On Robots

01 Jan 2020

"Puzzles to solve and matter to organize. I have a body in this world with muscles that have the potential to incur energy upon matter. I could do more to reach the bounds of the world that my ego inhabits to discover what the self is." - Thomas Grothe (September 9 2019)

Artificial general intelligence(AGI) is a machine that mimics the cognitive functions that humans associate with the human mind, such as "learning" and "problem solving". Billions are spent on artificial intelligence research a year. Even though tremendous progress has been made in the past decade, we still don't have cars that can drive themselves in any environment as was promised by experts in the field. This illustrates how formidable the task of creating AGI is.

A robot is a programmable computer that can perform a series of tasks. A robot is different from artificial intelligence because it's not supposed to simulate the human mind. The robots most people tend to imagine are ones that attempt to mimic the movements of the human body.

Mimicking the movements of the human body is easier than mimicking the human mind. Most people want an AGI because they believe these machines will be able to solve problems that human beings can't solve. Even if that's true, an AGI will have free will so it might decide that it doesn't want to solve a problem humans want to solve. That is the opposite of a tool! It might be better for mankind to hold the monopoly on knowledge creation; where we can use that knowledge to program robots to perform tasks we want rather than constructing a device that can potentially outhink us.

We create tools to make our lives easier. Early technological revolutions were in agriculture and manufacturing. Later revolutions have been mostly digital. The current technological revolution merges the manufacturing with digital. The outcome of the current revolution is humans using computers to design and make all physical objects. (1)

Software is very easy to scale compared to hardware. Software can be copied on billions of devices. Copying and installing Linux to a new device is much easier than copying a Tesla factory. That is why most top ten valued companies by market capitalization are software companies. There are no manufacturing companies to be found! Even though manufacturing cannot be scaled as well as software, how close can we make it if we tried?

To make the game even more stacked against manufacturing, computers are universal devices due to the Church-Turing thesis: any universal computer can do anything another universal computer can do. Analogously, imagine a factory; a universal factory that can make anything. That would be a very powerful device indeed. The process in building a universal factory is very hard compared to building a universal computer because the real world is messier than abstract mathematical objects like computable functions. Progress in using computers to design and make physical objects is being made via 3D printing, 4D printing, topological quantum computing; programmable matter in general. We don't know for sure what the first universal factory will look like. (2)

Von Neumann's universal factory is a machine that can create any other machine that can be programmed in its cellular automaton and described in a finite but arbitrarily long tape. Human beings have built a lot of amazing objects; pyramids, spaceships, superconductors e.t.c. It follows from this that a human being without free will and creativity (a human being without a mind) is a universal constructor. From this definition, Humans are more powerful than a universal factory because a machine that can build another machine from a program doesn't need to be creative to do it's job; it doesn't need to design the machine and write a program that builds it, it just executes the program it's given. (3)

Fab Labs allow anyone to make (almost) anything. The (almost) hints that they are our next step to building a universal constructor. Assuming all the layouts of an official Fab Lab are the same, we can build robots that are specifically designed to navigate the Fab lab and place them in the lab. Robots that are as good as humans in vision smell, feel can be universal constructors because they can theoretically build anything humans can build given a program that describes what needs to be built. Humans can program the Fab lab by building something in it while The Fab Lab will record their movements. Once the object is built by humans, The Fab lab can make instructions from the recording and be able to clone the object that the human made. The only resources required from humans is the knowledge of how to build something and the raw materials required to build it.

It's a nightmare when technology goes out of control. A machine that decides it has no use for us is about as bad as it gets. Machines and humans should work together where humans come up with novel solutions to problems and machines solve those problems scalably. For this reason, it's better to invest in creating a universal constructor to augment our minds rather than an AGI to take over it.

Thanks to Zuriel Omole for reading drafts of this.



(1) Digital Fabrication and the Future of Work, J. Cutcher-Gershenfeld, A. Gershenfeld, and N. Gershenfeld, Perspectives on Work, Labor and Employment Relations Association, pp. 8-13 (2018).

(2) Nielsen, Michael A. “The varieties of material existence”, 2018.

(3) Deutsch, David (2012). "Constructor Theory". 1210.7439