When I find myself with time off, my mind often wonders. I tend to think of really unconventional, impractical, and sometimes even bizarre ways to do conventional things. I become captivated by the question: "is this idea theoretically possible?" Over spring break in 2024, I had one of those unconventional ideas.
I had just watched a video about a company developing analog computing technology (meaning base ten, or decary, rather than binary, or digital) and the way these devices provide a substantially more efficient way to do matrix multiplication, making them an attractive technology for supporting neural network based artificial intelligence. The idea is that if you can control the resistance and voltage across a conductor, then measure the current flow, you have measured the quotient of the voltage and resistance, which could be two numbers you wish to divide. Extending that idea, if one wishes to multiply, one can simply apply the reciprocal resistance of the second number. I was amazed by the elegant simplicity of the idea, and frankly I find it beautiful when the 'back to basics' solutions can ultimately wind up enabling the greatest functionality. The idea that one of the most powerful technologies humanity has created could best be run using circuitry inherently simpler than anything in our computers now is just so stunning. It's the same level of beauty I find in Microsoft's Project Silica, which is prototyping the use of laser etched glass for a recyclable, long-term data storage medium. But I had a thought: what if instead of controlling the resistance using charges, you did it with heat? I knew it would be wildly impractical, wildly inefficient, and would never be useful for any practical application. But imagining how it could theoretically be done presented an intriguing mental challenge. I was also inspired by Steve Mould's water based binary adder, which you can find here. More than anything, that's why I took the project on. I am always looking to push my abilities, imagine something outside the box, and learn. I succeeded in theorizing the circuitry, defining three of the four base operations, and even defining integration for such a contraption. I thought of heuristics to enable the greatest possible efficiency, and I even thought of the most suitable materials and design criteria for such a computer. So, with that in mind, below is the rough write-up I did of my work. Enjoy my spring-break rabbit hole.