AI can do a lot of things — it fuels our voice assistants, understands pizza, and might even steal our jobs!
And you know what else it can do? Ruin my childhood, and possibly my subconscious.
This was the realization I came to this morning when this was shared in our team Slack channel:
today's nightmare fuel: AI generated Garfield panels #StyleGAN pic.twitter.com/KSA3KhjUHl
— Daniel Hanley ? (@calamardh) July 9, 2019
At first glance, it seems like just the text is scrambled. But look closer at the images:
The 💜 of EU tech
The latest rumblings from the EU tech scene, a story from our wise ol' founder Boris, and some questionable AI art. It's free, every week, in your inbox. Sign up now!
Absolutely chilling. We’re deep in the uncanny valley and I want out.
It’s trippiest if you consider that Garfield himself remains untouched in most of the frames, making it seem like he’s some sort of demon orchestrator cat, warping and twisting the world around him into a psychedelic nightmarish mess.
The tweet was sent by Daniel Hanley, who trained the model himself using an AI called StyleGAN, an alternative generator architecture for GAN (or generative adversarial networks, which you can read more about here):
Yes, I may have overtrained the model. Unlike other #stylegan models I've trained, this synthesis model doesn't produce much variation if random noise is disabled (video). If noise is enabled, we get much more interesting panels, but the interpolations aren't smooth. pic.twitter.com/mEEYOjFDtt
— Daniel Hanley ? (@calamardh) July 9, 2019
Had enough? I’ll leave you with this:
Get the TNW newsletter
Get the most important tech news in your inbox each week.