The series of images showing Donald Trump being arrested and his daily life in prison, as well as the image of Pope Francis in a white down jacket, were both created in the same AI image generator: Midjourney.
For a long time it was possible to test the service for free and ask it to generate images of a variety of celebrities and world leaders. But after the Pope and Trump photos went viral, access has been severely tightened.
The free version of the AI tool is being discontinued. A long list of words, such as “prison” and “arrested”, may no longer be used.
Eliot Higgins, founder of the online detective network Bellingcat, also claims to have been banned from Midjourney after he asked the service to generate the jailhouse photos of Trump and then spread them on Twitter.
Some tech journalists now believe that AI technology should be handled with as much caution as an atomic bomb. The fact that anyone can create fake pictures of, for example, politicians can mean a threat to democracy, they believe.
At the same time, others argue that the credibility and ability of the AI-generated images to influence the audience is overestimated and described in far too apocalyptic terms.
When SVT Nyheter tests partly Midjourney version five, partly Stable Diffusions Playground, we can quickly generate relatively realistic images of a number of prominent politicians, as well as, for example, the Swedish climate activist Greta Thunberg.
But – when it comes to Sweden’s prime minister, it stops. Why?
In the video above, we answer the question – and so you can see what the AI ”thinks” Ulf Kristersson looks like.