At work, colleagues were having fun with the new AI chatbot chat GPT .
They fed it little tasks related to what we do, like “write an explanation of how Freedom of Information works in the UK” or “create a strapline for our new website where people can look up councils’ climate action plans”.
As Communications Manager and head copywriter for the organisation, this made me feel more than a little uneasy. At the same time I’m as curious and entertained as everyone else by this tool, which seems to have come on in leaps and bounds since it was last in the public consciousness.
In my own time, I even enjoyed a surreal conversation in which it described me a comic book artist and invented (as far as I can tell) two graphic novels out of thin air.
Click the images to see them at a larger scale
Does… does AI lie?
Fun and intriguing, right? But my day job pays my mortgage, and while colleagues are kidding around with this stuff, it feels like the writing is on the wall – for them as well as for me: coders, designers, moderators are all at risk.
Now that we’ve created this potential, I don’t think there’s any going back. Who needs a brain, a person that gets tired? We’re not quite there yet, but the way the quality has leapt forward in just a few months makes me believe it will soon be pretty flawless.
There is one thing I’ve noticed lacking in text-based AI and I’m interested to know if it can ever be realised. One of our websites makes it simple for people to email their elected representatives, and the chatbot was invited to compose a letter to their MP (something that might, I guess, actually be useful for our users that have lower literacy). In this case it was about black mould on the walls of the supposed correspondent’s flat; AI turned out a well-written and factual message but… something was missing. There was no emotion behind it. Having seen thousands of the types of emails people send to their MP, I know that they often begin with the human experience of their problems: “it’s heartbreaking to see my little boy cough himself to sleep” or “The flat looks so awful I don’t feel like I can have visitors round”, etc.
Spam. That’s going to be the other massive problem as well. At the moment I clear out our email daily at work, having to open up the ones with titles that look borderline convincing because they’re in the same sort of language that our genuine correspondents might use. Moderators of the near future might find themselves wasting time writing back to bots.
So that’s the day job up the swanee – but at least I can find respite in my hobbies, eh? Or maybe not.
During a recent work webinar with other charities, another attendee happily mentioned that she’d asked AI to design a logo – she seemed surprised when I mentioned how artists and designers are feeling about that sort of usage.
Artists are scared and angry, as the same escalation happens before our eyes with AI like DALL-E and Midjourney. This technology pulls from a massive archive of work by real artists, dead and alive, to conjure up a fully realised image in a matter of seconds.
Professional illustrators are, understandably, terrified that they’ll be out of a job. For me? It’s not my livelihood: I’m more scared that there will be no reason for me to draw any more.
But then when I think more about it, I use my comics as a way of communicating, from what is inside me out into the world. Just as an AI currently seems incapable of replicating human feelings about black mould, it certainly can’t replicate me.
My artwork isn’t about trying to make a technically proficient picture (though obviously the better I become at drawing the more pleased I am with my work); it’s the communication of an idea or ideas.
Here’s an AI generated comic:
This does not spark joy. Admittedly, I didn’t spend much time refining my prompt.
AI also, at this point in time, can’t reliably draw hands. It seems to ‘know’ they have fingers, just not when to stop adding them or how they might be used. The sausage fingers are a particularly amusing feature of this stage of AI and I would be delighted if it just kept getting better and better but was never able to overcome that one flaw – how perfect would that be?
I see artists trying to fight back against the AI generators with a legal argument that they did not consent for their work to be fed into the big soup pot from which they take the elements of their outputs.
It must be especially galling if you’re an artist with a distinct aesthetic which anyone can summon up by typing “in the style of”, but I suspect this argument, which plays to people’s sympathies, is not going to work – it’s too technical, too small an objection in the face of the enormous juggernaut of progress, and money.
Some, enterprisingly, have been feeding in images from Disney – that most litigious of companies – to see if they’ll bring the might of their legal team to the table. It will be interesting to see if this causes any blip in the AI’s progress – but overall I doubt it. This is a Pandora’s box that once opened cannot be shut. If it can be done it will – and now here we are.
So – stop worrying and embrace it? The first AI comics have been created – of course Dave McKean (whose work was already so reliant on digital collage) was quick to get onto that. You might have seen the uproar from the illustrator community when someone proudly posted a picture book on Twitter.
One response went like this:
Of course artists are quick to carp, and it’s understandable that they’d point out errors – but are they just educating the next generation of AI, providing user feedback that will inform future development?
For me, I’m reaching a conclusion that AI is going to transform my paid work – who would employ someone to spin out a blog post about their work when it can be done by a machine so easily? – and the work of practically anyone who works in the creative industries, including web design, development, copywriting, book writing, illustration. We will be getting a taste of what blacksmiths had to endure when the motor car took over from horse driven travel, or skilled typists experienced when the word processor became common. One would love to think the result might be liberation from the current structures of a five day work week but weary experience tells us it’s unlikely. Perhaps it’s time to be retraining in something that machines will never be able to do – once I’ve figured out what that is.
But as I say, when it comes to artwork, there’s no AI that can express my feelings. OK, it can produce a great script. I bet if I said ‘give me a script for a comic about the feelings and concerns of a 54 year old woman in Brighton UK’ it would come out with something uncannily convincing. But they wouldn’t be the thoughts and concerns that come out of the unique set of synapses and chemical compounds that make up Myf. Only my synapses and chemical compounds can do that.
I could say ‘give me a set of stories of protest from the Rhondda Valley’ and it might be able to come up with true life examples – but they wouldn’t be the ones I discovered when visiting the Rhondda, talking to whichever real people I happened to meet, through chance, about their lived experiences. Truth has to count for something, right?
Of course, the script is only half the work of making a comic, and once one has generated the text you could imagine feeding it into an image generator to make much quicker or more adept work in whatever style one desired, than the sad old method of being hunched over a desk for hours, honing one’s craft over a period of decades.
I’m less sure about this. It’ll never do quite what I do. In some ways I feel a bit like an unpredictable black box of a machine myself – I tell myself what I want to draw and am always slightly surprised by what emerges. My drawing is a set of habits, shapes – faces often look alike (and surprisingly similar to ones I was doodling in my teens) if I’m not putting effort in; I like certain types of line, etc.
I’m not saying that my own style is one that people would necessarily want to replicate but they certainly could. The big question for me then is: with AI at my fingertips, would I be asking it to make work that looks like mine, or work that looks like mine would be if I’d studied drawing harder and spent more of my life practising?
I suppose the future will come, and then I’ll know.
PS: of course I couldn’t resist giving those prompts to ChatGPT. The results are almost the opposite of what I expected – the first is unsophisticated and without much value (but could probably be improved greatly with better prompts) and the second would actually provide new avenues of research to follow up on.
Prompt: Give me a script for a comic about the feelings and concerns of a 54 year old woman in Brighton UK’
Response: Sure, here is a script for a comic about the feelings and concerns of a 54 year old woman in Brighton, UK:
Panel 1: 54 year old woman sitting at her kitchen table, staring at her computer with a worried expression on her face.
Caption: “I can’t believe I’m turning 55 next week. Where did the time go?”
Panel 2: Flashback to the woman in her 20s, standing on a beach with a group of friends, laughing and having a great time.
Caption: “I remember when I was young and carefree. Life was so easy back then.”
Panel 3: Back to the present, the woman is now standing in front of a mirror, frowning at her reflection.
Caption: “Now I’ve got wrinkles, grey hair, and my metabolism has slowed down. I don’t feel like myself anymore.”
Panel 4: The woman is now sitting on a couch, talking to a friend on the phone.
Caption: “I’m so worried about the future. What if I can’t afford to retire? What if I get sick and can’t afford medical bills?”
Panel 5: The woman is now outside, walking through a park.
Caption: “But then I remind myself that I’ve still got a lot of life left to live. I can’t let my fears hold me back.”
Panel 6: The woman is now back at her kitchen table, smiling and looking at the computer with determination.
Caption: “I may be 54, but I’ve still got a lot to offer. I’m going to make the most of every day and not let my age hold me back.”
Prompt: Give me a set of stories of protest from the Rhondda Valley
Here are five stories of protest from the Rhondda Valley:
- The Rhondda Uprising of 1910: This was a series of strikes and protests that took place in the Rhondda Valley in South Wales, in response to poor working conditions and low wages in the coal mines. The protests were led by the South Wales Miners’ Federation, and resulted in several concessions being made by the coal owners.
- The Tonypandy Riots of 1910: These riots were sparked by a dispute between coal miners and the mine owners in the Rhondda Valley, and resulted in a series of violent clashes between miners and the police. The riots were eventually quelled, but not before several people were injured and much property was damaged.
- The Rhondda Valleys Hunger Marches of the 1930s: During the Great Depression, many people in the Rhondda Valley were struggling to make ends meet. In response, a series of hunger marches were organized, with protestors walking from the Rhondda Valley to London to demand better working conditions and higher wages.
- The 1984-85 UK miners’ strike: This was a nationwide strike by coal miners in protest of the closure of many coal mines in the UK. The strike had a particularly strong impact in the Rhondda Valley, where many miners were involved in the protests.
- The anti-war protests of 2003: In the lead-up to the Iraq War, a series of anti-war protests took place in the Rhondda Valley, with people demonstrating against the planned invasion. Many of these protests were organized by local peace groups and trade unions.
2 thoughts on “2022 in review: something to worry about”
I understand your concerns. I think people are having fun for now with something that can easily go out of control. Sigh.
Neither of those results are what we call ‘stories’: they are assemblages of information that match the pattern you might have previously encountered such information in before. Which is why the rhondda results are more useful, they have no viewpoints or observations in them. Just datapoints. And very surface-level at that.
I think anything beyond this one-dimensional presentation is a long way off yet. Or the prompts would have to be so much longer and more detailed that might negate the value of using AI at all?
Comments are closed.