The New AI Buzzword Is ‘Slop,’ and It’s Messing With You. What to Watch Out For – CNET [CNET]

Disturbing pizza toppings and humans with extra fingers are just a few examples of AI slop.

Fake images of former and current world leaders getting arrested. Glue as a pizza topping. AI-generated images that just can’t stop adding extra fingers to hands.  It’s junk, and now there’s a catch-all term for bad, useless or misleading artificial intelligence: “slop.” The term is spreading across tech blogs, mainstream media and Reddit, where countless threads point out egregious instances of AI gone wrong.

If slop sounds familiar both as a term and in its meaning, that may be because it’s a cousin of spam, which emerged way back in 1993 as a word for unwanted, often auto-generated emails that have been clogging up digital inboxes for decades.

AI’s growing power means that its ability to create new text, images, video and other types of content and throw them onto the web without much, if any, human interaction could lead to more and more slop clogging up fake web pages, social media accounts, message boards and anywhere else we dwell online. 

Don’t fall for slop

One problem with bad AI, however, is that people may not be able to tell it apart from legitimate content. When AI “hallucinates,” or offers up bad or out-of-context information, it’s not always obvious. 

Sometimes AI can be misled by satirical or purposefully misleading data pulled from websites or other sources, or it can simply be biased by the type of data it’s been trained on. 

Read more: Glue in Pizza? Eat Rocks? Google’s AI Search Is Mocked for Bizarre Answers

It’s can be hard to verify whether an image or video is faked by AI sometimes, and with text, it’s not always clear how the information is being sourced. It’s always worth making sure that information offered up by, say ChatGPT is current and that it’s been sourced from a reputable site or set of data.

Other things you can do to avoid falling into the slop:

Stay sharp, and look for AI quirks: Strange phrasing and irrelevant tangents can be signs of AI-written text. Look for unusual facial movements and bizarre background blending when examining video or photos that look suspicious.

Where is it from? Check the source. Is this information posted by The New York Times, or a site you’ve never heard of?

Check it out: Google the post’s information if you’re suspicious, and look for backup from recognized authorities.

We may not like it. We may grow to be overwhelmed by it very, very soon. But at least now there’s a name for all this slop.

Read more: AI Misinformation: How It Works and Ways to Spot It