The International Conference on Robotics and Automation in Vienna, where researchers from all over the world present their achievements every year, has an eye-catching tradition: the robot parade. The colourful procession that winds its way through the conference halls gets longer every year. And increasingly, the robots are taking on a human appearance.
At my house, the robot parade looks a bit less impressive for the time being. A robot vacuum and mop, a robot lawnmower and a window-cleaning robot are packed with sensors and smart software to do their jobs efficiently, but they still have their limitations. My robot vacuum can’t make its way up to the first floor on its own. If I forget to open a door? Then that room simply won’t be cleaned. And the window-cleaning robot doesn’t yet leap from one window to another Spider-Man-style. Each device excels at a single task, but don’t suddenly ask it to do something else.
“Try describing in a few lines how a robot should tidy a child’s bedroom”
That is the appeal of humanoid robots: a single machine performing a laundry list of tasks within our living environment. That environment is tailored to our bodies. A bigger problem is that traditional programming is insufficient for the variety of tasks and situations such a robot faces. Just try describing in a few lines how a robot should tidy a child’s bedroom. Physical AI is a promising technology for achieving greater flexibility, but the step towards truly human-like skills is a significant one.
Rodney Brooks, een roboticapionier aan het Massachusetts Institute of Technology, wijst erop dat robots niet alleen de finesse van onze handen missen, maar ook het gevoel voor kracht en aanraking. Dat ze dat zullen leren door video’s van menselijke handelingen te bekijken, noemt hij wishful thinking. Vooruitgang zal dan ook niet uit data alleen komen. Ze vraagt om een integratie van sensoren, actuatoren en AI: robots die kijken én voelen, die bewegen én hun kracht doseren, en die leren door te dóén, in de echte wereld. Zover zijn we nog niet.
Die scepsis remt de investeringsgolf allerminst. In 2025 haalden roboticabedrijven wereldwijd 40,7 miljard dollar op, goed voor 9 procent van alle durfkapitaal. Daarmee positioneert de sector zich naast AI als begeerd investeringsdomein. Met zoveel middelen kan het ontwikkeltempo alleen maar versnellen. Toch geloof ik niet in een ChatGPT-moment voor robotica. Humanoïde robots zullen niet van de ene dag op de andere transformeren tot universele assistenten. Het gaat stap voor stap. Dat betekent niet dat robots pas nuttig worden als ze perfect functioneren. Zelfs beperkte capaciteiten kunnen waarde creëren, thuis en in de industrie.
Rodney Brooks, a robotics pioneer at the Massachusetts Institute of Technology, points out that robots lack not only the dexterity of our hands, but also a sense of force and touch. He describes the idea that they will learn this by watching videos of human actions as wishful thinking. Progress will therefore not come from data alone. It requires the integration of sensors, actuators and AI: robots that see and feel, that move and modulate their force, and that learn by doing, in the real world. We are not there yet.
That scepticism is by no means holding back the wave of investment. In 2025, robotics companies worldwide raised $40.7 billion, accounting for 9 per cent of all venture capital. This positions the sector alongside AI as a coveted investment domain. With such substantial funding, the pace of development can only accelerate. Yet I do not believe in a ChatGPT moment for robotics. Humanoid robots will not transform into universal assistants overnight. It is a step-by-step process. That does not mean robots only become useful once they function perfectly. Even limited capabilities can create value, both at home and in industry.
At the end of the robot race, we may not find a world with a single type of all-rounder, but a colourful procession of humanoid robots, which are more versatile than today’s domestic robots, yet still have specialisations. Just as there is not just one type of car today, but a range of city cars, vans and sports cars.
Aan het einde van de robotrace wacht wellicht geen wereld met één type alleskunner, maar een bonte stoet van humanoïde robots, die veelzijdiger zijn dan de huidige huishoudrobots, maar toch specialisaties hebben. Net zoals er vandaag niet één auto bestaat, maar een waaier aan stads-, bestel- en sportwagens.