AI and robots can develop and fix themselves.

Today, AI starts to develop itself. But what if we were to take that ability to the physical systems? An even more interesting aspect is a “cannibal robot” that can collect spare parts and improve itself by utilizing another robot’s components. That means that the new robot can search for spare parts and improve parts from robot wrecks. That kind of thing is fundamental because it allows a physical robot to fix itself. AI-controlled robots that can collect spare parts from garbage and fix themselves will be the new tool for other planet research. The robot that can fix itself can also operate in remote areas on Earth. 

The AI that controls robots can make a checklist of what causes damage to the robot’s body. Then the robot can search for parts that can replace that weakness. If a robot requires a stronger shell, it just searches for plates that it can connect to its body. In the same way, if a robot requires more power, it can search for more servo-engines that can use its manipulators. The ability to search and benefit from raw materials from the environment makes robots more survivable and flexible. The AI develops itself in the same way as a robot that collects spare parts from its environment. 

AI is a computer program. And the computer program is a virtual robot. When AI develops its own code, it actually searches code models from the digital environment. In that case, the AI makes a model or matrix about things that it needs. Then the system will ask if there is some AI that has those abilities. Then the AI asks the code that gives the desired ability to the AI. The AI can also ask another AI to generate the code that it requires. And then the AI connects that code into itself. The problem is how the AI can detect things that it needs. And the second thing is that the AI must create the right form for the query that it makes for other AIs. 

The AI can also search things like internet databases about the code that it needs. And then generate that code in its code generator. In that case, there are two AIs: the first one that connects the code to another AI’s source code. Then the system boots another AI. And after that, the other AI makes the changes for the first AI’s source code and boots the system. The code also requires testing, and in that case, there must be at least two different systems. Because if there is a malfunction in the code, the other part of the system can operate without errors. That system can also remove malicious code from another system. But if there are no errors, the system can scale code changes through the entire system. 

https://www.euronews.com/next/2025/07/24/cannibal-robot-scientists-develop-a-robot-that-can-grow-and-heal-by-eating-others

https://www.livescience.com/technology/robotics/watch-this-robot-cannibal-grow-bigger-and-stronger-by-consuming-smaller-robots