Two Johns Hopkins electrical and computer engineering faculty members briefed congressional staff at the Hopkins Bloomberg Center on Aug. 4. Hosted by the Johns Hopkins Data Science and AI Institute, Engineering Lifelong Learning, and the Office of Federal Strategy, the session focused on advances and risks in AI systems that speak and see.
Rama Chellappa, Bloomberg Distinguished Professor in electrical and computer engineering and biomedical engineering, interim director of the Data Science and Artificial Intelligence Initiative, and chief scientist at the Johns Hopkins Institute for Assured Autonomy, described computer vision’s impact, from aiding people with low vision and identifying missing children to enabling safer autonomous systems. He emphasized real-world limits: data hunger, opaque decision-making, bias, privacy, and vulnerability to adversarial attacks, and posed the core test: “Will it work everywhere, and will it work for everyone?” Chellappa called for standards and robustness research so vision systems generalize across settings and populations.
Berrak Sisman, assistant professor of electrical and computer engineering and leader of the Speech & Machine Learning (SMILE) Lab in the Center for Language and Speech Processing, outlined how AI is transforming speech and language technologies. The same tools that enable natural, expressive voices for accessibility and health care can also be used to create audio deepfakes, making detection research as important as generation, she said. Sisman argued that emotional intelligence, tone, empathy, and safety guardrails are the next frontier as large language models become the backbone for how machines reason and converse across education, medicine, and cybersecurity.
Sisman and Chellappa offered a shared imperative: pairing innovation with assurance. For congressional staff shaping policy, that means investing in high-quality datasets, transparency, and detection tools so AI that speaks and sees does so responsibly, and delivers benefits broadly and safely.