Happy puppies and crazy geese: pushing the boundaries of AI absurdity
What is the happiest dog you can imagine? Does it beam with joy on a heavenly plane or frolic in a field of psychedelic flora?
If those images are hard to imagine, fear not, or maybe get a healthy dose of it: Artificial intelligence can bring even the most absurd scenarios to life in vivid color, and on social media some are seeing just how far this can be pushed.
While AI-generated images can often be unsettling due to their eerie realism – think the Pope in a Balenciaga puffer jacket – many are finding joy in a new form of low-stakes image crafting. This fall, ChatGPT released an update that allowed people to enter prompts for more detailed images than before, and it wasn’t long before some started pushing the chatbot to its limits.
In November, Garrett Scott McCurrach, the CEO of Pipedream Labs, a robotics company, posted a digital image of a goose on social media with the statement: “For every 10 likes this gets, I’ll ask ChatGPT to make this goose a little crazier.” As the post was liked tens of thousands of times, the goose went through some growing pains.
The first update was quite modest and gave the goose a colorful birthday hat and a big smile fitting for a Disney character. However, this was the case with the sixth prompt got a second pair of eyeballsdonned roller skates and bathed in a collage of undulating light, brass instruments and ringed planets.
Previous versions of AI chatbots placed the onus on users to provide detailed artistic cues. Mr McCurrach, who uses AI in his work, said using the latest version of ChatGPT was like “talking to someone else with the brush tool.”
“I think this is a really good example of where AI is going,” he said. “We can be much more vague; we can give it more atmosphere than a concrete idea. Then it can start making the assumptions to get where it needs to be.
Regardless of the starting point, the images all seem to end up more or less in the same place: in space, awash in psychedelic blooms. While Mr. McCurrach’s extremely silly goose was one of the first to undergo an absurd transformation, many increasingly crazy images have followed.
In one thread, a man fails to contain his awe at the power of nuclear energy, and is eventually split into dozens of clones, staring open-mouthed, on another plane of existence. Another shows one puppy he becomes so incredibly happy that he is more likely to penetrate the cosmos dissolve in a kaleidoscope of sacred geometry. In another, a chess pawn acquires such supernatural power and terrifying feeling that it looms over the board that once contained it.
Space, Mr. McCurrach said, is at the very limits of human understanding, and because AI is on the surface a collection of what we know, the edges of its imagination mirror ours.
“Watch Marvel movies,” he said. “Eventually they found space and time travel as the last frontiers of creativity.”
Eliezer Yudkowsky, an internet philosopher and self-taught AI researcher, saw these images becoming exponentially more absurd and wondered what the other extreme would look like.
Last month he did ChatGPT asked to give him “a very normal image”. The chatbot spit out a photo of a banal suburb. Pushed further, it produced images of a tidy desktop in a home office and then a white cup of coffee placed against a blank wall. Ultimately, after calling for “terrifying normalcy,” it produced what it described as “a completely blank, featureless white canvas‘, which according to the report ‘represents the essence of everydayness, taken to the extreme’.
One conclusion, Yudkowsky said in an email, was that “the field of AI can never walk across the room without stumbling across a profound question.”
Mr. Yudkowsky found ChatGPT becoming challenging and lectured him about the obstacles in defining “normality.” Mr McCurrach hit a similar wall with the goose, with the chatbot claiming it had reached the ‘peak of foolishness’. They both chose the same strategy to overcome the obstacle: argue. In both cases, ChatGPT succumbed to the pressure and ventured further.
While Mr. Yudkowsky sternly urged him to create increasingly “normal” images, commentators asked whether he was being too harsh on the defenseless chatbot.
“I think I wasn’t really torturing some poor AI artist who might suffer,” he said. “But it’s not a good sign for our civilization that we don’t seem to have any way to know for sure.”
ChatGPT, in turn, assures users that emotions and suffering are not part of the programming.