Without spending a lot of time digging into it, I don’t think that the pandemic was a major enabling factor.

Like, work on hardware-driven larger neural nets and generation of images had been going on pre-pandemic. I remember articles about Google researchers working on them quite some time back.

searches

https://en.wikipedia.org/wiki/DeepDream

DeepDream is a computer vision program created by Google engineer Alexander Mordvintsev that uses a convolutional neural network to find and enhance patterns in images via algorithmic pareidolia, thus creating a dream-like appearance reminiscent of a psychedelic experience in the deliberately overprocessed images.[1][2][3]

The DeepDream software, originated in a deep convolutional network codenamed “Inception” after the film of the same name,[1][2][3] was developed for the ImageNet Large-Scale Visual Recognition Challenge (ILSVRC) in 2014[3] and released in July 2015.

The dreaming idea and name became popular on the internet in 2015 thanks to Google’s DeepDream program. The idea dates from early in the history of neural networks,[4] and similar methods have been used to synthesize visual textures.[5] Related visualization ideas were developed (prior to Google’s work) by several research groups.[6][7]

https://research.google/blog/inceptionism-going-deeper-into-neural-networks/

Like, if you’ve used a diffusion model like Stable Diffusion and run low-iteration stuff in Automatic1111 or ComfyUI or something, that looks pretty familiar.

And that was being released in 2015, half a decade before the pandemic started, and was based on pre-existing work.

I think that the “why now” question mostly just has to do with hardware reaching the point where you can do some significantly-more-interesting things with neural nets, coupled with some mostly-iterative software improvements.

I don’t think that it was a “people are staying inside due to the pandemic and that drove a lot of change” in the sort of sense that, say, there was an impact on video game sales, something that I was talking about on here recently.

EDIT: If you want another data point, I was writing neural-net-based image enhancement software in the early 2000s. Far smaller neural nets than what people are using today for generative AI stuff, but…shrugs

And while I was writing code for CPUs, I distinctly remember a buddy of mine at the time who focused on parallel scientific computing — though he wasn’t working on neural nets — talking about using GPUs to do scientific computation back then, so I know off-the-cuff that there were definitely people banging on that in that timeframe. That’s over twenty years back, and those two things are the fundamental elements behind genAI image stuff.