Amazon Data Scientists from George Mason teach AI workshop at GeoInt Symposium

Body

Two George Mason alums and one student led by Kristof Ladny provided professional insights into how generative AI can help solve large-scale problems. 

Pictured from left to right: Kristof Ladny (MS DAEN '21), Justin Downes (PHD candidate, Computational Social Science), Joe Pelczar (MS DAEN '20). Photo provided by Kristof Ladny.

The GeoInt Symposium is a nationally renowned gathering sponsored by the United States Geospatial Intelligence Foundation (USGIF) to provide a space for industry, academic, and government representatives to share community and tradecraft.  

Geospatial intelligence is critical for furthering innovation and national security interests. Ladny, (MS, DAEN ‘21), and Joe Pelczar (MS, DAEN ’20) are both now data scientists with Amazon Web Services’ Professional Services team. Justin Downes, a current PhD candidate in computational social science at George Mason as well as a senior applied scientist with AWS, rounded out the team. 

Their goal at the conference was to illustrate how generative AI can work at a large scale to rapidly solve complex problems. At the heart of their panel was research using a multimodal AI model, which can take in both images and text simultaneously. “We wrote some code to render images around maritime scenarios that could send text about statistics surrounding certain watercraft,” Ladny said. “The model combined the images and stats and came up with insights into activities the vessel could be performing.”  

How could this be useful, though? “One case would be detecting illegal fishing,” he said. “Different watercraft in different areas show distinct patterns. The AI could show that someone is doing a fishing-type pattern in a place where they shouldn’t be.” 

At the conference, Ladny saw a lot of interest in generative AI across industry and government. If they come up with the right prompts, they can automate things that a programmer might have spent days trying to code in the past. But, he said, “it can be hard to predict the best prompts to put in for the best results. The symposium showed us that, if you have a place where you can test a lot of the prompts, you can help it perform very well.” 

These types of gatherings can help train both AI and the people who will use it to do so both effectively and ethically. And George Mason is training the trainers.