Faking it: What if AI clogs up the system and slows planning down?

“I wish someone was fishing right now," I say. I’m standing on the riverbank in Biidaasige Park, shooting photos for the magazine. 

 

I was sent a press kit of images, but the pictures showed the park before opening day and featured a freshly unboxed playground empty of children. Three months on, I’m here to capture the park in use, discover what’s popular and how it works. This is my third visit and it’s a beautiful day.

 

"I can do that," says a friend. He snaps a photo of the river and writes a prompt to insert a fisherman: "See? Fixed it for you."

In the image, a middle-aged man in green oilskins stands in the river. Cast aside (sorry) the fact that most urban fishers I’ve seen in Toronto are teenagers. They cruise on their bikes, rods on their backs, hoping to catch a twenty-pounder during the running of the salmon.

 

 

But it’s not the cliche fisherman that bothers me. This is the second time in a week someone has suggested AI in lieu of real photos. When a third person suggests adding fake children to the playground, I snap: "Shall I just insert some rubble and dead bodies and call it warzone coverage? This is journalism. If kids are playing on the swings, that is useful information."

 

The offhand comments gnaw at me. I resent the way AI is being used to cosplay eyewitness journalism; the way faking pictures has become normalised, filter by filter, leading to this. We used to protest when magazines airbrushed and thinned models. Now, we do it to ourselves.

 

We are entering an age of mass deception in which we routinely lie to each other. In this post-truth hellscape, anyone can create a convincing false narrative and misrepresent, not just the future, but the present. If it carries on like this, we will surely lose faith in photojournalism. How to hold power to account if they can plausibly deny the photos are real?

 

Last month, the journal Cognitive Research: Principles and Implications published research by Swansea University showing that people were no longer able to distinguish between AI-generated images of real people and actual photographs - even if they knew the subject in real life.

 

 

"The fact that everyday AI tools can do this not only raises urgent questions about misinformation and trust in visual media but also the need for reliable detection methods. as a matter of urgency," said Professor Jeremy Tree from Swansea’s School of Psychology.

 

AI fakery is already a problem in the property industry. Estate agents are doctoring the photos of real flats to encourage viewings. Consumers are raising the alarm regarding property listings that use AI to change the flooring, revamp the kitchen, even add windows. Although agents have argued these are useful illustrations of a flat’s potential, others are questioning whether this amounts to consumer misrepresentation. The images do not take into account the structure of the building or what alterations would be feasible.

 

AI-powered tricksters are a growing risk too – and the built environment is not immune. Arup was defrauded of £20 million last year when a virtual meeting with the company’s CFO, apparently instructing the transfer of funds, turned out to be a realtime, deep-fake video call with a fraudster.

 

"I think we really do have to start questioning what we see," said Arup’s chief information officer, Rob Greig to journalists at the time. After the incident, Greig tried to create a similar video of himself. "It took me, with some open source software, about 45 minutes:’ he said. And that was a year ago - given the pace of change, the process will be faster, easier and possibly even more realistic now.

 

You might think that we can deploy AI to detect fraud, but it has proved a poor judge of character so far, at least when it comes to separating the truth from lies. A recent study out of Michigan State University and the University of Oklahoma warns that certain large AI language models should not be used for lie detection. Indeed, AI thought most of the people telling the truth were lying.

 

 

At least AI can be trusted to run procurement processes, right? But there are fears that the speed and convenience of automation could gum up the system. This doesn’t just mean Design and Access Statements written by AI. The Guardian reported that the government’s plan to use AI to accelerate planning might in turn face ’’Al-powered nimbyism".

 

Paid services are offering to automatically raise planning objections using AI. These will scan for applications near a property’s address, draw up legal arguments against the development, and send off letters of objection (although some of these have included case law that doesn’t exist). Services start at just £45 and claim to level the playing field for citizens – a field that is set to get crowded.

 

AI is already being used by some firms to generate responses to RFPs to save time. This could help smaller firms with limited manpower, enabling them to bid on more projects. But an increase in bidders will up the workload on the receiving end. This would likely see RFPs being screened by AI to whittle down the number of applications – a process that could also discriminate against smaller, non-traditional firms (as shown in recruitment) depending on the algorithm criteria.

 

Is the future of planning and procurement to be found in machines speaking solely to other machines? Is it efficient to let AI modules submit and object to each other, data richocheting in the cloud, until it spits out an answer?

 

AI speaking to AI is not a good idea. Scientists are beginning to warn about model collapse: This is what happens when AI models run out of human-made training data and instead learn from AI-generated text. The result is writing that quickly degrades into gibberish. All this automated action has a cost too – data centres have already strained the grid and increased concerns about water shortages in places.

 

 

These thoughts trouble me as I put together the magazine, but the process of doing so is reassuringly human. Photographer John Sturrock made two trips to Paradise Birmingham to capture it for our Placetest. He spoke first to locals and professionals to understand the context, studying the map to agree the geographic boundaries of the story.

 

Sturrock hung around Paradise at different times of day, seeking out rooftop views and forgotten spaces, the backs of the buildings and down the canal, photographing the life of the place. We spoke about his experience and I sought to understand his perspective and consider his observations.

 

At Paradise Birmingham, he was struck by the nonstop walking, people on the move through the spaces, rarely stopping, a pedestrian thoroughfare. "Where are they all going?" I asked him. He wasn’t sure. Into one of the nearby venues.

 

Sturrock also noted there were very few trees or places to sit - apart from lots of steps. Birmingham is a young city, maybe people don’t require lumbar support or the warmth of a wooden bench. Or maybe this is why they move through the space instead of lingering. Who can be sure?

 

At the same time, for the written article, anthropologist Jon Mitchell (who grew up in Birmingham) was asking open-ended questions about the lives of the people who  live nearby, listening to their stories, without agenda. The words and pictures come together on the pages.

 

The Placetest that results from this process is never intended to be a definitive portrait. It captures a moment in time, embraces complexity, takes an impression of a place and preserves it on page. It seeks to be honest. It’s a portrait of right here, right now. Why would it be useful to fake any of that? The whole point is to learn about the user experience of a place.

 

What happens to society when little white lies become commonplace? When community engagement is faked, full of feedback from people that don’t exist, results manufactured, online surveys completed by bots, hijacked by people who aren’t local, photos doctored. This may already have happened.

 

If it carries on like this, we will lose faith in all photography. Some people already doubt the veracity of journalism. If we fail to believe in a famine captured on film, will this lead to political apathy? Will we fail to meet the moment; send aid? What corresponding shift will take place in democracy? In empathy?

 

The fabric of community has worn thin in parts through austerity-driven inequality. What will a further increase in distrust do to us? Humans have a natural bias in favour of believing each other – a bias towards truth. This supports the social bonds we need to survive. What if this reverses because of AI, and we think it’s all a lie?

 

Perhaps, as Marco Ferrari writes in this month’s Photobooth, the antidote to an AI future is analogue. It must be founded in dialogue; face-to-face; moments we can’t fake.

 

As for The Developer, with your help, we’ll keep commissioning writers who visit places and speak to real people, and photographers who capture life on the ground. If you believe in the value of what we do, please consider supporting us – as an organisation member, patron, advertiser or sponsor. The business model of journalism may be broken, but its purpose remains. Thank you to our members for helping us carry on.

 

---

This article appears as the Editor’s letter in our latest print edition of The Developer. Organisation members and Patrons receive the magazine as part of their benefits – please join them or order your magazine here.