Keynote: Frank Kirchner - Robotics and AI: Future Trends and Impact - RealAIGym Olympics, IJCAI 2023

In this talk

1. 0:30

   Turing’s recipe for learning robots

   Kirchner revisits Alan Turing’s proposal to study intelligence through robots with varied limbs, rich sensing, and real-world interaction. Robots would learn from experience and keep improving without a fixed endpoint, using feedback about whether actions worked.

2. 3:11

   AI draws on several fields

   Kirchner connects these ideas to Grey Walter’s small tortoise robot, which wandered, returned to a charging station, and resumed moving once its battery recovered. He describes AI as a meeting point for engineering and computer science with biology, neuroscience, linguistics, and psychology.

3. 8:08

   Complex bodies need learning and iteration

   Modern hardware makes Turing’s vision more practical, but adding degrees of freedom makes robots harder to control and software harder to manage. Learning methods can help with that complexity, while real-world trials reveal missing sensors or capabilities and send teams back to improve the system.

4. 11:01

   Build intelligence throughout the body

   Kirchner envisions robots learning tasks by observing people and eventually anticipating human intentions. He highlights compact actuators and pressure-sensing feet that calculate useful signals locally, reducing wiring and supporting balance. Distributed controllers also make complex behaviors easier to debug and can complement higher-level autonomy.

5. 17:11

   Use movement to support navigation

   A robot’s physical design can help it cross rough terrain, recover from awkward positions, and adapt its gait even before it has sophisticated planning. Kirchner also describes combining camera-based mapping with tactile and movement data to make maps more reliable in unfamiliar environments.