What the evolution of AI/ML means for government IT
Connecting state and local government leaders
The easy accessibility of cloud-native artificial intelligence and machine learning allows agencies to quickly design, test and deploy complex service solutions.
As recently as five years ago using artificial intelligence and machine learning to develop predictive algorithms was prohibitively expensive. Government agencies needed to provision large clusters of servers, implement Hadoop and data lakes and employ hundreds of data scientists. But now access to AI and ML has been democratized due to furious competition among the three largest cloud service providers: Microsoft Azure, Amazon Web Services and Google Cloud Platform.
The offerings from all three cloud giants have followed a similar pattern. First, the three CSPs all developed AI/ML application programming interfaces, which customers could connect to and offer services like voice transcription, translation and video recognition. An agency did not have to invest in the development of the algorithm, the cloud giants provided it. This allowed agencies to “dip their toes” into AI/ML projects and invest in the ones that showed the most promise in supporting the mission.
Just in the past two years, the big CSPs introduced cloud-native, automated machine learning. With Auto-ML, there is no pre-existing algorithm. Agencies create their own by bringing data pertaining to a mission challenge and then the CSP’s processes develop and validate the solution algorithm. Auto-ML puts the resources of the largest technology companies at the disposal of their customers, allowing hypotheses to be validated and service solutions designed much faster than in the past.
Every agency can benefit from Auto-ML. Any paper-based process can be digitized and the data used for greater efficiency. Recommendation engines can be built to make the workflow of any agency move faster and reduce errors. Brand new applications can be developed by agency IT teams, and/or intelligence can be embedded into existing government applications.
It’s become so much easier for agencies to validate a hypothesis using an existing algorithm from a CSP, much in the same way a private company would launch a minimally viable product, or MVP. Then as the business case is proved, agencies can graduate to a variety of cloud-native AI/ML services that allow for additional tuning of the algorithm. Cloud-native AI/ML is the driver for data-based decision-making for better government services and outcomes.
I’ve seen what the easy accessibility of AI/ML can do firsthand. When I served at the Transportation Security Administration, we were tasked with improving the credentialing process for enhanced access to critical areas. Using AI/ML we developed algorithms for assessing the severity of the applicants’ criminal backgrounds. The workload on human reviewers was reduced by 60%, and the algorithm forwarded the hardest decisions -- red cases -- to the most senior adjudicators.
At U.S. Citizen and Immigration Services, paper-based applications were an untapped source for electronic records processing. We digitized these records and used computer vision AI to detect document types, followed by document-specific natural language processing algorithms to automate the creation of digital information from the paper based applications. This digital information enabled electronic processing and workflows that greatly speeded the processing of applications. USCIS was also able to improve application scheduling efficiency by building AI into its interview scheduling process using historic no-show statistics.
The Navy spends billions annually to fight rust and corrosion on its ships. With a corrosion detection and analysis solution built with Google Cloud’s AI/ML platform and its native computer-vision capabilities, the Navy and its vendor partner successfully used drones to identify “corrosion of interest” from aerial images of vessels, with confidence scores of more than 90% with very few false positives. This complex integration between emerging software and hardware technologies was only possible due to the recent evolution and accessibility of AI/ML cloud services.
In the span of just a few years AI/ML has gone from fanciful science fiction to a solution only for the largest organizations, and finally to an easily accessible and economical tool for solving complex service delivery challenges. The pace of technological change and market competition between the big CSPs provide government IT leaders with a golden opportunity to improve their workflows.
Never before has such computational power been so available. Cloud-native AI/ML is creating a pathway for agencies to totally transform their service delivery.