Reflections on GEOINT 2017: Automation Technologies Give Analysts Super Powers

We are surrounded with reports of technology accelerating our lives. Search engines, like Google and Bing, help us find precise information instantly. Recommendation engines, like services from Netflix and Amazon, identify movies and products that are very similar to our tastes.

Intelligent agents like Alexa, Google Home and Cortana enable us to use voice commands to invoke the vast computing resources of Amazon, Google and Microsoft. Recent advances in machine learning provide capabilities to add textual annotations to objects detected in photos. It is clear these advanced technologies will automate many tasks we used to do manually.

The intelligence community is currently being inundated with data that seems to come from everywhere. It is nearly impossible for humans to manually sort through the petabytes of data created daily. This was a hot topic of conversation at GEOINT 2017. National Geospatial-Intelligence Agency Director Robert Cardillo asked for input on how to deal with this flood of data:

DigitalGlobe is in a great position to manage and make sense of the enormous amount of data coming in since we’ve been working on technology automation for years. Automation is a key element of the DigitalGlobe, an ecosystem of smart tools, geospatial intelligence and human expertise to find solutions to the world’s biggest challenges.

What does automation technology mean for our future?

In the next five years, automation will make the following positive impacts:

  • Illuminate new insights
  • Reduce information overload
  • Elevate analysts

 

Illuminating New Insights

In 1974, Ray Kurzweil created the company Kurzweil Computer Products, focused on the development of omni-font, an early stage optical character recognition (OCR) system. Using this machine learning technology, a computerized image (e.g., JPEG) of writing could be converted into a digital representation of the text. For decades, this type of technology has been used by the U.S. Postal System to “scan” and convert the handwritten addresses on 98 percent of letters into a digital version of the text to enable automated mail sorting and delivery. A variant of this OCR technology is used by law enforcement organizations to automatically scan photos of licenses plates into a digital representation of letters and numbers.

A visual representation of an automatic plate number recognition system

Today, more advanced machine learning technologies provide us with automated methods of generating insights from raw data:

  • Google’s open source machine learning library TensorFlow enables the creation of new software models that can automatically detect objects in images (e.g., find all of the street signs in a photo).
  • Collaborating with CosmiQ Works and NVIDIA, DigitalGlobe issued the SpaceNet Challenge to encourage the creation of machine learning models that can automatically detect building footprints and other objects in satellite imagery.
  • Using natural language-processing technologies like Python NLTK and SpaCy, computers can detect the positive or negative sentiment of a social media post.
  • DigitalGlobe’s Geospatial Big Data platform (GBDX) provides analysts with the ability to run automated object detection algorithms against petabytes of data with point and click simplicity.

These innovative technologies create new insights about our data, which open up additional opportunities during the process of analysis. Analysts, developers and data scientists working together in an open-source environment will expedite the evolution of automation technologies.

Reducing Information Overload

In the “Adventure of Copper Beaches,” the detective Sherlock Holmes famously says “Data! Data! Data! I cannot make bricks without clay!” In the 21st century, data is the critical fuel for analysis of all kinds. But, as the amount of available information continues to grow geometrically, it is easy to become overwhelmed with “too much data.”

New methods of automation can provide much-needed relief to overworked analysts. NGA Director Cardillo was recently quoted as saying that he would have to hire 8 million analysts to complete the analysis of all available satellite imagery of the Earth. If we have only 8,000 analysts available, could we ask them to work 1,000 times as hard to complete the task?

Of course not. But if we apply automation technology in a smart way, we can realize a massive increase in the impact analysts can make. The table below depicts some of the ways automation changes the game.

In a talk at GEOINT 2017, I provided some specific ideas on how intelligent automation will accelerate analysts’ workflow in the future.

Elevating Analysts

The Marvel comic “Iron Man” tells the fictional story of Tony Stark. Tony Stark is an eccentric engineer who creates an armored suit that provides the ability to fly and wield directed energy weapons. The Iron Man suit also has a resident Artificial Intelligence (AI) virtual assistant named Jarvis that provides critical co-piloting and information delivery functions.

Although fully human, Stark has superpowers he uses to support the band of heroes known as the Avengers. His advanced robotic suit and AI assistant enable him to perform tasks that he would have been unable to perform without technological enhancement.

The future of information analysis is not android robots or autonomous algorithms detached from humans. It is not merely faster CPUs and bigger RAM chips. The future of analysis is human analysts empowered and supported with automation technologies that allow them to address bigger challenges than ever before.

As Robert Cardillo said at GEOINT 2017, automation is a “transforming opportunity for the profession” of intelligence analysts. “Automation isn’t there to get rid of you; it’s there to elevate you. It’s about giving you a higher-level role to do the harder things.”