Objects for Pondering Computers, Consciousness, and Reductionism
Introduction
When I first encountered ChatGPT, I was in awe. For my part, I didn't think it was conscious, but it was at least a pretty amazing simulation. Perhaps it was a step on the way towards machine consciousness; that was unclear to me.
Humans are susceptible to anthropomorphizing of computer programs. The ancient program ELIZA (1966) was an early example of a chatbot that fooled enough people for the implications to be disturbing to its creator, Joseph Weizenbaum, a professor at MIT.
In 2022, an engineer at Google thought the company's LaMDA AI had reached sentience and was self-aware. Google said no. The engineer was fired, ostensibly for violating product confidentiality.
The entire issue of Consciousness is complex and contentious. To get started, I read some books. But to help me think more, I wanted to create some concrete physical objects as aides to pondering.
There's something about making and handling an object that helps me to engage with concepts that seem fantastical and non-physical. Here are the ones I've made so far:
Computer elements
Early computers were wired together from discrete electronic components. These were organized into fundamental circuits that in some form are part of every digital computer today. Each of these 3 boxes contain a simple representation of one of these circuits, a wiring diagram, and some object associated with the theme.
No matter how complex the behavior of a Large Language Model AI, the machinery itself is reducible to collections of circuits that are related to these.
Memory
Flip-flop circuit; engraved rock. The circuit has two stable states; pressing one of the pair of buttons causes the circuit to switch to the other state.
Decisions
Four logic circuits; gaming dice. Each logic circuit changes its output based on which buttons are pressed.
Clock
Multivibrator, with leading and trailing edge detectors; sand timer, petrified wood.
Reductionist Limit of a Human Being
Looking at the above electronic components helps make concrete the idea that computer hardware is just a collection of parts, and software is just a specification of electronic states. There is no magic, just engineering. A chatbot has no soul; it's just a more complicated collection of those parts and states.
Except for fierce reductionists, it's more difficult to apply the same logic to people. Inspired by an educational mineralogy set I had as a child, I put together a collection of minerals that contain every (required) type of atom that makes up the human body.
The label inside the box is the challenge: "Rearrange atoms to assemble a human."
Representation of a Neural Net
This project moves up to a level of abstraction higher than that of clocks and logic gates. It is a visual representation of a simple neural net. The software neural net was inspired by biological networks of nerve cells, and is a fundamental component of Large Language Models.
The top three nodes represent inputs; the bottom two nodes are outputs. As shown, the net counts the number of activated inputs: if any one input is on, the right output node is activated; if any two inputs are on, the left output is activated; if all three are on, then both outputs are activated. If no input nodes are activated, then a low level of noise flows through the net. The colors of the connections represent weights.
Does this box feel pain?
If a computer was conscious, how would we know?
Some associate the ability to experience pain as an indicator of consciousness. But how can we tell, even with a person, whether someone is "actually" in pain or merely pretending or simulating?
This box has a push button on top. Press the button, and a voice will emerge from the box saying "Ouch." Does this box feel pain?
Demonstrate this box to a small child, and the child may well answer, "Yes." Show it to a sophisticated adult, and the answer will be, "No." Show it to a sophisticated adult with a screwdriver, and the box can be readily opened up to reveal a small Radio Shack recording circuit connected to a speaker. When we see what is going on inside the box, the answer is confirmed. Case closed.
It's much harder to look inside the box of a vast computer program and see what is going on. The question of machine consciousness will ultimately depending on knowing what is going on inside us, when we claim to be conscious, and what is going on inside that vast computer program.
Consideration of the Soul
This box is a side journey that took me to a level of abstraction below that of electronics, to Maxwell's Equations, the fundamental laws of electricity and magnetism. I found it interesting to try out a metaphysical attitude, which is not one that I use. I don't think about whether a computer could have a "soul," but some people do.
The inspirations were the following:
- My freshman physics professor once quoted someone as saying, "The study of Maxwell's Equations is good for the soul." I found that to be an amusing juxtaposition of the scientific and theological modes of discussion, which don't usually meet up.
- I read an article about the art of creating miniature mosaics of religious figures. Mysterious Minis: Intricate mosaics shrouded in mystery
- Portable displays and sets for demonstrating physical principles are a mainstay of science education.
- Multiple religious traditions include the creation of portable travelling shrines.
This box is my answer to the question of what a small devotional shrine to Maxwell's Equations might look like.
When the box is opened, the lights come on. On the left side is a mosaic made from thousands of 2mm seed beads. On the right side are a microammeter connected to a coil and an electromagnet next to a hinged piece of iron, a strong magnet consisting of multiple cubes stuck together, and an RGB beam splitter. Pressing the button below energizes the electromagnet and pulls the piece of iron. Inserting the magnet into the coil and withdrawing quickly generates a current indicated on the meter.
Conclusions
Nope. Still thinking...