Tag: creativity

Posted on by — 1 Comment

AI, Art and AI Art – Part 4

ai generated portrait of a woman in green with a green, broad-brimmed hat, with flowers

This is the last of four posts dealing with AI and AI Art. It takes a different form to the previous three. In this post, I look only at the output from AI Art apps, without regard to how it works or what issues its use might raise. The post concludes with an overall assessment of AI art and my reactions.

In part 3, I briefly described how I tested the app, and mentioned the problems experienced with the output. This post was intended to expand on that, giving example text prompts and the resulting image. However, the theory and the practice have proved very different.

I’m sure I will be returning to the topic. I will try and come here to add new inks.

Is AI just a party trick?

Testing the AI art app

To recap, my initial aim in testing the AI art app was to push it as far as possible. I was not necessarily trying to generate useable images. The prompts I wrote:

  • brought together named people, who could never have met in real life, and put them in unlikely situations.
  • were sometimes deliberately vague.
  • were written with male and female versions.
  • used a variety of ethnicities and ages.
  • used single characters, and multiple characters interacting with each other.
  • used characters in combination with props and/or animals
  • used a range of different settings.

I realised I also needed to test the capacity of the AI to generate an image to a precise brief. This is, I believe, the area where AI art is likely to have the most impact. Doing this proved much harder than I expected.

In essence, generating an attractive image with a single character does not require a complex prompt. I suspect this is already being used by self-publishers on sites like Amazon.

Creating more complex images, at least with Imagine AI, is much more difficult. There are ways around the problem, but these require use of special syntax. This takes the writing of the prompt into a form of coding for which documentation is minimal.

Talking to the AI art app

This problem of human-AI communication is not something I’ve given any real thought to, beyond fiddling with the text prompt. This paper addresses one aspect of it. From this, it became clear that the text prompt used in AI art apps, or the query used by the likes of ChatGPT are not used in their original form. The initial text (in what is termed Natural Language or NL) has to be translated into logical form first. Only then can it be turned into actions by the AI, namely the image generation, although that glosses over a huge area of other complex programming.

This is a continuously evolving area of research. As things stand, the models used have difficulty in capturing the meaning of prompt modifiers. This mirrors my own difficulties. The paper is part of the effort to allow the use of Natural Language without the need to employ special syntax or terms.

Research into HCI

The research, described in this paper, points towards six different types of prompt modifiers use in the text-to-image art community. These are:

  • subject terms,
  • image prompts,
  • style modifiers,
  • quality boosters,
  • repeating terms, and
  • magic terms.

This taxonomy reflects the community’s practical experience of prompt modifiers.

The paper’s author made extensive use of what he calls ‘grey literature’. Grey literature is materials and research produced by organizations outside of the traditional commercial or academic publishing and distribution channels. In the case of AI art, much is available from the companies developing the apps. This from Stable Diffusion and this, deal with prompt writing.

Both of them take a similar approach to preparing the text prompt. They suggest organising the content into categories, which could be mapped onto the list of prompt modifiers referred to above.

The text-to-image community

As with any sphere of interest, there seems to be a strong online community. Given the nature of this particular activity, they post their images on social media. Some of this, being tactful, is best described as ‘specialist’. Actual porn is generally locked down by the AI companies. That doesn’t stop people pushing the boundaries, of course. If you decide to explore the online world of AI art, expect lots of anime in the shape of big chested young women in flimsy clothing. From the few images I’ve seen which made it past the barrier, the files used for training the app must have included NSFW (Not Safe For Work) images. What we get is not quite Rule 34, but skates close…

https://imgs.xkcd.com/comics/rule_34.png

What else?

It’s not all improbable girls, though. The Discord server for the Imagine AI art app has a wide range of channels. These include nature, animals, architecture, food, and fashion design as well as the usual SF, horror etc. The range of work posted is quite remarkable in isolation, but in the end quite samey. Posters tend not to share the prompt alongside the image. It isn’t clear therefore if this is a shortcoming in the AI, or a reflection of the comparatively narrow interests of those using the app.

Judging by the public response to AI, it seems unlikely that many artists in other media are using it with serious intent. That too will bias users to a particular mind set. Reading between the lines of the posts on Discord, my guess is that they tend to be young and male. Again, this limited user base will affect the nature of images made.

The output from the AI app

The problems I described above have prevented me from the sort of systematic evaluation, I planned. A step by step description of the process isn’t practical. It takes too long. The highest quality model on Imagine is restricted to 100 image generations in a day, for example. I hit that barrier while testing one prompt, still without succeeding.

In addition, I did a lot of this work before I decided to write about it, so only have broad details of the prompts I. I posted many of those images on Instagram in an account I created specifically for this purpose.

https://www.instagram.com/ianbertram_ai_

Generic Prompts

I began with some generic situations, adding variations as shown in brackets at the end of each prompt. In some cases, I inserted named people into the scenario. An example:

  • A figure walking down a street (M/F and various ages, physique, ethnicities, hair style/colour, style of dress)

Capturing a likeness

I wanted to see how well the app caught the likeness of well known people. By putting them in impossible, or at least unlikely situations, this would push the app even further. An example:

  • Marilyn Monroe dancing the tango with Patrick Stewart. I also tried Humphrey Bogart, Donald Trump and Winston Churchill.

I discovered a way to blend the likenesses of two people. This enables me to create a composite which can be carried through into several images. Without that, the AI would generate a new face each time. The numbers in the example are the respective weights given to the two people in making the image. If one is much better known than the other, the results may not be predictable, but should still be consistent:

  • (Person A: 0.4) (Person B: 0.6) sitting at a cafe table.

Practical applications

I also wanted to test the possibility of using the app for illustrations such as book covers, magazine graphics etc. Examples:

  • Man in his 50s with close-cropped black hair and a beard, wearing a yellow three-piece suit, standing at a crowded bar
  • Woman in her 50s with dark hair, cut in a bob, wearing a green sweater, sitting alone at a table in a bar.
  • Building, inspired by nautilus shell, art nouveau, Gaudi, Mucha, Sagrada Familia

To really push things, I wrote prompts drawn from texts intended for other purposes. Examples:

  • Lyrics to Dylan’s Visions of Johanna
  • Extracts from the Mars Trilogy by Kim Stanley Robinson
  • T S Eliot’s The Waste Land

I tried using random phrases, drawn from the news and whatever else was around, and finally random lists of unrelated words.

Worked example

This post would become too long if I included examples of everything from the list above, which is already shortened. Instead, I will show examples from a single prompt and some of those as I develop it. The prompt is designed to create the base image for a book cover. The story relates to three young people who become emotionally entangled as a consequence of an SF event. (A novel I’m currently writing)

Initial prompt:

Young man in his 20s, white, cropped brown hair, young woman, in her 20s, mixed race, afro, young woman in her 20s, white, curly red hair

This didn’t work, the output never showed three characters, often only one. If I wasn’t trying to get a specific image, they would be fine as generic illustrations.

Shifting away from photo realism, this one might have been nice, ethnicities apart, but for one significant flaw…

Next version

In order to get three characters, I obviously needed to be more precise. So I held off on the physical details in an attempt to get the basic composition right. After lots of fiddling and tweaking, I ended up with this

(((Young white man))), 26, facing towards the camera, standing behind (((two young women))), both about 24, facing each other

The brackets are a way to add priority to those elements with strength from 1 (*) to 3 (((*))).

The image I got, wasn’t perfect, but workable and certainly the closest so far.

Refining the prompt

My next step was refining the appearance, which proved equally problematic

(((young white man))), cropped brown hair, in his 20s, facing towards the camera, standing behind (((a young black woman))), ((afro hair)), in her 20s, facing (((a young white woman, curly red hair))), (((the two women are holding hands)))

I got nowhere with this. I usually got images where the man was standing between two black girls. In one a girl was wearing a bikini for some reason. In another she was wearing strange trousers, with one leg shorts, the other full length. I also got one with the composition I wanted, but with three women.

More attempts and tweaks failed. The closest to a usable image was this, using what is called by the app a pop art style. I eventually gave up. If there is a way to generate an image with three distinct figures in it, I have yet to find it.

This section is simply a slideshow of other images generated by the AI in testing. They are in no particular order, but show some of the possibilities, in terms of image quality. If only the image could be better matched to the prompt…

Consumer interest

I relaunched my Etsy shop, to test the market, so far without success. I haven’t put a lot of effort into this, so probably not a fair test. References to sales from the shop are to the previous version. At the time of writing, no AI output has sold. This is the URL:

https://www.etsy.com/uk/shop/PanchromaticaDigital

I also noticed on Etsy, and in adverts on social media, what looks like a developing market in prompts, with offerings of up to 100 in a variety of genre. These are usually offered at a very low price. The differing syntax used by the different apps may be an issue, but I haven’t bought anything to check. I saw, too, a number offering prompts to generate NSFW images. I’m not sure how they bypass the AI companies restrictions. Imagine, at least, seems to vet both the prompt and the output.

Overall Conclusions

It’s art, Jim, but not as we know it

In Part 3, I asked ‘Is AI Art, Art?’ It’s clear that many of those in the AI art community, consider the answer to be yes. They even raise similar concerns to ‘traditional’ artists about posting on social media, the risk of theft etc. The more I look, the more I think they have a point. The art is not I believe in the image alone, but in the entire process. It is not digital painting, it is, in effect, a new art form.

Making the images, getting them to a satisfactory state, is a sort of negotiation with the AI. It requires skill and creative talent. It requires more than simple language fluency, but an analytic approach to the language which allows the components of the description to be disaggregated in a specific way. Making AI art also requires an artistic eye to evaluate and assess the images generated and to evaluate what is needed to refine that image, both in terms of the prompt and the image itself.

The State of the art

As things stand, AI art is far from being a click and go product. Paradoxically, it is that imperfection which triggers the creativity. It means users develop an addictive, game-like mind set, puzzling away at finding just the right form of words. In Part 3, I referred to Wittgenstein and his definition of games. This seemed a way into looking at the many forms taken by art. A later definition, by Bernard Suits, is “the voluntary attempt to overcome unnecessary obstacles.” This could be applied to poetry, for example, with poetic forms like the sonnet.

Writing the prompt is very similar, it needs to fit a prescribed format, with specific patterns of syntax. In this post, I wrote about breaking the creative block by working within self-imposed, arbitrary rules. The imperfect text-to-image system, as it currently stands, is, in effect, the unnecessary obstacle that triggers creative effort.

The future

It seems inevitable that the problems of Human-AI communication will be resolved. AIs will then be able to understand natural language. I don’t know if we will ever get a fully autonomous AI art program. It certainly wouldn’t be high on my To-Do list. We don’t need it. A better AI, able to understand natural language and generate art, without the effort it currently takes, would be a mixed blessing. It would, however minimally, offer an opportunity for creativity to people who, for whatever reason, don’t believe themselves to be creative. On the other hand, too many jobs and occupations have already had the creativity stripped from the by automation and computerisation. Stronger AIs are going to accelerate that process.

It’s easy to say, ‘new jobs will be created’, but those jobs usually go to a different set of people. Development of better, but still weak, AIs will be disruptive. With genuine strong AI, all bets are off. We cannot predict what will happen. It is possible that so many jobs will be engineered away by strong AI that we will be grateful for the entertainment value, alone, of deliberately weak AI art apps and games.

Posted on by — 2 Comments

AI, Art and AI Art – Part 3

AI generated image, Japanese art style showing two women walking with a tower on the horizon.

This is Part 3 of a series of linked posts considering AI. It looks at AI art from an artist’s perspective. Inevitably it raises the, probably unanswerable, question ‘what is art’.

Part 1 looked at AI in general. Some more specific issues raised by AI art are covered in Part 2. Part 4 will be about my experience with one app, including a gallery of AI-generated images.

Links to each post will be here, once the series is complete.

Is AI Art, Art?

Is the output from an AI art app, actually ‘art’? I’m not sure if that is a debate I want to enter or if there is a definitive answer. Just think of the diversity of forms which are all presented under the banner of Art. What brings together as ‘art’ the cave paintings at Lascaux, Botticelli, Michelangelo, Rembrandt, J M W Turner, David Hockney, Damian Hirst, Alexander Calder, Piet Mondrian, John Martin, Olafur Eliasson, Beryl Cooke, Pablo Picasso, Edward Hopper, Carl Andre, Kurt Schwitters and Roy Lichtenstein? Or any other random list of artists?

One way forward is suggested by the idea of family resemblances. When he considered the similar question “what is meant by a game?” the philosopher Ludwig Wittgenstein used the concept. He argued that the elements of games, such as play, rules, and competition, all fail to adequately define what games are. From this, he concluded that people apply the term game to a range of disparate human activities that bear to one another only what one might call family resemblances. By this he meant that that things which could be thought to be connected by one essential common feature may in fact be connected by a series of overlapping similarities, where no one feature is common to all of them. This approach seems as if it would work for the list above. It is possible to trace a thread of connections which eventually encompasses all of them.

Whether such a thread could be extended to include work made by AI is not clear. I don’t intend to pursue it further here, but it may yet surface as a separate blog post

See also

https://en.wikipedia.org/wiki/Art

https://www.tate.org.uk/research/tate-papers/08/nothing-but-the-real-thing-considerations-on-copies-remakes-and-replicas-in-modern-art

Thought Experiments

As I wondered how the idea of family resemblance applied to works generated via an AI app, I realised that the act of thinking about something can be as useful as actually reaching a conclusion. Asking open questions without having an answer in mind helps us tease out what things mean, what they involve, and to explore our personal boundaries. This is the approach I’m going to take here, with a series of thought experiments.

Non-human creation

Work by animals has in the past been accepted as art, notably Congo, a chimpanzee and Pockets Warhol, a Capuchin monkey. Congo, in particular, seems to have had some sense of composition and colour. He refused to add to paintings he felt complete. However, it seems that animals cannot own the copyright to their work, at least in the US.

So, is the ability of animals, non-human intelligences, to create comparable with the production of art by computers? If not, what distinguishes one from the other?

One of the criticisms, directed at AI art, is that it lacks human emotion in its creation. That seems to argue against the acceptance of work by Congo or Pockets Warhol as art.

Is it too limited, for other reasons? What about the emotional response which might be experienced by an observer? Is the emotional response to an image comparable to the response we might have to a beautiful view? In the latter case, there is no artist per se.

Alien Art

If human emotion in the creative process is the defining factor in art, can anything created by non-humans, be art in human terms? I don’t believe so. To paraphrase Arthur C Clarke – The rash assertion that man makes art in his own image is ticking like a time bomb at the foundation of the art world. Obviously, if we stick to that view, we also exclude the work made by Congo or Pockets Warhol.

We don’t know whether life exists elsewhere in the universe, let alone intelligent life. But, for the sake of our thought experiment, let’s assume aliens are here on earth and that some of them are, in their terms, artists. For our purposes, let’s also assume that these hypothetical aliens see light in more or less the same range of frequencies as humans.

Going back to Arthur C Clarke, he discussed the potential impact of alien contact on human society in Profiles of the Future, originally published in 1962. Clarke also cites Toynbee’s Study of History. From our own history, we can predict that the response to alien contact would be dramatic. If alien art became known to us, it would inevitably also have an impact.

Such art would, by definition, be beyond our experience. It would be entirely new. We would know little of the cultural context for their art. Nor would we have access to the internal mental dialogue of these alien artists. What drives them is likely to be unknowable. Our relationship with any art they make, can only be an emotional response – how it makes us feel. I suppose it could be argued that we have some common ground with primates, which helps us relate to Congo and Pockets Warhol. Lacking that common ground, would it be possible for humans to respond meaningfully to alien art?

How does your answer sit with arguments about cultural appropriation of art from other human cultures?

Animal or Human?

Closer to reality, suppose we set up a ‘blind viewing’ of work by Congo and work by other human artists.

Would our observer be able to identify which was which? On what basis? Quality or something else? If your answer is based on quality (i.e. good/bad) does that not imply art is only art if it is good? Who decides if it is good?

In case you are wondering, the image on the left is by Joan Mitchell, that on the right is by Congo.

What if Congo was still around and his work used as a dataset to train an AI. That AI then generates work using that dataset. Would our observer be able to identify which was which? What about a three-way comparison, adding AI work to the Congo/human choice above?

See also this: https://www.nytimes.com/2022/08/18/t-magazine/art-activism-social-justice.html

Untouched by human hands…

Suppose, in some AI development lab, we link up a random text generator to an AI art app. The app is set up to take the random text and to generate an image from it. Each image is then automatically projected for a defined period of time, before the process is repeated with the next generation. Beyond the setup process, there is no human intervention.

What would an observer see? I suspect that, not knowing what was going on behind the scenes, they would see some remarkable images but also much dross and repetition. Isn’t dross and repetition, though, a characteristic component of almost all human endeavours? What does it mean if an AI does the same?

Are the individual images created in this way ‘art’? Would your view change, once you knew the origin of the image?

Ask yourself – what distinguishes an image generated by a human from one of otherwise comparable quality, generated by an AI? What happens if we compare the human dross with the AI dross?

Take a step back. Is the whole set up from text generator to projection equipment and everything in between, a work of art?

Does that view change if the setup moves from the lab to an art gallery, where it is called an ‘installation’? Why? The same human being conceived the idea. (I can assure you I’m not an AI.)

What would happen if a proportion of the images were randomly drawn from real digital works. Would our observer be able to distinguish the ‘real thing’ from the AI images? On what basis would that be made? What does it mean if they can’t separate them?

Original or reproduction?

Suppose, an AI app generates an image which looks as if it might be a photograph of a physical painting. Perhaps this one.

Suppose, further, that a human artist takes that flat image and paints an exact copy down to the last splash of paint.

How would an observer, ignorant of the order of events, see the painting and the digital image? Would it be unreasonable for them to assume the painting was the original and the digital image the copy? What does that say about the idea of the original? What if the AI image was the product of the random text generator? Does your view change if the painter wrote the original text prompt?

A further twist. Suppose that the digital image file was sent instead to a very sophisticated 3D printer to create a physical object that mimicked in every way the painting made by the artist. Where is the original, then?

For a long post on the difference between an original and a reproduction, go here.

Is AI art any good?

That is a question with several aspects.

Is it good, as art?

That can only be answered at all, if you accept the output as art. On the other hand, I don’t think a definition of art based on artistic quality stands up. All it does is shift the definition elsewhere, without answering the original question.

Is it good, technically?

That is almost as hard to answer. Look at this image. Clearly the horse has far too many legs. Is that enough to say it is technically bad?

So what about this image from Salvador Dali? Mere technical adherence to reality is clearly not enough.

Is it good at doing what it claims to do?

This section is based almost entirely on my experience with one app, but from other reading I believe that experience to be typical.

The apps seem to have little difficulty in handling stylistic aspects, provided obviously that those styles form part of the training data. Generally, if you specify say 1950s comics, that’s pretty much what you get.

Other aspects are much less successful. That isn’t surprising if you consider the complexity of the task. What’s probably more surprising is that it works as often as it does.

AI has a known problem with hands, but I found other problems too. A figure would often have more than the standard quota of limbs, often bending in ways that would require a trip to A&E in real life. Faces were often a mess. Two mouths, distorted noses, oddly placed eyes all appear – even without the Picasso filter! Certain combinations of model and style seemed to work consistently better than others.

Having more than one main figure in an image, or a figure with props such as tables or musical instruments, commonly caused problems. Humans in cars, more often than not, had their heads through the windscreen – or the bonnet. Cars otherwise tended to be driverless.

In a group of figures, it is common for limbs to be missing, or to be shared between figures. A figure sitting in a chair might lose a leg or merge into the furniture. If they are holding an object, it might float, or have two hands on it, with a third one elsewhere.

How close does it get, matching the image to the prompt?

In Imagine AI, the app I have using, it is possible to set a parameter which sets fidelity to the prompt against the quality of the image. I’m not sure how fidelity and quality are related, possibly through the allocation of processing resources.

I found getting specified attributes, like gender, ethnicity etc applied to the correct figure to be surprisingly difficult. Changes in word order can result in major changes to the generated image. Sometimes even the number of figures was wrong. Where I succeeded, there was no guarantee that this would be retained in further iterations. Generally, figures in the background of a scene tended to be dressed in similar colours to the main character and to be of the same ethnicity.

Getting variations in the physique of figures seems to be simpler for males than females. It seems very easy for depictions of women to become sexualised, compared to the same prompt used for a male figure. This is presumably a function of the training data.

What about the pictorial qualities?

Despite all the caveats, I have been surprised by the quality of the output, even quasi-photographic images and once the prompt is right, certain painting styles. Some styles still seem more likely to be successful, especially with faces and hands, or involving props like tables. Even so, and probably with some post-processing, much of the output could stand against the work of commercial illustrators and graphic designers, especially at the low cost end of the market. I have already noticed AI imagery in cover design of self-published books on Amazon.

It is the mimicry of techniques like impasto which give me the greatest doubts. I suppose it is early in the development of the field, but I saw no sign of anything which tried to use the essential characteristics of digital media in ways analogous, for example, to the use of grain in film photography. I suppose it could be argued that the widespread availability of reproductions has detached things like representations of impasto from their origins. In addition, digital imagery has been around for a limited period of time compared to traditional photography.

Impact on artists and art practice

As I said in Part 2:

For the future, much depends on the direction of development. Will these apps move towards autonomy, capable of autonomous generation of images on the basis of a written brief from a client? Or will they move towards becoming a tool for use by artists and designers, supporting and extending work in ‘traditional’ media? They are not mutually exclusive, so in the end the response from the market will be decisive.

I’m not sure that I would welcome a fully autonomous art AI. It wouldn’t do anything that humans can’t already do perfectly well. I can however see value in an AI enhanced graphic tool, which would have the capacity to speed up work in areas like advertising, film and TV.

Advertising and graphic design

In situations like this, where a quick turn round is required, I can envisage an AI generating a selection of outline layouts, based on a written brief from a client. This could be refined by say selecting an element and describing the changes needed. A figure could be moved, its pose altered, clothing changed etc. Once the key elements were agreed and fixed in position, the AI could then make further refinements until the finished artwork is generated.

Obviously this process could be managed by non-artists, but would be very much enhanced if used under the direction of an artist, working as part of a team. If the changes were made during a discussion, via a touch screen and verbal instruction, the position of the artist in the team would be enhanced.

Print on Demand

Print on demand services are common. Artists upload their files to a website, which then handles the production and shipping of products made using the uploaded image. Orders are typically taken on the artist’s own website or sites like Etsy. Products typically offered range from prints to clothing to phone cases. AI could contribute at various points in the process.

At the moment, a template has to be set up by the artist for each product they want to offer, which seems a perfect use for AI, probably with fine-tuning by the artist.

Preparing the full description for each product can be a complex process, especially when SEO is taken into account. Again, an AI could take on much of the donkey work, enabling artists to spend more time in making art. It may even be possible to partly automate the production of the basic descriptive text for an image. If an image can be created from text, it should be possible to generate text from an image.

Retailing

Many department stores offer a selection of images in frames ready to hang. The images themselves are rarely very distinctive and probably available in stores across the country. It is likely that the image forms a significant part of the selling price.

Assuming the availability of an AI capable of generating images to a reasonably high resolution, I can see stores, or even framing shops, offering a custom process.

“Tell us what you want in your picture, and we will print and frame it for you.”

Artists

Many artists already work digitally. I can see how an interactive AI as described above under Advertising and Graphic Design could be used to assist. A sketch drawing could be elaborated by an AI, acting effectively as a studio assistant. This could then be refined to a finished state by the artist.

Printmakers can already use digital packages like Photoshop to prepare colour separations for printing or silk screens. It should be possible with an AI to go beyond simple CMYK separations and create multiple files which can be converted into print screens or perhaps used to make Risograph prints.

Testing the AI App

I looked at a range of apps, initially using only free versions and generally only the Android versions. Some of them were seriously hobbled by advertising or other limitations, so couldn’t be properly assessed.

Imagine Art

AI Art Generator

GenZArt

Dream by Wombo

PicSo

Artimind (Beta)

Initially, I played with a series of different prompts to get a feel for how they worked. I then tried some standard prompts on all of them. I finally settled on Imagine, and paid a subscription. I’ll be basing the rest of this post on that app. I’ll include a couple of the worst horrors from others, but I won’t say which package produced them, since in all probability there will have been significant improvements that would make my criticism unfair.

The Imagine app in use.

My aim was as much to see what went wrong, as it was to generate usable images. I wrote prompts designed to push the AI system as far as possible. The prompts brought together named people, who could never have met in real life, and put them in unlikely situations. Some were deliberately vague. Others tried out the effect of male and female versions of the same prompt, different ethnicities and ages. I wrote prompts for single characters, for multiple characters interacting with each other, and for characters with props and/or animals and in different settings. I’ve given some examples below.

Imagine has different models for the image generation engine, plus a number of preset variations or styles. This adds extra complexity, so for some prompts, I ran them with different models, holding the style constant, and vice versa.

Outputs

Obviously, it isn’t enough to talk about these apps. The only test of their capabilities is to see what they produce. Part 4 will look at a selection, good and bad, of images and offer some thoughts on prompt writing as a creative process.

Conclusions

As with AI in general, AI art raises some interesting moral and philosophical questions. They may not be so fundamental as the Trolley Problem, but they will affect the livelihood of many people and will have a significant social impact. Finding a path through those questions, as the thought experiments show, will not be easy.

Much more quickly, though, we will get apps and packages that do specific jobs. Some are there already – colourising old B&W films for example. These are likely to have significant economic impact.

Posted on by — 6 Comments

Art, AI and AI Art – Part 1

Alien lizards roller skating in a Pride march with Pride flags

Introduction to Part 1

This post began as a limited review of what has become known as AI Art. In order to do that, I had to delve deeper into AI in general. Consequently, the post has grown significantly and is now in three parts. This post, Part 1, looks at AI in general. Part 2 will look at more specific issues raised by AI art, while Part 3 will look at AI art from the perspective of an artist. Part 4 will be a gallery of AI-generated images.

Links to each post will be here, once the series is complete.

What is AI?

It has felt almost impossible to avoid the topic of AI in recent weeks. Hardly a day passed, it seems, without some pronouncement, whether promoting the next big thing or predicting the end of the world. One thing seems clear though. Whatever concerns exist, the push to develop this thing called ‘AI’ is now probably unstoppable.

Looking deeper into the topic, I soon hit problems. AI seems to mean anything, depending on the motivation of the person using it. As a descriptive term, it is almost useless without a qualifier. This isn’t unusual in public discourse. I’m not going to consider all the bad-faith uses of the term, though. I will limit myself to trying to get some clarity.

What is now called AI is different to the former idea of ‘expert systems’. Instead of depending on humans trying to input rules to be followed, AI uses the raw data to find rules and patterns. This has only become possible with the advent of so-called ‘big data’. The term refers to data that is so large, fast or complex that it’s difficult or impossible to process using traditional methods. This is why research AIs are being used in, for example, drug research or cancer screening, using data from health systems like the NHS.

Weak and Strong AI

Essentially, there are two sorts of AI, ‘weak’ and ‘strong’. For a summary of the difference, I turned to AI. Here’s what the free version of ChatGPT said:

The terms “strong AI” and “weak AI” are used to describe different levels of artificial intelligence capabilities. They refer to the extent to which a machine or computer system can simulate human-like cognitive functions and intelligence.

Weak AI (Narrow AI)

Weak AI refers to artificial intelligence systems designed and trained for a specific or narrow range of tasks. These systems excel at performing well-defined tasks but lack general cognitive abilities and consciousness. Weak AI systems are prevalent in our daily lives. They are commonly found in applications like virtual personal assistants (e.g., Siri, Alexa), recommendation systems, image recognition, and natural language processing. They can appear intelligent within their specific domain, but do not possess real understanding or awareness.

Strong AI (General AI or AGI – Artificial General Intelligence)

Strong AI refers to artificial intelligence systems with the ability to understand, learn, and apply knowledge across a wide range of tasks and domains, just like a human being. A strong AI system would have cognitive abilities that match or exceed human intelligence. This includes not only performing specific tasks but also understanding context, learning from experience, and reasoning about various subjects. Strong AI would have a level of consciousness, self-awareness, and general problem-solving capabilities similar to humans.

Despite the claims of Google engineer, Blake Lemoine, it seems clear that we are nowhere near achieving Strong AI (AGI). All apps and programs currently in use, including ChatGPT, are examples of Weak AI. They are ‘designed and trained for a specific task or a narrow range of tasks.’ In other words, they are tools, just as much as the power looms of the late 18th century.

That does not mean that the introduction of even Weak AI is without risk. Like all disruptive technologies, it will create stress as older ways of working are replaced. Jobs will be lost, new jobs will be created, almost certainly not for the same people. If steam transformed the 19th century, and computers the 20th, Weak AI is likely to trigger a wave of change that will be an order of magnitude greater. We are beginning to see some of the potential in health care. It is already aiding in diagnostics, drug discovery, personalized treatment plans, and more.

Concerns raised in AI ‘training’

The ultimate goal in training an AI is to enable it to find patterns in data. Those patterns are then used to make accurate predictions or decisions on new, formerly unseen data. This process requires huge computational resources and careful design. The vast amounts of data involved, whether in training or in day-to-day use, raises significant ethical concerns.

Ethical concerns

Examples include:

  • Biases present in training data are likely to lead to biased outcomes. There is a need for transparency and accountability in training of AI systems, but the sheer volume of data involved makes this difficult. AI can help but…
  • Privacy concerns also arise in the use of training data. Improper handling of data or breaches in AI systems could lead to identity theft, unauthorized access to sensitive information, and other security risks. AI could also be used aggressively, to identify weak points in data protection, to break in and generally to create disruption.
  • Facial recognition systems could increase security by identifying potentially suspicious behaviour but at the risk of loss of privacy. Bias in the training data could lead to inequality of treatment in dealings with some ethnicities.
  • Adoption of AI in medical contexts potentially creates risk for patients. Avoiding risk requires rigorous testing, validation, and consideration of patient safety and privacy.
  • AI-generated content, such as art, music, and writing, raises questions about creativity and originality and challenges traditional notions of copyright or even ownership of the content.
  • Self-driving vehicles may enhance convenience and safety, but ethical issues arise in, for example, resolving safety issues affecting the vehicle occupant vs pedestrians or other road users.
  • AI can be used to optimize energy consumption, monitor pollution, and improve resource management. On the other hand, the energy demands of training AI models and running large-scale AI systems could contribute to increased energy consumption if not managed properly.

Fears about AI

Expressed fears of what AI might lead to range from the relatively mundane to the apocalyptic. The more dramatic risks generally relate to AGI, not narrow AI, although the distinction is often fudged. There are serious risks, of course, even with ‘Weak’ AI.

Impact on employment and skills

Widespread use of AI could lead to mass unemployment, (although similar fears were expressed in the early years of the Industrial Revolution). This is one of the concerns of the Screenwriters Guild in their present strike. Writers wanted artificial intelligence such as ChatGPT to be used only as a tool that can help with research or facilitate script ideas, and not as a tool to replace them. The Screen Actors Guild has also expressed concern over the use of AI in their strike.

To a limited extent, they are too late. Forrest Gump, Zelig, Avatar, The Matrix, and Toy Story, all in different ways, used elements of what will need to be in any workable film AI. There has been talk for many years about using computer-generated avatars of deceased actors in live-action filming.

Military AI

Military use of AI could lead to life-and-death decisions being made by machines with no human involvement. Associated risks include the possible use of AI being used by bad actors in cyber warfare. ‘Deep Fakes’, such as those of Volodymyr Zelenskyy and Barack Obama, are only the start.

Text-based AI

I have had only limited experience with text-based AI. I have used ChatGPT to support my research into AI, but I don’t know how it would cope with longer pieces like this blog post. None of the other programs I have tried have produced anything beyond banal superficialities. Too often the output is junk. There have been reports of the AI simply making things up, which paradoxically is probably the closest it gets to originality. With the current state of development, I would not believe any publicly available AI which claimed to generate original text.

Issues for education

Even so, packages like ChatGPT seem already to be causing problems in education. Unsurprisingly, plagiarism checkers are also being deployed, although they probably use AI too! So far as I can tell, these text generators don’t provide academic references as normally used, or any link to their original sources. If developers can crack that, then there will be a real problem, not just for schools and universities.

I asked ChatGPT to list some further reading on issues around the use of AI. It gave me what looks to be a good, varied list, probably as good as anything I might find by a Google search. I asked three times, getting differently formatted results each time. Version one gave me 2 or 3 titles under a range of headings, version two gave me a list of ten titles, while version three organised the list by books, articles and online resources. There is a significant overlap, but the content varies between the three versions. None of them are properly formatted to provide references, although enough information is given to find them.

I asked again, for a list in Harvard style, and got a list of 10 books and journal articles. When asked. I got details of the requirements for Harvard Style references. A further request gave me a list of other referencing systems in use in academic circles.

Conclusions

Strong AI has been a common fear for centuries. From Pygmalion or Pinocchio to the films 2001 or AI (in itself a partial retelling of Pinocchio), similar tropes arise. I covered the same theme in my short stories, here and here.

Weak AI

The use of even weak AI raises many complex moral and philosophical questions. Some of these, such as the Trolley Problem, had been interesting ways to explore ethical issues, but now, faced with self-driving cars, they have become real. Using AI in any form of decision-making process will raise similar problems.

There is still a long way to go to get ‘intelligence.’ If it happens, I suspect it will be an accident. Eventually, though, I believe even ‘weak’ AI will be able to provide a convincing facsimile. Whether dependence on AI for critical systems is desirable is another question.

AI as a research tool

In the more limited area of text generation, ChatGPT, used as a research tool, appears to work as well as Google, without the need to trawl through pages of results. On the other hand, it is probably the later pages of a Google search which provide the equivalent experience to the chance find on the library shelves. Without careful choice of words, it seems probable that AI search tools will ignore the outliers and contrary opinions, reinforcing pre-existing biases. But then, the researcher has to actually look at that chance find on the shelf.

I asked ChatGPT to list some further reading on issues around the use of AI and to include some contrary views. This it duly did, giving 2 or 3 references under a range of sub-topics with an identified contrary view for each. As a research tool, I can see that AI will be very useful. Its use will require care, but no more care than is required in selecting and using reference material found in any other manner.

On the other hand, it seems likely that, whether we like it or not, within a few years, AI packages will be able to generate a good facsimile of original text. A good proportion of journalism is already generated in response to press releases. How long before we get AI-generated news reports created in response to AI-generated press releases about AI-developed products? All untouched by human hands…



Posted on by — 1 Comment

The chain of creative inspiration

This post looks at how long buried memories can resurface and trigger creative inspiration.

Albert Irvin

Back in December 2018, I went to an outstanding exhibition at the Royal West of England Academy in Bristol, Albert Irvin and Abstract Impressionism.

In 1959, Irvin visited an exhibition called The New American Painting at Tate, curated by MOMA New York and toured to eight European cities. It brought the boldest and best new artistic talent from across the Atlantic to London. The exhibition redefined what was possible for a generation of British artists.

For Irvin, it was an epiphany.

Albert Irvin and Abstract Expressionism will bring together works by Irvin and the major abstract expressionist artists that inspired him, including Jackson Pollock, Willem de Kooning, Robert Motherwell, Barnett Newman, Sam Francis and Adolph Gottlieb from UK collections and works by Grace Hartigan and Jack Tworkov on loan from the USA, giving a unique chance to see so many of these important artists’ works in this country.

RWA website

Gateshead and public art

I had come across Irvin before, sometime in the late 1980s. I was still working as a planner then, in Gateshead, with responsibility for the Town Centre. One area I was working on was improvement of the physical environment of the centre. One of the projects I identified was a dismal and uninviting roadway under a railway bridge. This underpass led from a housing estate on the opposite side of the railway line to the town centre.

My idea was to line the underside of the bridge with vitreous enamel panels. Many stations on the recently opened Tyne and Wear Metro system used this system. The photo, of Jesmond Station in Newcastle, shows one example, with artwork by Simon Butler from 1983. Making these involves screen printing the image onto the surface before firing. The panels were robust and resistant to damage. In an unsupervised location like this one, they would have been ideal.

Jesmond Metro Station, Newcastle

This is where Albert Irvin comes in. I had already had contact with the Regional Arts Association, (Northern Arts), through a minor involvement in Gateshead’s Public Art Programme. I approached them again for advice on a suitable artist. The artist they suggested was Irvin. Sadly, I never saw his work in the flesh, only some murky colour transparencies in a hand held viewer. Even more sadly for me, the project never secured funding and soon afterwards I moved to Wiltshire. Looking at Google Maps, it seems that the bridge has now been rebuilt with a rather pedestrian tiled wall. The link shows the view in 2021.

https://goo.gl/maps/kPfKeS8dQYo6Wj5MA

Since then, I’ve only been back to the Town Centre once, when I photographed the soon-to-be demolished multi-storey, ‘Get Carter’ car park. The housing estate had gone, only partly redeveloped, largely with the usual mixture of sheds found anywhere.

Photographs of the iconic 'Get Carter' car park in Gateshead, before demolition
Photographs of the iconic 'Get Carter' car park in Gateshead, before demolition
Photographs of the iconic 'Get Carter' car park in Gateshead, before demolition
Photographs of the iconic 'Get Carter' car park in Gateshead, before demolition
Photographs of the iconic 'Get Carter' car park in Gateshead, before demolition
Photographs of the iconic 'Get Carter' car park in Gateshead, before demolition
Photographs of the iconic 'Get Carter' car park in Gateshead, before demolition

Memories

I had forgotten much of this until I went to the show. Nothing I saw reminded me of those murky slides, but something triggered the memories, and they came flooding back. The comparison between the superb paintings and the tiny 35 mm slides, though, was dramatic. I was, I must admit, unimpressed by the slides, so the paintings in the show had a huge impact on me. I must have sat in front of Almada, for instance, for a good half an hour, mentally disentangling the layers, trying to work out the way they had been built up.

Another painting, Kestrel, looks much simpler, but that simplicity is deceptive. In this case how it was made is irrelevant, all that really matters are the colours, which simply sing out.

It was via this exhibition that I became aware of Irvin’s screen prints. I was lucky to see them, since they were only displayed up to Christmas. Printmaking seems always to have been an important part of Irvin’s creative practice. For me, they captured just how well he handled colour. It is to them, too, that I return for inspiration, in the form of the catalogue of his prints published in 2010 by Lund Humphries. The way he worked is well described in the book, and also seen in this YouTube video.

Irvin and screen printing

The two prints below are striking examples of his exuberant approach to colour and also capture some of the recurring motifs he uses – starbursts, quatrefoils and mock Chinese characters abound.

I’ll come back to all this in another post, looking at the way I think these and other influences have worked their way into my own work.

Posted on by — Leave a comment

Working with stripes

Abstract collage blue block with red strips

I’ve just made a set of digital images built around stripes. Working digitally is my fallback position when I’m prevented from doing anything else, whether by time, health or anything else. I had no great expectations for these. I picked on stripes for the same reason I first picked on the cross, and recently the fuji-like peak silhouette. They were a recognisable starting point.

When I posted them to my Instagram account, I made reference to Sean Scully and Johnnie Cooper, in particular the latter’s ‘Fractured Light: Johnnie Cooper, Collages 1992–1997‘. There are of course many other artists who use the stripe in their work. See here, from the Tate, for example, or here, from a US site.

I’ve been adding some of my recent digital images to the shop. I’m not sure if these will make it though. The outcome was interesting, although not quite what I’d been expecting. That’s not a problem, of course. I like these on screen, but I think they might need some physical texture to really come alive.

They started life as gel prints, which I cut up to make collage. You can see those here. To make the digital images, I used the scans of the collage made for the shop listings. I brought these back into Paint Shop Pro and then edited and recombined them in various ways.

I’m not sure of the next steps if I don’t offer them digitally. One option is going back to collage. The tissue I use to remove excess paint from the gel plate before printing would work well over solid blocks of colour, whether painted or collaged. It’s certainly a path worth exploring.

It also occurs to me that scans of the tissue could also be used in digital prints, taking the cycle round again.

Posted on by — Leave a comment

Make time for play

For me setting aside time for play is a key part of creativity. It’s a way to get past my inner censor. It allows me to fail. That’s important because without failure there is no measure of success.

It’s almost a month since I spent time in my studio. Initially I took a break to think, because I found myself repeating the same thing. The work looked superficially different, but the process was the same, and so less and less enjoyable. Some health issues then intervened, so my time away from the studio became even more protracted.

I’ve already blogged about making digital prints. They are where I came from as a printmaker, so an important part of my practice. Going back to them while not in the studio was still a form of play. It gave me the freedom to think about ideas of shape and form and composition without investing too much time. Or money for that matter, since decent paper is not cheap. In the end, even though I was ‘only playing’ the outcomes were very satisfying, and I ended up with two ‘suites’ of prints. One is a set of square prints which relate quite strongly to the monotypes I have been making all year. The second set are panoramic in format, but oriented vertically. I wanted to avoid any landscape references and make these wholly abstract.

Digital abstract print
Aksinto – digital abstract print

This isn’t the first time I’ve used play to generate work. Back in 2014 I made a set of what I later called Tinies. I was painting then and very bad at judging how much paint to put on my palette. Rather than waste the leftovers I took, as I realised later, what were monotypes from the palette using some heavy mixed media paper I had to hand. Later I cut these down into small squares, each about 25-30 mm on a side. My original intention was to reassemble them into a collage.

I never made any progress, although I did play around with the pieces for a while. I kept the pieces though, then later still, mounted a selection of these to fit into a 6” x 6” frame (150 mm). When I took these to an ‘art boot sale’, to my surprise they sold very well. Many were sold before I decided to number the rest into a series – Tiny 2014.

The next year I acquired a number of pieces of mount board, originally samples of different colours. I used these to make a set of collagraph plates, experimenting with materials like tile cement. Printing these allowed me to play gain, experimenting with colour combinations, trying out the effects of overprinting colours. These became Tiny 2015.

Tiny collagraph
Tiny 2019 No 12 – collagraph

Tiny 2017, was another set of ‘found images’, this time cut from failed monotypes made with oil based inks, while Tiny 2019 was a return to the small collagraph plates. So far there have been no more.

Now though, I’m itching to get back to physical printing. I find it immensely satisfying to see an image gradually emerge out of the clutter of bits of paper, stencils and general rubbish I use to make my monotype prints. How I do that will be covered in another post.

I hope though that I can still retain the freedom from the last few weeks of ‘playtime’.

Posted on by — Leave a comment

Getting past Lockdown Lethargy

No one who is in the least bit creative will be surprised if I say that the last few weeks have been a bit barren creatively. So I was pleased to finally get some work done making some larger gel prints. I’ve added some images to the portfolio page but you can also see them on Instagram alongside other work in progress. As a build up to the larger prints I have also made a large number of smaller ones around 7″ x 5″ (18cm x 13cm) which are for sale matted to fit a 10″ x 8″ frame. These are now being added to the portfolio page. These smaller prints are being sold as part of the #artistsupportpledge or to raise funds for my local foodbank. I’ve added a contact form below if you see anything you would like to know more about.

Small prints on the Portfolio page

Cross series