AI will not replace you. A person using AI will.
This tweet, written by a specialist in the machine learning field, has been doing the rounds lately. Shared and co-signed by a lot of the enthusiastic early adopters of AI image making tools, it's finding a new lease of life outside of Twitter as a response, of sorts, to the concerns of equally enthusiastic early opposers of these tools. A lot of these early concerns are being raised by artists, designers, makers and practitioners within the creative industry. And I think the issue with this tweet being used in an attempt to assuage the fears of creative practitioners is that, I'm not sure the value systems that underpin the tech/AI/ML fields really align with those of the creative industry. A prediction like this for one field, no matter how correct it may be, can have very different implications for another.
If I was an enthusiastic AI image making early adopter, generating mountains of high quality artwork with relative ease, while dismissing the concerns of established, practising artists about my new favourite tech as nothing more than the complaints of pitchfork wielding haters of progress, I would probably love this tweet. It's short and to the point, it has a kind of poetic quality to it, and its got some big mic drop energy going on, in a "and that's the end of that!" kind of way.
But it also reduces a very nuanced, complex topic from a discussion that we should all be having into a simple, memeable slogan. One that people can share in place of forming more fully fleshed out opinions of their own. One that people can just point to and say "see". And that's without mentioning its sort of aggressively cavalier attitude towards the very real prospect of people losing their jobs.
A person using AI will likely replace one who isn't. But what will that person's role look like?
I'm going to take the position that the motivation behind a tweet like this is really to say "those practitioners who don't adopt AI in their workflows will be left behind, and from a professional standpoint, far less valuable than those who do". As far as the tech industry goes, I'm certain that AI/ML skills and experience will continue to be very valuable within its industry, even with the rise of tools like GPT-3 (and the imminent release of GPT-4). I imagine that the margin for error in these fields is very narrow, the difference between output that is 'good enough' and 'perfect' is huge, and the cost of making a single mistake is disastrous on a financial level that I cannot comprehend.
The problem is, none of these variables really align with a large amount of work that makes up creative visual industries, particularly for juniors and mid-weight practitioners.
In the short term, artists, designers and makers that implement AI tools into their workflows will absolutely be more valuable than those who don't. The speed at which these tools are adopted and utilised professionally will be partially driven by client needs and demands - artists using these tools will be able to produce more work, much faster, and they'll be better placed to make the necessary visual adjustments to the output of these tools that are still in their infancy, and as such, still require a certain amount of correction.
But the other factor driving their widespread adoption at all levels of professional work will be the mind-blowing rate at which they are evolving and improving. For now, they're still in their infancy. For now, they're resolution-limited. For now, artists who adopt these tools won't experience much of a change in their day-to-day work. But how long is 'for now' realistically going to last? The results that a beginner with no art skills can generate in just a few hours with Midjourney or Stable Diffusion would have seemed laughable to most only 12 months ago.
So in regards to these realities, tweets like this don't answer anything meaningful in relation to the real issues at hand. In fact, for many of the people most affected by the topic at hand, they just raise more questions. And I think when people hold up tweets like this as some sort of attempt to say "everything will be okay!", it's not too much to expect they should have to answer some of those follow-up questions.
How long will that person be able to make meaningful contributions to the final work?
It's not inconceivable to assume an artist using AI will see their role reduced as the demand for more content at faster rates keeps up with the progress of the tech.
Most artists actually enjoy the challenge and process of creating work from nothing. At some point, motivated by both progress and demand, artists using AI tools will likely see more of the finished output of their work being taken care of by AI tools. Artists with AI image making tool skills will still be more valuable than those without, as they'll be able to make final corrections and adjustments on-spec to work that's perhaps 90% complete.
The problem is, 'picking a below average AI image and improving it' reduces their craft to what essentially amounts to retouching. Most artists, who likely developed their skills because they got something meaningful out of using their tool set, mechanical skills, trained eye and heightened sense of visual literacy, probably won't find the process of touching up work that's already 90% finished all that enjoyable, and they certainly won't find it particularly challenging.
How long will that person even be necessary?
If that role could be replaced so easily, how much job security will the newcomer even have?
I'm no futurist, so this is way out of my wheelhouse. But it can't be a stretch to imagine that one of the ideal end goals of AI tools is to be able to work with no human intervention at all. I can't be alone in thinking we've all been given the keys to the playground in part to just expedite the process of these tools not needing us anymore, right? We are training Midjourney, after all.
If you browse popular AI subreddits or discords, there's kind of a recurring theme that comes up in conversation amongst its most eager early adopters: this sense of discovering a new enthusiast, or potential professional, pathway. It's kind of like "I can generate images that are comparable to these famous and/or working artists. It won't be long until this is an entirely new role. I guess I can be part of this too now! This future is amazing!" And there's a sense of pride in their ability to wrangle with the semantics of the tools they are using, like they are explorers uncovering exciting new secrets. And a lot of this is displayed alongside feelings of "why are artists so annoyed? They should be happy this is happening, it opens up the field for everyone!"
But aren't Midjourney users, through their progressively more complex and refined prompting, essentially just making the tool easier to use for future newcomers? How will the early adopters react when these future newcomers, with little to no experience in the field, start effortlessly generating work that looks as good as those with years of experience? And won't this just be the same cycle playing out for the eager early adopters of AI tools as it once did for artists?
How much will that person be able to earn?
This kind of progressively lower and lower barrier to entry, in regards to high quality output, raises even more practical concerns, and highlights the issue with using the value system of one field to allay the concerns of practitioners in another. As I said earlier, I'm sure specialists in the AI/ML fields will still offer significant value to their employers. Regarding creative industries, in terms of individual practitioners on the execution side, I don't feel confident it will go all that well for a vast majority of individual practitioners. Particularly for junior and mid weight roles.
On the other hand, I have a feeling this is going to work out fantastically for both clients, and employers.
As it stands right now, there's a huge difference between someone with 10 years of professional experience as an artist or designer and someone with less than one year. And you know, we use that as a scale to pay them for their work. In the near future, as these tools get better, the need for unique, individual input or intervention will undoubtedly become less important.
However, as the barrier for entry and the insight required to generate high-level output with these tools continues to get lower and lower, I can see clients eagerly pushing for lower and lower rates. When someone can learn to generate good to great quality work with a few days of focused practise with SD or MJ, why would anyone want to pay them close to what they used to pay for practitioners with years of experience behind them?
How long until we start regularly replacing multiple people with one person?
Fashion has always been a very forward thinking early adopter of new tech: VR, AR, NFTs, Metaverse and all the other things I know absolutely nothing about. Fashion will undoubtedly lead the charge in adopting AI generated imagery, from editorial to flat lay to conceptual high-art imagery, and I don't think it will just be seen as a novelty. In fact, Vogue Business just published an article titled AI’s revival raises questions for fashion’s creative class, discussing AI's ability to "streamline and stimulate the creative design process". I've already seen a huge amount of "ethical fashion" Midjourney generations, and resolution aside, with the right models and prompts, Stable Diffusion does realistic depictive and detective photographic qualities to a very high standard.
The article states that many view the tech as a fantastic way to generate inspiration or pre-vizualise much more accurately. But it also points out the time and monetary savings afforded to those who use the tools. Ultimately, I feel that clients are going to dictate which of these two benefits are given higher priority over increasingly more future productions. And I'm fairly certain the vast majority of them aren't going to be swayed by inspiration.
Photographic shoots are the result of dozens of experienced and highly talented individuals: photographer(s), assistants/digi operators, models, stylists, both hair and make up artists, not to mention producers and creative directors. One of the more common talking points I've heard around this possibility usually relies on something along the lines of "they can be creative directors for the AI". Which is true, I suppose. But this goes back to my first question: how many people, who have spent years gaining both specialist technical and mechanical skills in these fields, are going to find it satisfying sitting around a monitor and cherry picking their favourite outputs from Discord?
And further to that, if one of the biggest appeals for AI tool use is its ability to save clients time and money, what client is going to want to pay for an entire team of specialists to sit around and essentially be Pinterest board curators?
We're all moving towards the same outcome
Regardless of intention, or the motivation behind its co-opting, or the difference in value systems between two professional industries, this tweet still freely admits plenty of people are going to lose their jobs. This isn't something any of us should be dismissive of. And while its intent seems to suggest a 1:1 ratio of job loss to job gain, as though that makes it okay, I would suggest that's nothing more than naive, blind optimism, with little grounding in reality. Which I suspect will be more along the lines of "hundreds of thousands will lose their jobs, but that's a sacrifice the tens of thousands who will replace them are willing to make".
Championing simplistic statements as some kind of substitute for meaningful conversation or action - as though 'being correct' is more important than simply 'being a decent human being' - is just a way to deflect from actively dealing with the very real, very difficult challenges that inspired the statement in the first place. It's essentially saying "I know something bad is about to happen. I'm not going to actually do anything about it. I'm just going to predict that it will happen." And when it does, people can point, and say "see".
When it comes to the topic of the inevitability of AI implementation in creative fields, we should expect more than just simplistic, reductive statements propelling us forward. Especially when they're laced with all the weirdly submissive trappings of techno-defeatism. If someone feels they have the authority to make such bold predictions, they should at the very least have the integrity to address the very real, very uncomfortable, livelihood threatening implications that the widespread adoption of these tools in creative industries will likely have.
Look, I get it - we're all kind of just watching this unfold and there's a very real sense that it's being dictated, in part, by people who have far more power in deciding whether this displaces entire segments of industries now, or later. And that's probably true. I don't know. These tools are going to make a very select group of people very, very, very, very wealthy, and contribute unimaginable amounts of profit to certain corporations bottom line. There are plenty of us that feel safe in our capacity to earn a living right now, and plenty of us who feel our livelihoods, and ability to work meaningfully, very very threatened. Instead of simply trying to just be right all the time, or focusing on the new avenues we think these tools have given us to explore, maybe we can ask those first in line to be impacted by them what they're most concerned about, and see if we can do something to help address it, while we still can. We're all standing in the same line anyways. It's clearly not slowing down.