machine Learning Studio to predict which automakers submitted incorrect gas mileage estimates. We’ve used Google Deep Learning APIs to recognize images. And we’ve used dozens of machine learning libraries available in the R and Python programming languages available in the AWS, Azure, and Google data science virtual machines to help the government with other mission critical tasks.

The future is AI, and the cloud eases the building and deploying of AI point solutions, as well as enterprise-class solutions. Industry partners like General Dynamics IT are excited about the potential of AI and the cloud, and are committed to using the AI tools available in the cloud to develop the solutions that revolutionize the way the government operates.