Gecko Robotics co-founder Troy Demmer testified before the House Committee on Homeland Security Wednesday as part of a hearing on how the government can better utilize artificial intelligence.
"The way America collects data on infrastructure today hasn't fundamentally changed in 50 years," Demmer said. "Despite advances in data collection technologies like robots, drones, sensors and smart probes that can detect corrosion through layers of concrete we are still largely gathering data manually, even on our most critical infrastructure like dams, pipelines, bridges, power plants and even military assets."
Demmer, who currently serves as CPO of the North Side-based robotics company, shared statistics comparing how Gecko's robots gather data versus the aforementioned manual means. The manual process used by an unnamed "major national security partner" of Gecko's created approximately 3,000 data points per asset, while using Gecko's robots on the same asset created over 8 million.
"AI models are only as good as the inputs they are trained on," Demmer said. "Without better data even the most sophisticated AI models will be at best ineffective and at worst harmful."
Security was discussed extensively at the panel. Other witnesses included Michael Sikorsky, CTO of cyber security firm Unit 42, Ajay Amlani, president of the America side of iProov, an online authenticator and Jake Laperruque, deputy director of security and surveillance at the Center for Democracy and Technology, a DC based nonprofit.
"We rushed as innovators to get the internet out and we didn't think about security and look at the spot we're in," Sikorski said. "I think that's why it is really important that we build AI and have it secure by design... everyone is rushing to get this technology into their products and out to consumers but we need to think about it as we build it."