Artificial intelligence spotted inventing its own creepy language

An artificial intelligence program has developed its own language and no one can understand it.

OpenAI is an artificial intelligence systems developer – their programs are fantastic examples of super-computing but there are quirks.

DALLE-E2 is OpenAI‘s latest AI system – it can generate realistic or artistic images from user-entered text descriptions.

DALLE-E2 represents a milestone in machine learning – OpenAI’s site says the program “learned the relationship between images and the text used to describe them.”

A DALLE-E2 demonstration includes interactive keywords for visiting users to play with and generate images – toggling different keywords will result in different images, styles, and subjects.

But the system has one strange behavior – it’s writing its own language of random arrangements of letters, and researchers don’t know why.

Giannis Daras, a computer science Ph.D. student at the University of Texas, published a Twitter thread detailing DALLE-E2’s unexplained new language.

Daras told DALLE-E2 to create an image of “farmers talking about vegetables” and the program did so, but the farmers’ speech read “vicootes” – some unknown AI word.

Fresh vegetables green salad, peppers, mushrooms and eggplant on kitchen tabl
After being told to create an image of ‘farmers talking about vegetables,’ the speech read vicootes,’ an unknown AI-word.
Getty Images/iStockphoto

Daras fed “vicootes” back into the DALLE-E2 system and got back pictures of vegetables.

“We then feed the words: ‘Apoploe vesrreaitars’ and we get birds.” Daras wrote on Twitter.

“It seems that the farmers are talking about birds, messing with their vegetables!”

Daras and a co-author have written a paper on DALLE-E2’s “hidden vocabulary”.

They acknowledge that telling DALLE-E2 to generate images of words – the command “an image of the word airplane” is Daras’ example – normally results in DALLE-E2 spitting out “gibberish text”.

When plugged back into DALLE-E2, that gibberish text will result in images of airplanes – which says something about the way DALLE-E2 talks to and thinks of itself.

Some AI researchers argued that DALLE-E2’s gibberish text is “random noise“.

Hopefully, we don’t come to find the DALLE-E2’s second language was a security flaw that needed patching after it’s too late.

This article originally appeared on The Sun and was reproduced here with permission.

Steve Liem

Learn More →