One of the topics in the public discourse that annoys me to no end is the general assumption that any AI agent will act like an evil human and seek to exterminate humanity. I watched an interview with Ben Horowitz from A16Z where he suggests it might be better to call AI probabilistic computing as it is much closer to what is going on. The reason I get annoyed by the general discourse is that everyone seems to assume that we even know what intelligence or cognition is. Which is completely false. There are no working theories of intelligence and cognition that even describe human behavior in a holistic fashion. We have no idea where consciousness comes from or what it even is. AI is simply the next tool in our tool box that simply happens to work in certain contexts and for certain purposes, even if we have no clue how or why it works. But also this is nothing new. When we started to use fire, we didn’t know how it worked, just that it worked. The first theories about electricity were ridiculously wrong, but we used the technology anyway. It is the same with AI. It will be a great tool to explore intelligence, cognition and consciousness, but it will be a long journey before we get there. For now, we don’t even know how a model with hundreds of billions of parameters and trained on a set of data that is smaller than that doesn’t overfit as classical statistics says it should. We need to admit the limitations of our knowledge. So, as a reflection today: do you know how AI works?