This strikes a blow to the entire ecosystem. Now, while I’ve been behind the camera for things I’ll not mention, what strikes me more is how much happens behind the scenes in any given industry.
You know how news has largely gone to shit? Well, part of that – corporate consolidation aside – is that everyone who was supposed to be the last line of defense was either laid off or shunted to a hub while reporters fresh out of college are expected to suddenly be photogs, videographers and social-media experts.
It has backfired spectacularly (the only reason we didn’t run an A1 hed with the wrong name of a gallery in Ashland, Ore., is that I was laying out the page in Texas but had already been the news ed there a lifetime ago and knew damn well what it was named and physically where it was [see also: getting street names wrong because you’ve never lived there]). That’s not at all generative “AI” but indicative of how these things go.
Quality goes down, and in this case there might be a few extra fingers, but more to the point, what are they doing? You’re a fashion retailer, and you’re going to graft the product you’re selling onto a body? And “AI” is going to make this look realistic?
That’s not how this works. Wow, are companies going to be surprised to discover that marketing via generative models raises more problems than it solves. The tech isn’t there yet to make your shirt look form-fitting on a digital twin. I mean, run it enough times, and you’ll get close (monkeys, typewriters, Shakespeare), but at that point, what are you really saving in terms of outlay? You’re paying an “AI” firm for GPU cycles sted professionals who know what the fuck they’re doing.
Not to mention the poor timing on this. Generative models have valid applications, but the current landscape is one where more and more people are coming to understand the limits of what it can do. There’s a needle very close to the bubble at this point, so this is not when you want to lean into tech that is proving itself not up to the task just in text and even worse in other spheres.
Simply put, humans are still better at everything in the arts (I’ll grant the utility in the sciences of being able to perform analysis in orders of magnitude less time in certain cases) than any of these models. And photoshoots are an art, not a science.
They’ll no doubt push ahead anyway. But the ROI isn’t what they think it’s going to be.
I agree with you in sentiment, but unlike you, I DO believe AI can easily get your child slave work shirt to fit better on the non existent model made out of an amalgamation of all your previous photoshoots plus stolen data.
It keeps getting more and more accurate every day.
Even if it doesn’t fit great straight out of the generator, you already have a Photoshop touch up artist who will fix that and any extra fingers. It works out. H&M may have to pay some extra 20% to the Photoshop designer in India, but will spare themselves paying a photographer and a model by 100%. Plus any other people in charge of recruiting models and organizing shoots now can get their hours reduced too.
I really really hope things end up like you say they will. Nothing would make me happier than seeing these idiots hitting a wall.
H&M claims in the article that they’d still pay the models prevailing wages to use their likenesses without them having to do anything, which on the surface sounds like a good deal. Further recruitment seems unnecessary when you have a stable of 30 digital twins who will never age.
And sure, a Photoshop user making slave wages can fix whatever issues the “AI” spits out up to a certain point. That’s not to say it will look as … uh … natural as airbrushed, toned pics.
That’s just the model getting paid- which is good - but the pay rate isn’t the same as the one for the shoot. It’s the rate they get paid for the images to be used, something similar to royalties so to speak. It will greatly benefit a minority of popular models, and drastically reduce the job pool for new models.
The photographers are getting totally screwed though, as well as the make up artist and anyone else involved in a shoot.
The photographers are getting totally screwed though, as well as the make up artist and anyone else involved in a shoot.
I mean, that was the central thesis of my original response.
I’m just waiting for OpenAI or someone else to jack up prices in the name of profits so high that they will be the equivalent of Nvidia for business AI.
Suddenly hiring for creative roles will be back in full swing because companies won’t be able to afford it. Unfortunately, in the end, the shareholder always wins
Profits? What are these profits of which you speak? These firms are losing money hand over fist over other fist and promising investors they’ll make money Soon™.
I was working in tech ahead of the dot-com crash in 2000, and this is exactly what we were seeing. The tech wasn’t ready for what a number of companies were claiming. As my NDA with a defunct startup holds no water 25 years later, I’ll point out our business model was tablet development on the hardware side and roughly where search got to around 2015 on the software side.
The all-hands meeting where it was announced we were pivoting to just being a search engine (established players were plentiful at the time; this is before the rise of Google) led me to quit. It was very clear to me – at 20 and as a drug-addicted raver – that this was a road to nowhere.
I’ve seen this movie before. MBAs are fucking idiots.
I’ve seen this movie before. MBAs are fucking idiots.
Truer words have never been spoken.
Moreover, the company name made no sense once the hardware portion was removed from the equation. They shut down six weeks after I left when the VC funding ran out and they couldn’t raise another round.
How much VC/Microsoft funding does OpenAI have? Surely they can’t afford to keep burning cash like they are now? When will they make AI so expensive to use that it will become useless?
The only AI startup I care about is Mistral. Amazing models, altruistic attitude towards AI and software. In my opinion, one of the few great companies out there. Give them a few more years and I’ll consider them in the same league as Mullvad.
Altman claimed a need for $7 trillion to really get the job done. When Musk can’t cover the bill, maybe you’re just delusional.
Microsoft has sent at least $13 billion their way. And got the reviled Copilot for their troubles. They’re pulling back.
It’s completely absurd to me that as someone who didn’t even finish undergrad, the state of the emperor’s clothes is obvious, yet we’re playing this game of pretending there’s value where none exists. Nvidia is the only winner here. My guess is their shareholders are also going to get fucked without lube sometime in the near future.
Tablets obviously came to pass, but 2000 was not the correct time with the state of the art (resistive screens, anyone?). LLMs are currently here. I’d imagine we will get to the point of “AI” becoming AI, but it isn’t now. And you can’t will that into being, but at least we’re gutting research funding at the federal level to further gum up the works.
I tried out Mistral ahead of responding, going so far as to install the app. It’s uninspiring. The tech just isn’t there yet. I asked it about my job search, and it provided suggestions I’ve already been doing for years; upon providing clarification, it spat out more useless obvious things. Is the target market people who’ve never had a job before when asking such questions?
– sent from my Mulllvad VPN
I’m loving your commentary on the subject.
When I mentioned Mistral, I meant their attention to open source. Their new model (Mistral small) can be run on consumer hardware with similar results to ChatGPT if trained on good data. AI isn’t that useful outside of me asking it to write one-liners but I haven’t had the experience you have.
Consumers can ultimately help by speaking out / boycotting brands that do this ( CHEAPSKATES).
Didn’t Black Mirror have an episode about this?
