Hao Dong, PI (external link)
Joining the Institute during the autumn semester 2026 | Multimodal learning, Computer vision, Robotics
Meet our new PIs: Hao Dong believes the next frontier is physical AI
Hao Dong joined ELLIS Institute Finland and Tampere University in August 2026.
“Artificial intelligence has transformed our digital lives, and the next step is physical AI,” says Hao Dong. The new assistant professor and principal investigator at ELLIS Institute Finland refers to robotics and autonomous systems that operate in the physical world, interacting with humans in tangible ways beyond chatbots. Dong’s research aims to develop physical AI systems that are robust, reliable, and adaptable. Such systems must be able to keep working when tasks change, respond to uncertainty and adapt to new environments while still performing safely and dependably.
Dong’s research ambitions took shape during his master’s degree at Aalto University. He was attracted to study in Finland by the strong concentration of research groups in robotics and machine learning. The collaborative culture and abundant research resources helped him establish a strong foundation in scientific research, gain confidence in conducting rigorous experiments, and successfully publish his findings. Following a PhD at ETH Zürich, Dong has returned to Finland, where he is now building his own research group, with an emphasis on students who have solid mathematical foundations and creativity to think outside the box.
Enabling the safe deployment and continuous adaptation of AI systems in changing environments remains a central challenge that Dong seeks to address, particularly in critical domains such as robotics, healthcare, and industrial monitoring. His current research focuses on long-horizon tasks, where autonomous systems must operate reliably over hours, days, or even years rather than minutes. Open questions include how AI can retain and use knowledge over extended periods, how it can detect and monitor for failures, and how it can adapt while staying in the bounds of reliability and safety.
Dong takes pride in his commitment to open science, having publicly released benchmark datasets and codebases that have garnered tens of thousands of downloads, generating real-world impact beyond individual papers. Alongside open research practices, Dong believes researchers need resilience and strong communication skills. “In research, 90 per cent of the time we have negative results, but we shouldn’t give up, just stay confident and make a plan for next steps,” says Dong. “Engaging with others and collaborating across areas of expertise worldwide is the best way to push the field forward.”
Contact:
