The most powerful story could be one helluva an illusion. So what d’you do?

Dr David Dunkley Gyimah
4 min readMay 19, 2023

--

It was one of the world’s most seminal stories when South Africa spun off its yoke of Apartheid (racism legalised). In two days it would be legally consigned to history. I’m in South Africa circa 1994 reporting. Or was I?

It was once held that the camera never lies. So much for that. Today the image does. Last week I was contacted by a BBC Journalist who wanted to interview me by showing me a group of photorealistic images he’d generated

I was slightly ahead of him. If you’re asking me to distinguish between an AI photo and a real one, even some of the best trained eyes for the Sony World Photography Awards had a hard time.

You know the story by now. Artist Boris Eldagsen generated an image called “PSEUDOMNESIA | The Electrician. It’s a mesmerising portrait of two women from the 1940s. It’s every bit the photo you wished you’d taken.

In Eldagsen’s case he had help from Open AIs Dall.e-2. He refused the award, using the publicity to highlight a real concern. What can you now trust?

The image above of me in South Africa is indeed real, but how easily malleable is it? And it’s not just the wholly imagined image that concerns me, but the ability to manipulate an existing photo to make a point.

So instead of us trio examining a crime scene where a bomb has gone off, we could be facing different directions with different expression. We’re there doing something else, but not as the original prints state so.

This week the debate got serious. OpenAI CEO Sam Altman told US senators about his deep concerns with AI’s potential to manipulate views. Google lost one of its AI founding fathers.

And then there are livelihoods. Draped across the front inside pages of the Metro newspaper after King Charles III’s coronation were imagined photos of the royals partying. Too amusing not to pass by and have a chuckle, but the thought occurred to me. Imagine being a professional photographer. You’ve got the shot, and the next day one of the popular tabloids features images generated by AI.

There are also so many complex issues to navigate. Is an AI image copyright and to whom? Obviously the creative. But what if, and there is a debate going on here, the algorithms trained to create images was based on a sample of copyrighted imagery?

MidJourney, the AI tool, is all the rage at the moment and one of the most celebrated artists is Nigerian filmmaker and artist Malik Afegbua whose stylish photos of elderly Black people excites the imagination. This is what commercial enterprises and fashion shows could be doing.

Midjourney stole a march on the industry; it’s no longer free to use, presumably because of volume of users, but it’s also sparked a conversation on craft skill.

This photo below was taken at a gym in Ghana. I love boxing. Twenty years ago I was the videojournalist/ filmmaker for boxer Lennox lewis fighting Tyson.

This image could so easily be generated using any good generative AI software. The publishers don’t have to pay a photographer. They just need an account and someone in house who’s got form in prompting the software.

Similarly this image taken at the South Bank and a night of Afro beat. Easily replicable.

So what does that mean for creatives? And before you think, yes that’s progress, as a creative technologist I won’t disagree. Every era has found an existing form under threat. It happened with photography and Art.

Impressionism was a response to the fact photography captured real life perspectives similar to renaissance painters and in particular realism and neoclassicist painters like Jacques-Louis David.

When mobile film cameras were possible in news, a respectable BBC writer lamented there goes descriptive narrative, because the camera could show you warts n’ all.

When YouTube launched professionals knew that speech they gave about TV being difficult to make wouldn’t last.

What then makes AI generated images such a panic pusher? Previous leaps in technology have involved mastering new tech, with considerable effort at source. Today and in the future it’s plain natural language?

I’m keen to find out the first university course that offers AI prompting skills to creative enterprises in which no other legacy skills are required. That would be something! But you can see it happening.

How do you address AI images? I guess it won’t be long until we hear about an AI filter catcher. It’s already happening with ChatGPT from Edward Tian, a 22-year-old Princeton University student. The BBC interview, which involves a handful of professionals talking about AI comes out this weekend. I’ll keep you posted.

--

--

Dr David Dunkley Gyimah

Creative Technologist & Associate Professor. International Award Winner Cinema journalist. Ex BBC/C4News. Apple profiled Top Writer,