‘Demonic’ DALL-E Leaves Netizens Shook as AI Appears to Create Its ‘Own Language’
CC0 / /
Subscribe
The DALL-E tool is a neural network “trained” by research and deployment company OpenAI to generate realistic images and art from text in natural language. Developers have previously found DALL-E capable of combining unrelated concepts in plausible ways and creating anthropomorphized versions of animals and objects
Giannis Daras, a computer science PhD candidate, took to Twitter earlier this week to share examples of what he controversially described as the AI’s “own language.”
“‘Apoploe vesrreaitais’ means birds,” Daras tweeted. “‘Contarra ccetnxniams luryca tanniounons’ means bugs or pests.”
A known limitation of DALLE-2 is that it struggles with text. For example, the prompt: "Two farmers talking about vegetables, with subtitles" gives an image that appears to have gibberish text on it.
— Giannis Daras (@giannis_daras) May 31, 2022
However, the text is not as random as it initially appears... (2/n) pic.twitter.com/B3e5qVsTKu
Daras, who co-authored academic research on the matter with Alexandros Dimakis, claimed that certain text prompts trigger the AI to produce images it has linked to the phrase.
“We discover that this produced text is not random, but rather reveals a hidden vocabulary that the model seems to have developed internally,” Daras and his colleague wrote.
From intrigue to talk of the Rapture being upon us, netizens had a wide range of reactions to Daras’ characterization of the AI.
“This is absolutely f*cking wild and I love it,” tweeted Mitchell Phillipp (@MGPAlpha). “‘Our AI made up its own language’ WHAT.”
Twitter user Paul Rowe (@SidikiXavier) comments on Giannis Daras' DALLE-2 theory.
© Twitter/SidikiXavier
The theory was also picked apart by a number of skeptics.
Twitter user Benjamin Hilton (@benjamin_hilton) comments on Giannis Daras' DALL-E 2 theory.
© Twitter/@benjamin_hilton
“If a toddler starts to learn nonsense and speak gibberish, their parents will get worried. Some AI researchers get excited instead,” tweeted Tiago Peixoto (@tiagopeixoto), who said the theory was “specious.”
