Suppose you had to produce a local computer from scratch, completely independent from existing manufacturers like Intel. (If you need a specific location for this hypothetical, let's say Greater Philadelphia.) No chip fabs in China, Taiwan etc. The tech level and computing power doesn't matter, you can tailor the application to the power of the computer; indie local production is top priority.
What is the best approach? A simple conventional silicon chip? Vacuum tubes? Analog? Mechanical?
@Bender Ok, this is genuinely cool, but what can we use it for besides teaching children (or adults) how computers work? I don't care how frivolous the application is, but what actual task could a marble computer complete in a way that is at least marginally better than a human doing it manually?
I guess we need a way for it to trigger actions in the real world, like maybe a physical switch that gets flipped by a marble hitting it (or enough marbles piling up to tip it over).
@skyfaller mixture of valves (tubes), electromechanical assemblies and punched cards/tape with other magnetic storage, rather like a cross between Collossus machine used to decrypt coded messages from the German military during WW II and other early computers used for databases and financial processing..
@skyfaller From studying hardware I really get the impression that the primary concern would be data storage, for which I suspect the only real option in this hypothetical would be mechanical. Magnetic tape?
You'd probably want to use a loop of tape for a register bank (resembling x86's granddaddy), which only leaves an ALU to build. That should require only a handful of transistors or vacuum tubes!
This design should be incredibly slow, but extremely achievable!
@skyfaller I like this thought experiment, I think you need more perpetrators though. You say no chip fabs in China or Taiwan, so expand on where the materials can be sourced from? Lets say the entire thing you are building is made from raw materials. Not all the raw materials of a modern computer can be sourced in North America. Thats fine for this hypothetical, we can build a non modern computer, but then where are the remaining materials being sourced?
@liaizon I suggested we recycle what we need from local sources, but people pointed out that if we're using existing tech, we can just salvage old computers instead of building new ones out of recycling. And truly that's great as far as it goes, but I want to build and maintain a local supply chain for making new computers. I still feel like a circular economy should be important to the thought experiment tho.
Building a modern computer is not my priority, that's why I mentioned retro methods.
@steve Decades sounds like too much time; you might be able to build a conventional RISC-V plant by then, for all I know. Let's say you have five (5) years warning to get things rolling before you have to start producing weird local computers. A time traveler tips you off. ("Why didn't you warn me sooner?" "Do you know how to work this time machine? Neither do I, I just hit the Back button.")
You have enough resources to make something interesting, but not enough to replicate the status quo.
@skyfaller if I got to pick it would be biocomputing. Maybe algal? Just a tank of cell colonies doing math and powered by sunlight, fresh water, and nutrient broth. Need more processing or storage power? Split the colony into two tanks and grow some more.
It feels like the trick is finding an application that plays to biocomputing strengths, while minimizing its downsides. Perhaps something related to chemical reactions? I don't know enough to speculate.
@skyfaller I also don’t know enough, but let’s say you’re right and chemical reactions are the key. Then our next job is to frame other problems in those terms so that they might be computed in this manner. I am reminded of the molecular and chemical computing models Luca Cardelli worked on awhile back. Does any of this stuff feel related to you? http://lucacardelli.name/
@mjgardner So are you recommending bringing back human computers? Or are you suggesting a machine writing on paper with pencils, and then something like Scantron to read it?
I genuinely think human computers are a good idea in a few situations, like I think people should be able to use abacuses and slide rules when practical. Perhaps there are novel approaches to human computing that haven't been adequately explored that could expand practical applications.
@skyfaller I was facetiously suggesting a physical implementation of a Turing Machine, the mathematical model that underlies all computer algorithms. Simulating it (Turing-completeness) is the bare minimum necessary for any system of instructions to express all possible computer tasks.
So a real one can mathematically run anything you want. Anything less compromises your “from scratch” requirement.
In other words, this amateur computer scientist was yanking your infinite chain. 🤓
Add comment