Google’s troubled AI feature has given rise to a hilarious new gay meme format… meet Slurpy Faggi

Photo taken at San Diago Pride showing Grogu from The Mandalorian in Pride themed rainbow outfit

Google’s new AI Overview feature was rolled out to users this week, and it’s certainly made waves: though possibly not for the reasons Google was expecting.

Unfortunately, it seems like the feature still has a lot of bugs, as instead of giving an accurate response, the tool seems to be incorporating weird and wonderful information from, amongst other things, old Reddit posts.

A few days ago, the tool returned a now-viral answer to a question about how best to ensure that cheese sticks to pizza. The AI answer suggested using glue in the tomato sauce… a “solution” that we want to make absolutely clear we do not recommend.

https://twitter.com/PixelButts/status/1793387357753999656

Internet sleuths did a bit of digging, and discovered that the suggestion seemed to have originated from an 11-year old Reddit post by a user called “F**ksmith”.

11 year old reddit post that advises people to use 1/8 cup of Elmer's glue in pizza sauce to get the cheese to stick to a pizza
The advice seemed to originate from this 11-year old Reddit post (Reddit)

Also, Gizmodo reports that the tool is bringing up false historical information that would instantly make any student fail their midterms.

If you ask “which U.S. president went to the University of Wisconsin-Madison”, the AI bot will answer that 13 presidents have done that, but then it goes on to claim that those 13 presidents attained 59 different degrees, mostly after they died. Gizmodo gives the example of 17th president Andrew Johnson, who apparently earned several degrees between 1947 and 2012, even though he died in 1875.

You may like to watch

This, in turn, has led to an inevitable influx of memes in the form of fake AI results, many of them coming from non other than gay Twitter, or gay X, as we should probably now call it.

Slurpy Faggi and Dr. Butto

Film still of Star Wars characters C-3P0 and R2D2
Is R2D2 actually Dr. Butto? (Lucasfilm)

One viral tweet that took in a lot of people (including us, at first – we’re willing to admit), suggested that the AI Overview tool had invented two gay Star Wars characters called Slurpy Faggi, and his boyfriend, Dr. Butto.

https://twitter.com/computer_gay/status/1793683085999083935?t=vE1Uz5wLY60I9dg3TLCFeQ&s=19

According to a screenshot shared by an X user called @computer_gay, the AI Overview tool responded to his Star Wars query with: “Yes, there are some LGBTQ+ characters in the Star Wars franchise, including characters who are openly gay, lesbian, or androgynous.” This is true.

It went on to give an example of a gay character: “Slurpy Faggi: The first openly gay character in Star Wars. Slurpy is in a committed relationship with his boyfriend, Dr. Butto.” This is not true.

Naturally, X/Twitter responded the way it usually does, with lots of jokes. One person tweeted: “That’s what they called me in highschool”. Another shared a photo of C-3P0 and R2D2, captioning it: “Wasn’t it obvious?”

A third referenced the fact that Star Wars creator George Lucas invented a fictional type of music known as “jizz“, so they wouldn’t put it past him to create Slurpy and Butto (not that he did, of course).

Poe and Finn
Star Wars heroes Finn and Poe are famously not gay. (Disney)

Another fake AI Overview tweet applies similar results to a list of Nintendo characters. Apparently Yoshi is a “tender, non-binary lesbian” and Wario is a “messy, polyamorous bottom”. Sounds about right.

https://twitter.com/GraphCrimes/status/1793866408289620214

The AI Overview feature is powered by Google’s AI tool Gemini, intended to be Google’s answer to Chat-GPT.

Gemini was developed as part of Project Astra. The Project Astra website says that the goal is to develop “AI agents that can quickly process multimodal information, reason about the context you’re in, and respond to questions at a conversational pace, making interactions feel much more natural.”

Google also boast that “with a score of 90.0%, Gemini Ultra is the first model to outperform human experts on MMLU (massive multitask language understanding).”

However, based on this initial test, it seems we’d best hang on to at least a few human experts for now – unless we want to spend the next few years eating glue-flavoured pizza and failing history tests.

This article was updated at 3.30pm on 24 May to confirm that the Slurpy Faggi and Dr. Butto result was not, actually, generated by Google’s AI Overview.

How did this story make you feel?

Sending reaction...
Thanks for your feedback!

Please login or register to comment on this story.