Fishy business

Depending on how you look at it, the images are brilliantly surrealStrange, like a dream., grotesque, blasphemous Showing disrespect for God or a religion. or deeply disturbing. They show Jesus with a scaly body or claws, waving at his followers from the sea, or lying down surrounded by small crustaceansCreatures with segmented bodies and limbs and armour-like exoskeletons. They include lobsters, crabs and prawns.. Welcome to the latest internet phenomenon - Shrimp Jesus.

There are other equally strange pictures popping up on Facebook and elsewhere. Some show very old people blowing out birthday candles, others dramatic aeroplane crashes. What they have in common is that they have been created by artificial intelligence (AI).

Two experts at Stanford University, Renee DiResta and Josh Goldstein, have just published a study of these images.¹ They looked at 120 Facebook pages which showed at least 50 of them. The pages had names like Love Baby or Interesting Planet, and had, on average, 129,000 followers.

The images prompted hundreds of millions of interactions. One post alone received 40 million views, putting it in Facebook's top 20 for last summer.

But, DiResta and Goldstein warn, the AI pictures are not just being created for fun: "that the capacity to produce captivating, novel, and immersive imagery, cheaply and instantly... is also what makes the technology appealing to spammers and scammers".

Some "were attempting either to sell products that do not appear to exist, had stolen the pages they operated, or were attempting to manipulate their audiences within the comments". And while the Jesus Shrimp pictures were obviously faked, comments on others indicated that not every viewer realised they were not real.

In fact, says Jason Koebler on online publication 404 Media, "Hundreds of AI-generated spam pages are posting dozens of times a day and are being rewarded by Facebook's recommendation algorithm".

The people responsible for the images often add comments such as "I made this with my own hands", which naive users - often older people - then like, prompting the algorithm to send them more of the same. The images sometimes come with links to pages where visitors are persuaded to buy non-existent products and share their personal details.

In a podcast,² Kevin Roose warns that it is just a short step from Shrimp Jesus to political misinformationIncorrect or misleading information unintentionally presented as fact. It can be contrasted with disinformation or deliberate lies.. As an example, he cites an AI-generated image of an underground city beneath the Capitol in Washington. Though no such place exists, it is the kind of thing that conspiracy theorists are all too willing to believe.

In contrast, some people could become so distrustful of what they see that they start to believe even real pictures are fakes.

Facebook and Meta have now announced that they are creating tools that will automatically detect AI-created content. But, says Kevin Roose, "They're never going to be able to catch everything. And at least so far, it doesn't seem like they're trying all that hard."

Do we need labelling laws for content?

Yes: We have them for food and medicine; AI-generated content could potentially harm a far greater number of people, so it makes complete sense that it should come with some sort of warning.  

No: Most people are not taken in by these kind of posts. Such posts are basically just for fun, freeing the imagination and encouraging people to be more creative.

Or... Unfortunately such laws would be impossible to enforce. The internet exists beyond the control of any government, and even if there were a labelling system, scammers would find a way around it.