The AI learns like a child.

Why did the old-timer ATARI chess console beat Chat-GPT? Or why that old-fashioned Chess console could beat humans in chess? The reason for that is the same as in cases where our robot reapers will always get stuck when it works. If we ever play against those antique game consoles we don’t win them. The old-fashioned ATARI involves a couple of mechanic games. But we cannot predict how it moves its buttons if we don’t play against those consoles. That means we win those consoles because we learn how that console plays its game. Those consoles use traditional linear computer programs. If some button is hit the system removes code lines that were meant for that button. 

The old-fashioned ATARI shows that AI requires similar learning methods as humans. So why are our robot reapers unable to do their job? When we program those systems we must stop thinking like programmers who use linear, symbolic programming languages. We should take control of that reaper, and drive the area by pushing that system through the grass area. The system must have navigation tools that help the system to determine its place in the yard. Those navigation tools can be three or four radio lighthouses that help the robot determine its position without the GPS. 

The system can also have a GPS that helps to locate the robot if somebody steals it. When the owner pushes the first mows that helps the robot’s system to determine how much energy it needs. That helps it to plan the battery reload position. The system also needs information about the escarpments and potholes. Those things might be easy for humans or big robots. But for small robots, those things can cause trouble. When we teach AI, we must remember that there are many variables that don’t mean anything to us. But those things are very important for robots who must make complicated things. 

Many complicated things like working in cramped places are automatized in our bodies. But if we want to make a robot plumber we must program every movement that plumbers make doing jobs. That thing requires new programming tools like AI-based systems that can follow the plumber while working. Then that system must copy those movements to the robot’s body. This is one thing that requires advancements. Traditional programming tools are not suitable if we want to describe multiple actions to robots. 

https://www.rudebaguette.com/en/2025/06/chatgpt-just-got-wrecked-by-a-1977-atari-vintage-console-destroys-modern-ai-in-the-most-ridiculous-chess-match-ever/