Real Eyes Realize A.I.
how the rapid development of AI has everyone second guessing reality
Michelle’s piece last week about artificial intelligence developing recipes got me thinking about how AI might impact my job. I’m the resident photographer here for Bad Manners and recent advancements in AI has plenty of people on edge, especially creatives.
If you’re wondering why everyone’s talking about AI images, blame a viral photo of the Pope that duped half the internet. The advancements of new tools like ChatGPT and MidJourney have sparked a global conversation about reality and the ethics of using AI tools.
Everyone has concerns about the rapid advancements in this technology. Italy officially banned ChatGPT while European Union authorities have created a task force dedicated to AI oversight. Even Elon Musk has asked to pause developing more powerful AI tools as a “risk to society” buuuut it’s hard to take him seriously, especially after Musk announced he’s developing his own rival AI company. Sounds more like someone just wants to play catch-up.
Humans have been training machines to learn from us for a while now. The first use of the word “artificial intelligence” appeared in 1955 and a year after that, Arthur Samuel revealed a computer playing checkers on national television. We’ve been giving computers data and asking them to feed our own data back to us for decades. This isn’t new but modern AI platforms have the entire internet at their disposal to solve whatever data requests we can throw at them. So when someone prompts an AI system to generate “US Presidents playing Mario Kart together but only speaking in gamer vernacular and bickering like siblings”- well, voila.
Those faux-President voices are close but not perfect. The faux-Pope’s hands look fucked up. You can just tell some things just look weird, advanced AI still has some blind spots. Even fake Instagram models like Miquela, Ailice, or Imma just look… kinda off. And I don’t mean ‘fake’ as in airbrushed into oblivion, I mean like they do not exist. They’re avatars with curated personas and their accounts are managed by humans.
Enter the uncanny valley.
We’ve mentioned it before but the idea of the uncanny valley is when something mimics human appearance, speech, or text BUT we instinctually recognize that it’s not human. You’ve experienced this feeling if you’ve been to a wax museum or answered a robocall contacting you about your car’s extended warranty.
From a distance it’s almost convincing but humans are perceptive and skeptical enough to know when something is bullshit. So let’s feed AI some prompts for food photography and see what it shits out.
Not bad but we all look at that and recognize that everything is just a little fucky. Closer inspection shows oddly shaped veggies, inconsistent lighting, and what appears to be a mound of black pepper? And what’s with the foot of that bottom right bowl? Your brain wants to recognize the image but it detects bullshit.
LET’S SPIN IT AGAIN:
I’ll admit, the countertop reflection of the glass is impressive. These are convincing at a glance but the longer you look, the more something in your brain is telling you “this isn’t real”. The backgrounds are generic enough but the focused subject looks confusing. And why does the wrap on the bottom right have so much texture? And the bottom left sample can’t decide if the shadow should be square or match the plate. Also please explain what those 4 drinks are in the background the AI generated more liquids than it did food.
But let’s zoom out and try to incorporate people with food:
Looks more like a painting than a photo. Fingers really confuse current AI, it either adds more than necessary, not enough, or just blurs the hell out of hand altogether. If you notice in the top right sample, AI merged the man’s hand with the loaf of bread in front of him lol. The horror.
One more SPIIIIIIIIN!
What is that hunk of meat (?) floating in the top left sample, it’s like AI knows what an al pastor is but rendered less than half of it. The faces and hands look distorted in every sample and I’m learning in real-time that AI also seems to struggle with text. These menus are gibberish. I do appreciate the lighting and wet cement textures but real eyes realize AI.
This is an interesting exercise in AI delivering ‘likeness’ of food and humans cooking, ultimately our built-in bullshit detectors win. And yes, I’ve heard all the hand-wringing arguments for legislating artificial intelligence. Honestly it all just kinda feels like the announcement of the VCR- a disproportionate response over new tech backed by hypotheticals. Also, false advertising laws about misrepresenting products, , especially food, to protect the public have existed for years.
We’ve all shared a cynical laugh about how the food advertised sometimes isn’t what we get. But menus are legally required to be representative of the product they’re selling. Or they get sued which happens with a regular enough frequency to protect consumers. We could use new vocabulary beyond "“the uncanny valley” with these new advancements. Food fraud? Deceptive dishes? Faux ex machina? I’ll workshop it and get back to y’all.
I can’t speak to other creative fields like AI successfully mimick’ing Drake or lying to a Task Rabbit, but I feel confident that my job as a food photographer (and Michelle’s as a recipe developer) are safe, for now.
Change can be scary, especially something none of us fully understands. Including whatever clueless nerds created AI but most of us already use artificial intelligence every day and you probably don’t even realize it.
Next time you pull your phone out to snap a photo, watch how quickly the camera will auto autofocus or automatically find people’s faces. Yup, AI bitch. Scrolling your social media feed? The content is AI curated. Own any smart devices at home like Alexa, Siri, home security or thermostat you can control with your phone? All that shit is AI and you’re perfectly comfortable using it because it’s incrementally made your life easier than it has harder.
So ~relax~ because if there’s anything that should keep us up at night, it’s probably broken arrows. Besides, robots don’t fucking want our jobs either.
Sure AI has potential to cause harm but so does everything in our lives. Bananas are naturally radioactive and if you eat enough you’ll die from radiation poisoning. You drive a car that’s a glorified molotov cocktail on wheels but you have faith that the benefits outweigh the chance of disaster.
I’m not buying into the “sky is falling” narrative towards emerging technology. Embrace AI. Adapt. Use it as a tool. Not only can it make your life easier but you’ll have a sharper sense of detecting bullshit as AI evolves, and so will you.
Our Sunday subscribers will get a recipe for Vanilla Bean Crème Brulee, sent out ONE DAY EARLIER just in time for Mother’s Day.
Not a subscriber? Come sit at the cool kids table, always room for one more.
Don’t forget to follow us on social too.