New News About AI for Photographers who Like News About AI
I tried to write that headline so it would sound like AI. Because irony.
You did not read anything in these pages about the recent photo-specific zeitgeist watercooler moment (the royal family Photoshop thing) not strictly because I’m a snob who’s above hawking titillating bullshit, but mostly because it struck me that there was just no “there” there. Plus, it felt a bit like any additional attention would have just been adding to the noise. And now with the benefit of hindsight I understand how true that was.
Mentioning it at all feels a bit like uncivil discourse, in fact, but I wrote that so I can now write this: We make our culture. It doesn’t happen to us. We don’t receive it. We swipe and click it into existence every day. You’re doing it now by reading this instead of the entirety of human consciousness that exists elsewhere online.
Propelled by the power of social media and the replacement of human tastemakers with algorithms, we have altered our culture at its most basic level. It’s not just that we’ve changed the culture, we’ve changed the way we change our culture.
What I mean is, we are served what we ask for. Our clicks are our votes. Our attention still drives this economy.
This all hit me like a ton of bricks the other day as I listened to
on surviving the media apocalypse. It’s a must-listen as far as I’m concerned.1 In the episode, PJ interviews new media icon Ezra Klein who explains how media is not something we receive passively, but rather something we participate in making.“I think over time,” Klein says, “one of the things that we are going to have to discover is an identity of ourselves as generators of the internet, not consumers of the internet. Meaning, on a small level, every time you read one thing over another, or watch or listen to or spend time on, you are creating more of that thing and less of other things.”
As Vogt paraphrases, ”What we pay attention to online we are willing into existence. What we ignore online we are asking to die.”
So if we like to think of ourselves as thoughtful, cultured, interested people, but instead of reading the Washington Post we watch short videos about royal non-scandals, we demonstrate to the algorithm exactly what buys our attention: more content, less quality.
“I think that we have come to see what we do online as morally neutral,” Klein continues, “with the only moral decisions being made by the providers of the content. But what we are clicking on, reading, where we are spending our time, where we are spending our money… ‘Well we can’t be blamed for that, it’s clickbait!’ If you decide that the way you are going to access the entirety of the internet is spending time on social media sites, and occasionally clicking a link, then what you will get is an internet composed of social media sites. Which is more or less what we have now. At some point, if we don’t want this, if people don’t want this, then they have to do something else. If you find what I’m saying here bad, but you do not subscribe to any local news, you are not neutral in this. You are the problem in it.”
I too am the problem. I’ve been voting with my clicks but slouching around the internet like I’m permanently incognito. I’m not though, I’m fully participating, and the algorithm is keeping track. Time to start clicking like it matters, because it does.
If reading dreck is bad, writing it is surely worse. And since I don’t want that on my conscience I’ve been pretty selective about what I put out here.2 I have resisted doing the things one is supposed to do when attempting to draw a crowd in the creator economy. I don’t like the game and I’m not going to play it, largely because I’m dumb enough to keep doing this (at least for the moment) without it being a profitable use of my time. But I am simultaneously going to try to be a little less precious with what I publish.
That was a really long way of saying thank you for being here and participating in shaping our culture. I’ll keep serving my own interests, and hopefully yours too. But never ever the algorithm.
Five Things I’m Thinking About AI at the Moment
• The Importance of Proper Prompts
I had a real eye opening experience about the importance of well-crafted prompts for generative AI. By way of context, allow me to tell you that my interest and appreciation and general level of worry about AI most closely matches a pendulum in full swing. Most of the time I’m somewhere in the middle, but sometimes I fly out to the extremes and become either deathly afraid of what AI has wrought or dismissive of it as a flash in the pan. I’ll admit to not being anywhere near smart enough to figure out which is right. (And, frankly, I think anybody who claims to be able to do so is at best intellectually dishonest and at worst trying to sell you something.)
I have long thought that, at least when it comes to writing, AI just doesn’t have it. From what I’ve seen with my own eyes and tested with my own AIs, I’ve determined this computer generated stuff works great when it’s formulaic (thank you letters, invitations, certain types of business correspondence), but terrible when it comes to actually writing prose in a way a human will want to read.3
But then I studied up on the importance of writing effective prompts and did a bit more testing and found Chat GPT in particular better able to generate text that reads more like something useful. Maybe not yet artful, but much better than the “use a thesaurus to replace every other word” kind of AI content I’ve mostly encountered. (Maybe this is selection bias, and I’ve actually encountered good AI writing but didn’t know it was AI. Possible, maybe even likely.) The point is, while I still think “AI prompt engineer” is a wild job title, I’m coming to understand better how it’s an actually meaningful skill. Likely a skill that will be expected as part of other job descriptions, if I had to venture a guess. Maybe it will become something that’s taught in business school, like Excel spreadsheets or the discounted cash flow model of business valuation. Or maybe it will require such nuanced expertise that it will actually become a full-fledged job all its own. Again, I’m not smart enough to know. But I think if I were in a dying industry and needed to look for what may be the next big thing, particularly if I thought of myself as somewhat thoughtful and creative, I might spend some time learning the ins and outs of AI prompt writing in an effort to generate something deliberately useful.
• AI Images are Synonymous with Scams
Regular readers know the high regard in which I hold Ryan Broderick’s
newsletter. In his recent piece about AI’s Looming Reputation Crisis Mr. Broderick makes some points that I don’t seem to be seeing other thoughtful web 3.0 types making. Namely, that AI today is more like a crypto scam than a cultural revolution.“There is an entire world of perfectly functional and very boring AI tech that will probably slot into our lives without much societal upheaval,” Broderick writes.
“But many large AI companies use the shinier, flashier side of generative-AI as a way to advertise themselves,” he continues. “This is a problem, and where crypto and AI are most similar. At this point, I’m confident saying that 75% of what generative-AI text and image platforms can do is useless at best and, at worst, actively harmful. Which means that if AI companies want to onboard the millions of people they need as customers to fund themselves and bring about the great AI revolution, they’ll have to perpetually outrun the millions of pathetic losers hoping to use this tech to make a quick buck. Which is something crypto has never been able to do."
“We may have already reached a point where AI images have become synonymous with scams and fraud,” he says. And I think it’s a good point.
I hadn’t considered AI as an extension of other money making scams of recent years, but it makes sense. Any time these unethical sharks smell blood in the water, they circle. I wonder if legitimate AI will be able to “outrun” the scammers or if it’s gonna go the way of the NFT?
• AI is Improving in Some Ways, Getting Worse in Others
Last week I checked in with one of my preferred AI toys (open.ai4 not to be confused with Open AI) to see how it had improved in the last year since I first ventured into the world of artificial intelligence image generation. I was surprised to see how much better it had gotten—in some specific ways (skin, hair, hands, eyes). In other ways it’s weirdly much worse (generic, formulaic, sanitized). The corporate interests have clearly done what they always do, which is take something exciting from its wild west phase and turn it into a watered down shell of its former self. Repeating my previous Avedon and Irving Penn prompts led to more polished yet somehow much less interesting recreations of the vibe of these great artists. If the vibes are off in your AI system, I think it might be doomed, no matter how photorealistic the illustrations may be. When every AI face begins to look like every other AI face, it’s in trouble.
Presumably these applications have been programmed not to infringe on the existing copyrights of those of rich and powerful and famous artists, so we can’t quite get “a photorealistic portrait of a woman striking an active pose in the style of Richard Avedon.” But fear not, even if these AI can’t copy the works of the great masters due to copyright worries, they’ll still happily recreate the work of us little guys, no problem. Sometimes they’ll even just duplicate existing work. See the below image of the Joker, for example. It happened when The New York Times asked AI to generate an image of the Batman villain, and instead it delivered a note for note replica of a copyrighted image.
• It Feels Like I’m Being Singled Out for Destruction
I’ve been encountering a lot of advertisements lately (like, a lot a lot) on the socials about AI generated headshots. Save a trip to the studio, they say, and eliminate the cost of those pesky photographers. Are you all seeing these too, or is it just because I shoot headshots for a living and the algorithm has gained sentience and decided to taunt me mercilessly until it can dance on my grave? One part of me knows that people want actual photographs of themselves, which I believe to be true. But I also know, from sometimes torturous first-hand experience, that a meaningful percentage of the population really doesn’t like how they look and really don’t enjoy being photographed. I believe these people would jump at the opportunity to never pay for my services again. The desire to appear thinner/younger/smoother has never been more profound. This could be a real thing, putting a dent in my business’s bottom line. Do I investigate how to harness this power for myself? I’m unsure how to proceed. I do know the uncanny valley may be my saving grace, but what’s the threshold on that? What’s a tight enough tolerance when compared to reality before we say, “You know, that does look pretty good. I’ll take it!” Yikes. My guess is, more than anything, it will depend on the trends. If it becomes socially acceptable (or really, desirable) to have a particular style of AI-generated headshot, well then that’s what we’re gonna get. Being fashionable is good for the bottom line.
• The Limits of How AI Learns
This New Yorker profile of the hilarious and brilliant Percival Everett is worth a read outside of today’s context, but it was an offhand comment he made about AI that I wanted to mention. It might just be because, as exciting as I find the prospects of useful artificial intelligence, and as curious as I am about how it may change the way we work, some part of me feels like there’s a nonzero chance we’re being sold a sow’s ear disguised as a silk purse.5
“It has no ideas!” Everett explains to a student who asks about using AI for writing.
“Did I ever tell you about John Searle’s Chinese-room thought experiment?” he asks the class. “Imagine that I’m in a room. I have two windows. Through one window, I receive a character. I have a vast array of characters in front of me. Through trial and error, I learn that when I receive this character, I put out that one, until I learn that when I get this message, I put out that message. I can become perfect at that. But what have I not learned?”
The students remain silent. When Everett talks again, it’s with a mischievous, satisfied smile.
“How to speak Chinese.”
Hat tip to Austin Kleon for yet again pointing me to this smart, entertaining and informative something.
• Postscript
It became public recently that the photo magazines for which I used to write were laid waste for nothing. The publisher who bought the titles fired employees, shuttered the publications, then switched off servers and drove several well-established (and even mildly profitable) magazines instantly into the ground. After ceasing publication the new owner sued the old one. Yada yada yada, cut to last month and an out of court settlement followed a statement rescinding prior defamatory statements, and the magazines are back in the hands of the previous owner. But by virtue of the aforementioned decimation, the “magazines” aren’t magazines at all, and must be nearly worthless now. Jobs lost, publications scrapped and readers abandoned. All for what? The removal of a library’s worth of useful information across multiple disciplines effectively wiped from the earth. I can’t help but feel a deep sense of loss and regret because those magazines didn’t need to disappear. They are gone solely due to hubris and ineptitude. What a travesty.
By the way, this conversation between Vogt and Klein goes on to do a tremendous job of explaining why the economics of media— newspapers, magazines, tv, et al—don’t work in our new economy. Ezra Klein: “It’s just really hard to make money advertising as a media organization. The issue is not that you’re competing with the New York Times, it’s that you’re competing with Facebook, you’re competing with Google, you’re competing with TikTok, you’re competing with Netflix. That’s really tough. So what is your differentiated proposal to the advertiser? In the past it might have been, ‘I am how you reach the people of Indianapolis. And you want to reach the people of Indianapolis.’ But do you know who knows when people are in Indianapolis? Meta. Google. They know when people are in Indianapolis and they know how to follow them all around the internet.”
I can hear you all now. “Selective? Really? You have?”
There’s an idea that human reading may not actually matter (see the longwinded intro, above) as we will soon have computers (AI) generating content to be read by other computers (Google, algorithms, etc) with humans not really factoring into the equation at all.
Interestingly, while the site was fully functional a few days ago, as it has been for the last year, as of this writing it’s offline and I can’t seem to determine exactly why. Maybe it’s ominous, or maybe the Chat GPT folks felt like open.ai was a bit too similar to Open AI and filed suit. But either way, this just reinforces the aforementioned point about AI currently living in close proximity to scamville.
I don’t believe it to be a bunch of bull, but I do think it has some characteristic hallmarks of overhyped nonevents of the past. So it could be this. And I clearly, at least on some level, am hoping that’s the case.
Ugh. Fantastic.