QUESTION: I’m a designer, is it okay for me to use AI? 

ANSWER: Short answer? The value that genAI adds to our lives is not great enough to offset the exploitation that was core to its development, and remains core to its continued proliferation. Shorter answer? No. 

As everyone likes to say right now, AI isn’t going anywhere soon.

Unless you somehow run your whole business without the internet (in which case you wouldn’t be reading this in the first place…) it’s very likely you know what I’m about to say. Artificial Intelligence? As frustrating as it is, it’s not going anywhere anytime soon (unfortunately). I’d love to just yeet the whole concept into the sun, if I’m honest, given the entire system as it exists now is fruit of the poison tree (which we’ll talk about momentarily) but there’s very little to be done about its integration as one person, and I don’t want to get on my soapbox too much without giving you actual actionable things to think about.

So… like I said, you’ve heard it, I’ve heard it, you’ve probably said it just like I’m saying it to you: AI isn’t going anywhere anytime soon. And the idea of speeding things up, making work take less effort – I get why it’s tempting. We’re all trapped in capitalism, after all, and there’s nothing capitalism likes more than efficiency.

That doesn’t mean it’s the way forward, OR that we’re stuck with it.

But just because it’s not going anywhere anytime soon, doesn’t mean we don’t have the individual power to make choices that can help undermine, one step at a time, its hold on our industry. (Or… your industry… if you’re not a designer and you’re here anyway.) Why do I put it like that? Well, as you probably ALSO know, one of the major reasons this “tool” is being shoved down our proverbial throats every time we turn around?

It’s all about investment. And oh boy, is there investment tied up in this madness. Mountains of it. So of course it’s being deployed everywhere and anywhere. It’s the new buzz term. A few years back, you could get a “smart” washing machine, that would calculate the time your wash needed to run based on the weight of the load – now? Well… it does some calculations, and uses a computer, so the companies slap “AI” over the old “smart” label, because it’s the new flashy fancy thing that’s all the rage. Is it AI in the sense that things like Midjourney and Claude are AI? Not by a long shot. There’s a big difference between GENERATIVE AI, and AI that uses an equation to make a couple of deterministic calculations.

One of them, is pretty much a fancy calculator. The other, was built on a foundation of large-scale intellectual property theft. I think you can guess which is which.

Let’s talk about ETHICS in Artificial Intelligence.

And that brings us to ethics, which… when it comes to answering the question “should you use AI as a designer” (which I realise it seems like I’ve strayed from a bit, but trust me, we’re going on a journey together!), is paramount.

There is a massive ethical difference between Traditional AI (“tAI”, moving forward), which we have had for a very very long time, and Generative AI (“genAI”, moving forward) which is mostly what people are talking about these days, since it’s the new shiny toy you’re using when you put a prompt into Claude.

To create a baseline for what we’re going to talk about in the “AI Reality” series, I want to define all of our terms really clearly – so in the interest of clarity:

Traditional AI (tAI):

The traditional format of AI has existed for ages – we’re talking decades. A whole ton of “smart” items, like the washing machine I mentioned above, have had these kinds of evaluative algorithms built into their processes since before we thought about it as AI at all. But this is because it’s much simpler to distinguish from magic, than genAI is.

The technical setup of tAI relies on programmers creating pre-determined algorithms, that use calculations and rules to output predictable behaviour.

Crucially, these algorithms DO NOT “LEARN”. They’re, in that sense, static, until a programmer tweaks them. Examples are everywhere, from the washing machine I keep coming back to, to things like the early generations of streaming service algorithms. I can’t speak to how, for example, Spotify’s algorithm works now, but I can guarantee that at the beginning, your suggested music was based on inputs from you – music you liked, things you’d listened to on repeat, etc. Pinterest worked (and may still work) the same way – the suggested items on your feed were selected to show to you based upon your behaviour on the platform, and a set of rules designed by programmers that used things like keyword categorization to determine what else you might also like.

Obviously, these algorithms over time have become more and more complex, and taken into account more and more variables. But at the beginning, and at their core, they have been evaluative – they didn’t CREATE anything, and still don’t. They can’t hallucinate, because their behaviour is derived from set data, and cause → effect.

To oversimplify a bit, Traditional AI is pretty much: “If (input), then (behaviour)”.

• If (weight of laundry load exceeds 8kg), then (output overweight load error)
• If (user “likes” two shows tagged “horror” and “comedy”), then (output homepage “your next watch” section populated with “Comedic Horror”)
• If (user “likes” Paris Paloma album), then (output “daily playlist” using Paris Paloma tracks, and similarly tagged items)

Obviously, these are massive oversimplifications of how the variables are evaluated by tAI to create the outputs – but those “formulas” above can function as practical examples. The algorithm makes a calculation, using set data, and that calculation results in a specific output.

And that’s why I call it fancy math – because it is. They’re closed systems, that largely use platform-behaviour-specific data-points. As they’ve gotten more and more complex, it’s felt they’ve gotten more and more personal, especially in contexts like social media algorithms, that seem to know what sandwich you had for lunch and what you plan on naming your first born. But it’s all based on your behaviour, and what that behaviour indicates at the level of platform variables.

If you’re still following, buckle in, because it gets more unhinged (less hinged?) from here.

Generative AI (genAI):

Generative AI is a completely different beast. It’s still algorithmically based, but not in the same way – the algorithms that make genAI work are significantly more detailed, use vast datasets to train, and operate using probability, not determinism, like tAI. Essentially, it’s a big ole pattern recognition machine, and that’s part of why its critics regularly say things like “it’s a machine that guesses what word comes next” – because in a sense… it is. In another sense, it’s far, far more complex than that.

Because it operates on a probabilistic basis, the process of training genAI models, some of which are called “Large Language Models” (LLMs) involves steps like receiving feedback from humans, on whether it did a good job with its guess. I do, by the way, know that calling genAI output a “guess” might feel dismissive, especially if you’re someone who’s come to rely on genAI for certain aspects of your daily life – but I want to underscore that I don’t mean it in a dismissive way. Here, the word is used with neutrality, and if in reading it you feel that I’m painting genAI negatively… it might be a good idea to examine why that feeling is coming up.

But I digress – the pattern recognition machine can be trained on different types of media, from text, to images, to human voices, and you’ll have seen the results of this across the internet. One of the variety of problems with this, is since the idea is for the system to learn to both recognize and utilise patterns in a probabilistic way, it requires a TONNE of data to start it off, in the “training” stage. So in generation one, what did this mean? What’s the easiest way to get the unbelievably aggressive amount of information and media that these neural nets needed to digest?

The answer is scraping that data from the internet – which raises huge questions about intellectual property and copyrights, because initially, this was done without permission or licensing.

Now that we’re beyond the first generations of this technology, the question becomes more complex, because there are indeed models trained on ethically sourced data – however, could those exist without the initial theft-based systems? Is all Generative AI fruit of the poison vine? And this doesn’t even get into the environmental impacts, which we’ll discuss in another conversation – because it’s far-reaching, and at this point, incredibly problematic.

Because of all this and more, I personally believe that genAI cannot yet have a real place in the arts. I recognize though, that this is an ethical and moral stance, and not a truly practical one.

Okay, but what about moral perfectionism?

As this is a moral question – I would be remiss if I didn’t address moral perfectionism, and its general impossibility in our current capitalist structures. This is especially prescient because as I sit writing this, I’m using a computer, which contains rare Earth minerals, and those minerals… I don’t personally know how they were mined. I’m not certain it was by someone receiving fair pay for their work, or in a safe environment – I’m not even certain they were mined by adults. You might be reading this on a mobile phone, that suffers the same flaw. So in other words, there are very likely similar elements of theft in the development of many, many of the items we engage with on a daily basis – not necessarily theft of intellectual property, but theft of time, theft of health, and theft of personal safety.

And I recognize that using these tools and participating in the systems that have brought them into existence is contrary to the stance that we shouldn’t use genAI due to the moral and ethical quandaries, theft included, present in its development. After all, I am, right now, using a device with similar quagmires, and most of us do, every day. These devices are so integrated into our lives in the western world, managing our bank accounts, paying for purchases, tracking our calendars, and keeping us connected – sometimes it’s difficult to remember the time before those items took such powerful places in our lives.

Formulating a “good” argument

There is no “good” argument, that completely aligns with my personal ethics and values, that I can make that creates a true delineation between using an iPhone, and using chatGPT. All exploitation, regardless of scale, is wrong. Whether it’s one company underpaying mining workers in a remote location, another choosing a country with poor workers’ rights protections for their product manufacturing plant, or a genAI company choosing to scrape protected intellectual property to train their models – it’s ALL in contrast with my values.

And to add to that, in our history as humans, there are very few true advancements and developments that did not come with an aspect of exploitation of one community or another, usually underrepresented, minority communities who are easier to exploit without consequences. Numerous medical advancements and treatments that we benefit from today began with horrifically unethical practices – everything from infecting unsuspecting children with communicable diseases like Hepatitis, to performing surgeries and procedures on slaves and other individuals and groups with limited agency. As time has gone on, these atrocities have been recognized for what they were, and there are now strict protocols in place for testing in medical advancement to prevent (most of) these horrific acts in medical development.

But they happened. And we benefit from the outcomes of that exploitation – just like we benefit from the work of the individuals mining those rare Earth minerals I mentioned earlier. To bring this back around to moral perfectionism, if my life were in the balance, and the treatment for whatever was causing that had originated in an exploitative environment… I’m choosing to live. And I would guess you would too.

Which… is again in direct contrast with the whole “don’t use genAI because its origin sits squarely in the territory of exploitation” appeal.

genAI Exploitation is a matter of SCALE

The only argument I can make here, is one of scale, and I’ll openly admit it’s poorly positioned at best – but not necessarily irrelevant due to that poor positioning. The argument is this: The scale of exploitation required for genAI to develop, and continue growing, is so vast, it cannot come into balance with the value it provides.

This argument isn’t without flaws – which I’ll address momentarily. But to expound upon the argument itself, the scale of genAI exploitation is so far reaching, it’s difficult to get a true handle on where the problem begins or ends. This is especially true because the research on the impacts of, for example, Data Centres utilised to implement these technologies, is still actively developing. Meaning, we don’t actually KNOW the true scale of the impacts they will have over time.

To summarise though, a non-exhaustive list of exploited individuals and groups involved in the development and growth of genAI technologies looks a bit like this:

• Artists whose work was used in training without permission
• Writers whose work was used in training without permission
• Videographers whose work was used in training without permission
• Voice actors whose work was used in training without permission
• Journalists whose work was used in training without permission
• Designers whose work was used in training without permission
• Underrepresented communities around the globe whose land has been usurped in various ways for data centre implementation
• Larger communities within 10km of data centres, who experience environmental impacts of their operation against their consent and without their input
• Individuals employed in the arts whose jobs have been replaced and/or are in the process of being replaced
• Individuals employed in tech industries whose jobs have been replaced and/or are in the process of being replaced
• Individuals being effectively forced to use AI against their consent due to its overarching absorption into systems that daily life in modern society requires
• Individuals being effectively forced to use AI against their consent in their work roles, due to the investment large conglomerates have made betting on the technology’s success
• Individuals whose medical data has been sold and/or shared with genAI companies by governmental entities or private entities (without transparent consent)
• Individuals whose financial data has been sold and/or shared with genAI companies by governmental entities or private entities (without transparent consent)
• Individuals whose likeness has been sold and/or shared with genAI companies without full transparent disclosure and consent
• Individuals who have been led to believe that genAI is a trustworthy source of information, who have experienced poor outcomes due to this not yet being the case

As I said, this isn’t an exhaustive list, I have surely missed some groups who have been, or are being, exploited by genAI development. But oh boy, is it already long – and you may have noticed, pretty much everyone falls into at least one of these categories, especially when we consider proximity to data centres, given their unbelievably fast proliferation which is required to support the uncontrolled growth of genAI usage.

The Trolley Problem & genAI Development

In the context of exploitation in medical development, we can think about the classic philosophical “trolley problem,” variations of which are used to explore questions of the greater good. A trolley is barrelling down a track, that splits in front of you. Your hand is on the lever, controlling which direction it goes. There are people tied to each track – on one side, one person is bound to the track. On the other, a group of ten. All of the people are strangers to you. You are in control of what happens next – does the trolley run over one person? Or ten?

Many, many variations of this have been analysed by philosophers and thinkers throughout the ages – but it applies particularly well to medical ethics, because in the context of life-saving treatment development, it’s a relatively straightforward comparison (though not without its flaws, like any thought experiment). Is it ethical to risk or forfeit the life of one person, to save the lives of many? What about ten, to save the lives of ten thousand? A hundred, to save millions?

It would be easy to say “well, genAI can benefit everyone, so it is in the interest of the greater good for it to continue developing.” But genAI “benefitting everyone” in the sense that a life-saving medical development “benefits everyone” does not require the level of proliferation or investment currently occurring in the genAI space – this level of growth and financial commitment is of the style only profit-focused capitalist endeavours can produce.

Cost/Benefit Analysis, at a MASSIVE scale

The comparison would be a lot easier if these technologies weren’t being marketed aggressively for consumer use, and instead were exclusively in the hands of scientists and researchers. Obviously, it IS in their hands as well – genAI is being tested for medical use, for evaluation of large datasets in meteorological contexts, and for the development of myriad other important endeavours that ARE in the interest of humanity as a whole. But these uses don’t require the same level of data centre proliferation, or the same types of training content acquired without licensing. Therefore, it wouldn’t require the same scale of exploitation outlined above, to reach outcomes that contributed to our development as a society.

Of course, keeping these technologies exclusively in the hands of researchers and scientists couldn’t result in the same profit that marketing them directly to a massive volume of consumers can – and because we live in a capitalist greenhouse, the strategy that results in the most profits, regardless of harm, is the strategy that wins out… until regulation is put in place to attempt to adjust for that harm. Before this goes completely off point and descends into the land of “deconstructing capitalism and how we’ve landed where we are”, I’ll suffice to say that while regulation would be incredibly helpful in stifling the bleeding, since research focused on the impacts of data centres on our environment is still attempting to catch up, we cannot yet know how much damage may be difficult to reverse – or wholly permanent.

Moral perfectionism isn’t possible in 2026, without a full-scale rejection of modern life. AKA, going and living at the edge of the woods like a bog-witch, who trades frogs from passers-by for the telling of their future. We’re not going to get into the systems etc that make THAT a near impossibility, either, however, there are many. But the fact that perfection isn’t possible doesn’t mean we might as well abandon any attempts to mitigate exploitation and damage, simply because we can’t achieve an ideal of “no exploitation” and “no environmental damage”. It is, however, on the flip side of the coin, also not fair to fully reject a technology that could be genuinely transformative for humanity if placed in the right hands at the right times.

But this is where our power as consumers comes into play, which while limited given the sheer scale of genAI feature implementation by companies that we have little choice but to engage with in things like work tasks, is still on the board nonetheless.

We do not have to buy into the over-reliance on genAI for everyday tasks, or utilise it for tasks we can easily do ourselves, or through traditional means that don’t require the resource intensity of genAI processing. Unfortunately because of where we have been positioned by companies and entities in power, it’s not possible to fully remove genAI from our lives at this time. But it’s possible to push back in small ways – and because of all the above, I believe that we should.

“But if I don’t adapt… won’t I be replaced?”

This brings us to the question of adaptation, and the relative fear mongering that results in the thought process of “adapt or die”. It’s rife in conversations about genAI in our lives in general, and as artists. We’re told that it’s the new tool – like Photoshop was the new tool at one point, and cameras were the new tool at another. And if you don’t start using the new tool, you’re going to fall into the void of obsolete professionals who are too “old school” to be relevant.

Maybe a better comparison than Photoshop, is computers. When they took over, it was part of a massive cultural shift – one that similarly brought things like efficiency to police evaluation of records, and new capabilities in medical contexts that were life changing for all of us. The impacts were far reaching – like the theoretical impacts of genAI. And there are also those elements of exploitation that we touched on earlier, when it comes to manufacturing processes and materials sourcing for the creation of computer components.

Similarly, too, we are largely incapable of participating in modern life without carrying a tiny computer in our pockets – our banking runs from apps. Here in the UK, if you asked for a cheque book to keep track of your banking, you’d probably be laughed out of the branch. Even government ID verification for online management of taxes and other items require use of an authentication app to be even a TINY bit reasonable to manage the process. They’re fully integrated into our lives, in ways that would require massive cultural upheaval to change.

Which brings me to where we are now, and why we’re NOT YET at the “if you don’t adapt, you’ll be the loser” place. Because while genAI is everywhere – it’s not yet a REQUIREMENT. We’re not yet in a place where it’s the ONLY option for searching for data online, for example. We might get there eventually – and the idea of that kind of low key terrifies me, if I’m being honest with myself.

But the more we refuse its use, the more opportunity there is for this train to stop before it reaches the next station. The more we push back, before it’s been fully integrated into our lives in ways that are inextricable, the more control we have on the ways that it will be utilised moving forward.

So right now? Rejecting genAI doesn’t mean you’ll be replaced as an artist or designer – if you work for yourself. I can’t speak for the larger context in the corporate world, because that’s a very different kettle of fish. But if you are a self-employed designer, you have the power of choice right now, and I urge you very seriously to exercise it.

I’m not using AI in my processes, and I’d urge you to think hard before YOU do.

And all of this, and more, is why I do not use AI in my processes, and I would urge you to think very, very hard before you do. The “more” we’ll get to later – because as a brand strategist, one of the things I spend a lot of time evaluating, is the impact that communicative choices (including use of tAI and genAI) make on your audience’s view of your business. When artists use AI? Right now? The response overall isn’t brilliant – and that’s playing out in the large scale data we’ve got on the phenomenon, as we barrel into the future.

So before you use genAI to create the description of your product, or to generate a logo – think about the systems you’re participating in, and the sheer number of people exploited by those systems, before you click “submit” or “create” or whatever button your chosen platform uses to output generated content. There are many systems that you don’t have a real choice not to participate in, that are rife with exploitation. But they’re also now so integrated into our lives, that it’s not a real option, without negating your access to everyday life. We’re not there (yet) with genAI. You can still impact what comes next. You can still say “I will vote with my choices, and my vote is no”. You still have the chance to contribute to a future where these tools are properly regulated, with handbrakes built in – and right now, choosing to say “this task does not require this tool” is a part of that process.