Effective artificial intelligence tools require vast quantities of clean, organized, comprehensive data, and that’s a problem for federal agencies.
Much of the government’s data is either unstructured or locked away in bureaucratic silos that prevent groups from accessing the information they need. Combine that with legal obstacles to information sharing and a shortage of tech expertise, and building worthwhile AI becomes a steep uphill battle.
But a service recently launched by IBM could help agencies work around those barriers and deploy basic AI tools wherever they house their data.
On Feb. 12, the company made its Watson AI system available across all cloud platforms, meaning agencies running Amazon, Microsoft or other cloud services can now use IBM’s tech in their enterprise environment. The service comes packaged in IBM Cloud Private for Data, an open source, containerized cloud environment that groups can stand up in different parts of their IT infrastructure.
Though the product doesn’t offer the full stack of Watson capabilities, it comes with a handful of basic applications, including a personal assistant, AI management service and platform for building machine-learning tools.
Previously, agencies that wanted to access Watson either needed to purchase IBM’s cloud or send their data off to a commercial server for analysis, according to Claude Yusti, who leads the AI practice in the company’s federal branch. The model presented a barrier for agencies with unorganized data, he said, and even if they could corral their information in one place, transferring it outside the enterprise is often prohibited under federal regulations.
“The question [became] how can we take the technology and make it more flexible and adaptable,” Yusti told Nextgov. “We can’t change the way the government organizes itself, policies don’t change overnight.”
In recent months, federal leaders have upped the pressure on agencies to support the country’s AI market and adopt the tech into their own ecosystems. The White House on Feb. 11 unveiled a national strategy for artificial intelligence that called on government to double down on AI research, create frameworks for building “robust” algorithms and open more federal data to AI researchers.
But as agencies work to advance the tech itself, Yusti said the government’s risk-averse culture is limiting its own adoption of AI.
“There’s very little incentive in the government to take risks with AI unless it’s been mandated as a mission essential capability,” he said. As such, defense and intelligence agencies are devoting significant resources to advancing the tech, but other organizations have been slow to get the ball rolling, according to Yusti.
Most civilian departments also lack the expertise to integrate the tech into what are often arcane, bureaucratic internal processes, he added. Generally speaking, non-defense agencies disproportionately struggle to bring young technologists, who are likely more versed in emerging tools like AI than their older counterparts, into the workforce.
A report published Thursday by IBM and the Partnership for Public Service said some 130,000 federal employees will see their jobs impacted by AI in the coming years. Government workers have said they largely feel unprepared for the changes on the horizon.
While an AI-as-a-service product like IBM’s won’t address all these underlying challenges, Yusti said, it could help agencies overcome some of the architectural and infrastructure barriers to rolling out the tech.
“Perfection isn’t going to happen,” he said. “We need to act, we need to learn from doing incremental steps. That will allow us to gain the experience and scale by having some small successes. I think the attitude is to execute some work now versus defer until it’s just right.”
source: NextGov
