Look Who it Was Random video

Why Robots Need to Make their own Experiences | Oliver Groth | TEDxDresden

1,191 views · Published 27 November 2018 · 12:55 · Indexed 21 October 2026

Channel: TEDx Talks · 2018 · Nonprofits & Activism

Watch on YouTube

Playing a game of Jenga is an easy and enjoyable task for almost everybody. The famous block stacking game is hugely popular among children and adults alike - and more recently also among robotics researchers.

When Oliver started his PhD research, he tried to teach a robotic arm to play Jenga. Despite the stunning advances in artificial intelligence over the past decade, robots are surprisingly still struggling with this simple task.

Oliver believes that human physical intuition is the key which enables our superior object manipulation skills and sets us apart in situations where the accurate, yet brittle computations of a machine might fail.

In his talk, Oliver explains how he tries to bridge this gap between human and artificial intelligence with machine learning and argues that machines cannot only be taught but also need to ‘make their own experiences’, in order to develop an intuition about the world they operate in. Oliver is working towards a Ph.D. in Computer Vision and Robotics at Oxford, supervised by Prof. Ingmar Posner and Prof. Andrea Vedaldi. He started his scientific journey at the Computer Vision Lab Dresden under the supervision of Prof. Carsten Rother and received the Carl-Zeiss-Award 2017 for his Diplom thesis on image captioning. He also spent one year as a visiting student at the heart of Silicon Valley working at the lab of Prof. Fei-Fei Li at Stanford. Oliver studied computer science at the Technical University of Dresden. 
Oliver’s research focuses on the training of autonomous agents with the help of computer simulations. He is fascinated by the human “superpower” of physical intuition which enables us to execute a wide range of complex manipulation tasks. Oliver believes that this intuition can be acquired by machines in a human-like fashion: His machine learning algorithms play with elementary geometric shapes in computer simulations to learn about geometry and physics. This talk was given at a TEDx event using the TED conference format but independently organized by a local community. Learn more at https://www.ted.com/tedx

More from this channel