A funny image on Reddit suggesting what ChatGPT would be like during someone’s childhood had me thinking about how of-the-moment AI chatbots are. The look and feel of conversing with ChatGPT or one of its rivals only really makes sense when everyone is wirelessly online with what once would have seemed a ludicrously fast internet connection linked to a massive digital cloud network.
But it also made me think of my own 1990s childhood. What if instead of diving into my Encarta CD-ROM or hunting through the Scholastic Book Fair, I had a robotic intelligence with a huge database of knowledge sitting in the family PC between the Oregon Trail and SimCity 2000?
I decided to ask ChatGPT to imagine that world for me, and threw in a few other decades just to see what it would come up with.
1990s ChatGPT
For my prompt, I asked the AI to: “Imagine ChatGPT existed in the 1990s. What would it be like and what would people use it for?”
The AI didn’t disappoint in outlining a “DOS-style prompt window with glowing blue screen and white text in boxy font.” The program would apparently take a minute or two to load from a CD-ROM. ChatGPT explained that responses in the 1990s would probably have been text-only, and only as fast as your processor would allow. “A Pentium II could manage a decent sentence every few seconds,” it wrote. Instead of chatting in real time, you’d type a full question, hit Enter, and then watch it slowly generate an answer like downloading an image from the internet.
People might use it the same way they used the internet back then, whether helping you write a book report, giving you cheat codes for Earthworm Jim, or explaining the lyrics to the new Spice Girls song. The beige tower and beige plastic everything for the computer really cemented the look.
1980s ChatGPT
I wanted to see how ChatGPT would reinterpret itself further back, so I submitted the same prompt, but for the 1980s. ChatGPT described a world where it was a program accessed via a Commodore 64 or an early Macintosh. This version of ChatGPT would have been distributed via floppy disk, and you’d launch it from a command line after typing something like LOAD”*”,8,1.
Its description of the 1980s interface was even more primitive, explaining it would be a “black screen with white text, resembling the interface of an early word processor.” Users would type in full questions, hit Return, and watch each word of the response appear one at a time.” Asking who the president is would lead to long minutes of waiting before it said Reagan.
You can see the Commodore device above. Primitive is the best that can be said for it. Even Matthew Broderick in War Games would have had a better computer than this.
1920s ChatGPT
Just for fun, I asked ChatGPT to imagine an entirely fictitious kind of computer, one that might make sense in the 1920s. The typewriter meets oscilloscope seen above was the response. ChatGPT explained that, had it existed in the 1920s, it would probably have been accessed via “a mechanical teleprinter or teletype machine connected to an experimental vacuum-tube computational engine housed in a university basement.”
The interface, according to ChatGPT, would involve typing your question on a typewriter-style keyboard, sending it through a series of cables the size of garden hoses, and waiting while “the machine processes your input using punched cards, whirring reels, and more electricity than a small town. Your answer would be printed on a long strip of continuous paper, and the response would be limited to about 150 words due to the machine’s tendency to overheat. There would be a three-hour delay between each query.”
Though hardly definitive, I will say they each had the personality of their eras. The 1990s one felt friendly but a bit clunky. The 1980s one was more formal, with perhaps the need for some technical training. And the 1920s one felt like what an eccentric inventor would show off in a Jules Verne novel. I do sometimes miss the days of giant chunky PCs and CD-ROMs, but the idea of using dial-up modems for ChatGPT is truly nightmarish.
