We are using machine learning to control situations where there are a lot of variables.
Chief Federal Technologist, Snowflake
AI Governance, Cloud and the Future
Data democratization means everyone has access to these data and tools. There are a ton of great tools out there that help folks who maybe aren’t data scientists, but are data science-y and make better decisions at work.
Chief Federal Technologist, Snowflake
The growth of artificial intelligence and machine learning over the last few years is unmistakable.
Agencies have realized the potential and real benefits of using the advanced technologies to improve decision making, analyze large databases and address mission challenges.
Bloomberg Government estimates agencies are on track to spend about $5 billion in total AI investments this year alone.
The initial use cases and successes are driving even more investments.
For instance, the National Science Foundation requested more than $850 million in AI-related research and development funding in fiscal 2021, that’s about a 70 percent increase above its fiscal 2020 request.
The Defense Department’s 2021 budget request would boost its Joint AI Center’s funding from $242 million to $290 million, and the Pentagon is asking for another $449 million for AI-related projects at the Defense Advanced Research Projects Agency (DARPA).
All of this spending doesn’t necessarily translate into long-term successes.
That is why the administration and GSA created an AI community of practice. It launched in November and now has more than 400 members from 26 agencies.
The goal is to help agencies find and apply best practices, and examples where agencies have successfully deployed AI for customer experience, human resources, advanced cybersecurity and business processes.
Nicholas Speece, the chief federal technologist at Snowflake, said the opportunity to use AI and machine learning to improve mission delivery across all industries, government and private sector, is substantial.
He said it’s not necessarily about replacing what agencies are doing today, but doing in a smarter way by using data to drive better, faster decisions.
“We are using machine learning to control situations where there are a lot of variables,” Speece said on the Innovation in Government show. “For example, think about it in the medical community if you are trying to figure out for a certain individual what conflicts of different prescriptions they might have. Those conflicts across any number of variables, demographics, health and weight could be important factors for that individual. Being able to look at all of those variables in a very quick way and serve out recommendations to pharmacists is just one example of using AI and ML.”
Speece said one of the reasons why agencies have picked up AI and ML so quickly is vendors have built out the underlying architecture to be the foundation that these tools can run on top of.
“Agencies and organizations are left to build that top 30% that is more specific to their mission and really get that value out for less cost and in a more catered and more effective way,” he said. “Agencies need to maximize the use of as much open source, open interfaces and open standards through commercial products.”
Still, the move to broad use of AI and ML tools isn’t as easy as flipping a switch.
Agencies still must address a host of challenges ranging from security to customization by not taking advantage of agile or dev/ops methodologies to the often-used phrase of culture change.
“Making data go from a cloud data warehouse into a cloud AI or ML platform it’s just a matter of wiring up the right things. We stress to folks to use the capabilities inherent to those products,” Speece said. “Don’t let this be a black box in your organization. Don’t try to personify it and name it. It’s a process and a tool. It informs how your agency does business so treat it as such. Be transparent about what it’s doing and how it makes its decisions, and take feedback from every level on how it can do it better.”
It comes back to establishing trust and transparency both internally and externally when using AI and ML.
Speece said creating trust and transparency comes back to the idea of governance of the tools and technology.
He said agencies should build into the tools security features such as roles-based access controls that are governed by the meta data tagging.
“This allows us to put richer roles in place for things like AI and ML because as we breakdown siloes we are exposing a lot of potentially sensitive data to an algorithm that will produce an output,” Speece said. “So making sure the inbound and outbound parts of those is safe is important not just for citizens, but on the commercial side for customers.”
One of the reasons agencies have this opportunity around applying AI and ML tools to their mission is because of the adoption of the cloud services over the last five or so years.
Speece said agencies couldn’t have gotten the same scale and computing power without the cloud.
“Data democratization means everyone has access to these data and tools,” he said. “There are a ton of great tools out there that help folks who maybe aren’t data scientists, but are data science-y and make better decisions at work.”
Snowflake’s cloud data platform equips agencies with a single, integrated platform that offers the only cloud-built data warehouse; instant, secure, and governed access to all their data; and a core architecture to enable many types of data workloads. Snowflake: Data without limits. Find out more at Snowflake.com/federal.