This week Zdeněk Huspenina spoke at the AI in Practice 2026 conference in Zlín, organized by the Regional Chamber of Commerce of the Zlín Region. The topic of his talk sounded theoretical, but in practice it decides whether a robot will work in production or not: how physical AI differs from the AI we know from apps and chatbots.
No Try Again
When AI controls the physical world, it is a completely different game. It is not like a chatbot and cannot be approached that way. There is no "generate it again and better." The cycle either gets done or it does not. The pallet either falls or is placed well. The safety officer either intervenes or does not.
We have been building Robot George for eight months with this boundary in mind. The AI in him is not enough to just "answer correctly." It must understand space, react in real time, and function in the physical world. A sandbox alone is not enough for that. It has to be tested on real hardware in real testing.
This and much more Zdeněk included in his talk.
Why We Say It Out Loud
In the Czech Republic, AI is talked about primarily in the context of apps and chatbots. Physical AI is a smaller, younger, and technically more demanding field that is only forming here today. What Zdeněk said in Zlín about what we underestimated and what we succeeded at was for us a step toward an alliance with people who are solving similar problems.
"When AI is not enough to just 'answer correctly' and must understand space, react in real time, and function in the physical world, it is a completely different type of engineering work," said Zdeněk Huspenina, our Chief Business Development Officer.