If chatgpt were run by a 19th century mechanical computer
Part of Charles Babbage's Analytical Engine which was fully designed but never completed
Charles Babbage, aided by Ada Lovelace, designend the first Turing-complete general purpose computer in 1837. The engine was mechanical and designed to have input via punch cards, perform binary calculations, and store numbers in memory.
I was curious to know how many of these engines it would take to run OpenAI's chatGPT. To get the answer, I asked chatGPT. In summary, the answer is that, assuming each engine had a footprint of 20 feet by 10 feet, the engines would have to cover the entire planet, with some floating on the sea, or they could be stacked to form giant clockwork cities, which would then leave some room to grow food. Assuming they were steam powered, there might be enough coal on earth to run this monstrous Borgesian machine for 10 years. Of course there would be many other issues of implementation, including the need for users to manually punch cards, feed them in, then collect the printout some hours later. It's a massive interface problem. People would spend their whole lives queuing to execute simple conversations, eventually leading them to wonder why they did not just talk to each other and read books.
