LinkedIn·Friday, 14 August 2026·11d ago

Three talks, three completely different sides of AI. Yesterday, at our brown bag session, three speakers reminded me how wide the field of…

ML6

24,333 followers

Three talks, three completely different sides of AI. Yesterday, at our brown bag session, three speakers reminded me how wide the field of AI actually is. Arne Vandendorpe (ML6) talked about the work we are doing on our humanoid robot. The team swapped in new, heavier five fingered robotic hands to allow dextrous manipulation. Simple change, big consequence. Moving that extra weight around shifted the robot’s weight distribution significantly, so the locomotion policy had to be trained from scratch in simulation before it worked on the real hardware again. The talk concluded on the topic of loco-manipulation and whole-body-control with the robot showing off its newly acquired martial arts skills. Filip Nowak, from the nonprofit North Star, is building an AI character that speaks West Flemish for a theater play. There is no ready made dataset for the dialect, so the team is recording native speakers themselves, aiming for about five hours of audio as a baseline. It is a small number if you compare it to how much data most AI models train on, but it is enough to start preserving a way of speaking that is slowly disappearing. Violetta Nguyen (ML6) closed with computer use agents: AI that clicks and types the same way a person would, useful to automate User Acceptance Testing. Running one costs around 3 cents per step, up to a few dollars for a longer task, and the agents even have a way to notice when they are stuck in a loop and try something else. Different problems, same pattern. AI is showing up in the parts of work that are physical, unrecorded, or just plain old. Not the clean, well documented use cases, the messy ones. Stay tuned and don't miss our next events. Subscribe to our newsletter to keep up to date: https://lnkd.in/eMjxZ289 #BrownBagSession #robotics #AIAgents