Advancing robots.

Robots are an extension of AI. Or they are a physical connection to AI. That makes robots both interesting and dangerous. Every day we read stories about mistakes. That automated systems make. We might think those systems aren't a threat to us at work. We make everything better. We are more sympathetic. And that is one of the things. That we sometimes misunderstand. 

The robot must not be intelligent. That. It can do its work. We can measure. The robot's effectiveness easily. Counting the number of accepted tasks. How many accepted pieces it can make. That is an easy way to calculate effectiveness. 

When we think about manufacturing robots. Those robots. They are more effective than humans. They don’t even know what they are doing. But they are not intelligent. And this is why we use robots. We use robots because they don’t resist. This makes robots a useful tool. In. The hands of the military. 

Flyer 1

We can look back in history. When. We look at things like the first car. Or. The first aircraft. Could somebody believe that the Flyer 1 that flew in Kitty Hawk would someday revolutionize transportation, research, and warfare? 

That is a good question. When we look at the Flyer 1. We see. How primitive this system is. Then we can ask why it's primitive. Because.  It was the first heavier-than-air machine that flew using its own engine. We can compare that primitive system with a modern F-35 and say that the first aircraft was dangerous. It was useless. At the same time, forget one thing. The F-35 does not exist without that primitive machine. That flew a couple of meters at Kitty Hawk in 1903. It’s hard to be number one. 

The second fact that we see is how robots make mistakes. And we can laugh at those mistakes. We forget that AI-controlled robots and LLMs are less than ten years old. The robot itself is not new. The person who introduced the idea of robots as mechanical men is unknown. 

Centauri Robot. 

Leonardo da Vinci tried to make a mechanical man. His invention was a mechanical knight. The first functioning robot, “The Digesting Duck, " was built by a French inventor named Jacques de Vaucanson. If. We compare that robot with modern robots. That tool was very limited. But that mechanical duck could transport dynamite to enemy positions. This means that the user determines how to use the robot. And the people behind Vaucanson. They wanted a system that pulls firewire behind it. And carries explosives to the target. 

“The Goliath tracked mine (German: Leichter Ladungsträger Goliath, "Goliath Light Charge Carrier". (Wikipedia, Goliath tracked mine)

“FireAnt unmanned ground vehicle”. (Photo by Austin Thomas)(DefenseBlog)

Same way. We can think of the German “Goliath” kamikaze tank. The tracked mine. That travels. On tracks was a solution. That could be useful. Lack of fuel and parts. Caused. There were too few of those systems. The “fire ant” unmanned ground vehicle is quite similar to “Goliath”.It can run below the tank’s tracks. It uses more advanced systems. There is a possibility. That those “fire ant” mines could carry bazookas. To make them more effective. 

"Model of Leonardo's robot with inner workings, on display in Berlin" (Wikipedia, Leonardo's robot)

By the way. Vaucanson called his robot an “automaton”. The word “robot” is from the 1930s. And when we look at things like robot taxis and primitive worker robots. We forget one thing.  Those systems that we see in museums are primitive pathfinders. This means that when we look at modern robots and movie robots. We must understand that things like AI agents. And PHP protocols. And. Morphing neural networks make those tools more versatile than the 18th-century mechanical ducks ever could be. 

PHP is a protocol. That makes. It is possible. To run code on the server. This means that. The workstation outsources the computing to the server. This means that. The server and its owner seed everything. That server does. And what does the system do? It processes information. That. The user gives to it. 

When Vaucanson made his first robots. He. Didn’t have networks and AI. The materials that he used were limited. And that is one of the things. That. We must understand. We can make Terminators. If. We mean man-shaped robots.  The man-shaped remotely controlled robot is possible. And the only thing. That is not possible is the computer. That allows those systems to operate independently. So. We can make a remotely controlled version of Terminators. But. Technology becomes more advanced.

The ability to interconnect servers over the network. It allows us to break the limits of the systems. And that can cause situations. That. We cannot control. Advanced algorithms can recognize that the computer does not have enough power. And call other servers to assist. That causes a situation. Where. The calculation power jumps to a new level. 

In this case, it is possible. The server operators don’t see that thing. This means that the AI runs on a server platform. That power can be a million times higher. That predicted calculation power. 

“Digesting duck”. Or. A postulated interior of the Duck of Vaucanson (1738–1739)(Wikipedia, Automaton)

Or. We can say that the robot must not look like any animal. It can look like a Centauri. Like the robot in the second image. But there is one danger with robots and algorithms. Especially.  Algorithms from the late 20th century. They were made for individual workstations. Those systems had 640 kt. memory. They were not used for network-based AI solutions. When. A neural network runs those algorithms. It can change its operations. Much faster than a 1980s PC. This means that the platform makes those systems more effective. 

Then networked server solutions run the same algorithm in multiple places. The system can point. The same algorithm can follow multiple objects at the same time. And that makes them more effective. Another thing is that. The AI has more data than ever before. The AI solution is not the same as a Large Language Model (LLM). The solution is the interface that connects the solution to the LLM. 

Even if we run the AI agent. Or. Some AI solution on our servers. The solution itself is connected to the LLM. That runs in a data center. The LLM is licensed software. And the solution is the interface that communicates with the LLM. This means that the data that we think we own runs on remote servers. 

https://defence-blog.com/u-s-army-tests-robotic-tank-killers/

https://en.wikipedia.org/wiki/Automaton

https://en.wikipedia.org/wiki/Goliath_tracked_mine

https://en.wikipedia.org/wiki/Jacques_de_Vaucanson

https://en.wikipedia.org/wiki/Leonardo%27s_robot