AI and machine learning: Looking beyond the hype
- By Erin Hawley
- Dec 15, 2017
In every federal agency, critical insights are hidden within the massive data sets collected over the years. But because of a shortage of data scientists in the federal government, extracting value from this data is time consuming, if it happens at all. Yet with advances in data science, artificial intelligence (AI) and machine learning, agencies now have access to advanced tools that will transform information analysis and agency operations.
From predicting terror threats to detecting tax fraud, a new class of enterprise-grade tools, called automated machine learning, have the power to transform the speed and accuracy of federal decision-making through predictive modeling. Technologies like these that enable AI are changing the way the federal government understands and makes decisions.
To use tools like automated machine learning to their full potential to accelerate and optimize data science in the federal government, it’s important to start by understanding the terms used and what they mean.
Data science — the art of analyzing data
Data science is a broad term, referring to the science and art of using data to solve problems. Rooted in statistics, this practice blends math, coding and domain knowledge to answer specific questions from a certain data set. Advances in computing power have transformed this from calculator-based statistical modeling into predictive algorithms that transform historical analysis into forecasts about future behaviors.
Even the very first U.S. census conducted in 1790, using quill and paper, collected about 20 megabytes of data. Today the Census Bureau has a backlog of nearly 400 billion data points, offering a wealth of insights into the demographics and behaviors of a constantly-evolving population. And while the quantity of data has grown over time, data scientists are in short supply, leaving a large gap in the amount of data and insights available, and the people needed to derive those insights from the data.
AI — filling the gaps in data science
AI overlaps with data science by giving machines the ability to interact as though a human engaged in the process. The power for the machine to copy intelligent human behavior by applying mathematical models to extrapolate information from data is at the core of AI. However, AI goes further in the sense that it can make decisions and take action through these machines -- whether on premise in a data center or in the cloud.
An early form of AI was developed by Alan Turing, which decoded encrypted messages from German forces during World War II. Today, AI is coming to fruition in the government with the availability and accessibility of big data, and increasingly affordable storage and processing power. AI has been applied successfully for intelligence and defense applications like the Defense Advanced Research Projects Agency's Cognitive Agent that Learns and Organizes program, which now serves as the backbone of Apple’s Siri and unmanned aircraft.
Machine Learning — the next step in AI's evolution
As AI has progressed, a new wave of innovation has developed machine learning. A blend of AI and data science, machine learning doesn’t just use algorithms to make decisions — it can learn from past data points, refining its methodology and getting smarter as time progresses and more data is collected and analyzed. The game-changer here is predictive intelligence — the ability to anticipate and prepare for future events based on sophisticated algorithms.
Think of what the power of prediction could achieve in the federal government. For example, machine learning will help the IRS find anomalies in filed tax forms, automatically flagging potentially fraudulent submissions for further review. It also could blend satellite and sensor information to fuel predictions from the National Weather Service.
And it’s coming at a critical time. Data scientists need AI and automated machine learning when they are interpreting results, rather than manually managing and processing information. What’s more, with automated machine learning, anyone in the federal government can operate as a data scientist, leveraging the predictive models and insights data can provide.
These advances are creating countless opportunities for federal agencies to make faster, more accurate decisions that can dramatically improve mission capability, efficiency and security -- while also adjusting for the unfilled data scientists positions. Agencies can learn from the successes businesses have had and point AI at sophisticated cyber adversaries, pinpointing insider threats before they strike, proactively identifying fraud or anticipating terrorist attacks.
Given the national security and economic threats we face and the growing shortage of data scientists, embracing this innovation is critical for staying one step ahead. With the power of AI and machine learning, this capability is now at our fingertips.
Erin Hawley is DataRobot's vice president of public sector.