In 1923, a cartoonist named HT Webster drew a comic cartoon for the New York World newspaper depicting a fictional 2023 machine generating ideas and automatically turning them into cartoons. The title of the cartoon read, “In 2023, all of our work will be done with electricity” and in the cartoon, while the artist was sitting on his own, an “idea dynamo” was producing ideas for him, while a “cartoon dynamo” was drawing.
Interestingly, this business separation is similar to our neural networks today. In real 2023, the “idea dynamo” would likely be a major language model like GPT-3 (although flawed), and the “cartoon dynamo” would look like an image synthesis model like Stable Diffusion.
Here is the cartoon in question:
Interestingly, despite the rapid advances in generative AI technology over the past two years, image synthesis models aren’t quite that good at line art yet, as a cartoonist named Douglas Bonneville has often pointed out on Twitter.
However, advances in artificial intelligence models mastering hand-drawn cartoons can always come as a surprise. Unlike other future projections of the early 1900s that included personal butterfly wings and city-wide tube networks, Webster’s prediction seems pretty close to the goal.
6845
