 |
| A Really Simple Guide to AI for Artists |
With all of the buzz around
Artificial Intelligence, (AI), I decided to see if it really could solve the
time problem that many artists face. Could it really provide answers to the
most searched for answers that artists seek, such as “what art sells best”,
“how do I run an art business” and those two all-consuming questions that so
many new artists ask, how do I sell more art, and how do I price my work. Can
AI really take the place of an experienced artist who has lived the life and
borne the scars of such a complex industry?
Could I use AI to fill a
website with content? Could I use AI to make me a millionaire, so many
questions, but the answers AI has given me since it has become more readily
available and accessible to the public have often surprised me, and not always
in a good way.
I have used ChatGPT which seems
to be the “thing” of the moment, for a while now and others for a few years
before so it isn’t my first rodeo with artificial intelligence. I have
interacted with plenty of humans whose intelligence could be classified as
artificial, and I’ve been using this technology in various forms since the days
of Microsoft’s Clippy, remember the super-smart and oh so annoying animated
paperclip from the 90s that always knew better? Clippy is a perfect example of
where we were pre-millennium as opposed to where we are now.
AI isn’t new, what’s sort of
new is just how accessible it has become, and more than that, just how mind
blowing it is that you can interact with real AI from home. There have been
many iterations of publicly accessible AI platforms like ChatGPT over the years,
the more recent ones are the ones that really are beginning to move the needle
so these platforms are what I term as being, new-age AI. Older iterations were more akin to simple chat
bots or were so convoluted to use that it just wasn’t worth it.
 |
| FMV Baby! One of my latest creations celebrating the emergence of Full Motion Video in the 90s. Made possible by the development of relatively inexpensive graphics cards and the accessibility of PCs. |
Before we begin it’s probably
worth me quantifying any thoughts I have about AI up front. I come from a
long-time background of technology and have been involved in AI since what
seems like forever ago. In the 80s I was writing very simple AI in my computer
game code to detect collisions in simple games on 8-bit home computers, in the
90s I was doing the same with 16-bit computers. The difference between 8 and 16
bit computing power was compelling, the difference between the AI, not so much.
Today, AI is way too
complicated to even contemplate building it from scratch as we did in the 80s
and 90s, instead programmers mainly rely on pre-built engines to do the heavy work,
just as they rely on pre-built engines to create scenery and even video games.
It’s not that programmers today aren’t as talented as they were, the creations
they program have become so huge that it’s no longer in the realms of
possibility for a single programmer or even a team to take this kind of work
on.
When the old computers began to
be replaced with PCs, we didn’t see anything of note on the AI front for most
people for a very long time. It was more or less an out of reach science and
the stuff of science fiction for what seemed like an eternity, but things were
happening behind the scenes. Analysts were analysing, data scientists were
doing whatever it is they do, and then there were the lab folks, mostly wearing
big glasses and wild hair, who were tinkering and plotting that the machine
would one day inherit the earth.
AI in the 80s was basic,
sometimes even written in BASIC (A programming language that remains the
bedrock of coding today, except we stopped teaching it in schools long ago
because humans like the awe of oh shiny new, even when it’s not as good as old)
and then about five years ago we began to witness a sort of seismic shift in
how we thought about AI and how it could benefit those who could access it.
Suddenly, the algorithm was
placed on a plinth for the minions to worship as if it were some kind of God. It
began to mature at a faster rate, it was harnessed by big tech, the pandemic
raised its profile in medical research, and we have slaved for it since from our
interactions on social media to our frustration in getting Alexa to answer us
even after the third attempt. Don’t get me started on all of those chatbots and
their endless mind-loops.
AI has become smarter, but it
hasn’t done that on its own. AI relies on humans to code the learning and the
output. It’s not like a film where AI breeds and then takes over the world,
there’s very much a need for humans to sit behind the wheel and drive it, or at
least there is for now. It’s smart for sure, or maybe, clever might be a more
appropriate word, but the humans behind AI are mostly smarter, albeit slower, the
real problems begin with the end users who want to use it to create their very
own versions of chaos.
I’m not convinced that AI is
the problem that it is often framed to be, it gets a bad press for sure and
there are few who truly understand it, it’s technically beyond anything we
could have imagined in the early days of computers, or anything that we might
have imagined even just a decade ago.
The problem with AI rests in the hands of humanity and that’s what also
makes the entire concept of AI in public hands really quite terrifying.
 |
| 90s Digital Camera by Mark Taylor – Hand drawn using a digital medium, this was another technically challenging piece and took around 70-hours to complete. |
Machine learning isn’t new. The
concept of machine learning has been around for several decades, but the modern
era of machine learning began in the 1950s and 1960s when computer scientists
began to develop algorithms and models that could learn from data.
Prior to this, and to keep
things really simple because the world is complicated enough, algorithms were
very much a manual thing. Humans would build up a library of problem statements
with a list of viable options and then note the ones that worked. In short, AI
was a bunch of problems in an almost historical timeline so that you could see
what might work and humans then worked it out, often very slowly.
The difference between then
and now is that machines can make that comparison much faster and with more
accuracy than humans ever could. You could say that at one time AI was simply
called intelligence, but then humans decided to automate it with a computer.
Humans it seems are nowhere near as efficient as computers. Why do all that
math and cross checking when something about the size of your cell phone can make
a better decision far quicker.
One of the earliest examples
of machine learning was the development of the Perceptron algorithm by Frank
Rosenblatt in 1957, which was a simple algorithm that could learn to classify
images based on their features. That’s the kind of AI technology that we now
see with licence plate and facial recognition in crowds, the difference is that
Rosenblatt’s algorithm has been massively scaled and in ways that I’m sure
Rosenblatt would never have considered.
Another important milestone in the history of
machine learning was the development of decision trees by Arthur Samuel in
1959, which were used to make decisions based on a set of rules learned from
data. This underpinned AI for many years, but the machines are now programmed
to learn everything.
Since then, machine learning
has evolved and expanded rapidly, with breakthroughs in areas such as neural
networks, deep learning, and reinforcement learning, which have enabled
machines to learn more complex and sophisticated patterns in data. Today,
machine learning is widely used in a range of applications, from image
recognition and natural language processing to fraud detection, healthcare and
personalised marketing.
For the most part, it’s
difficult to comprehend just how much AI already encroaches into our lives
behind the scenes. It’s something we use but we’re not always aware of it, and
that is how it’s supposed to be, to provide us with a seamless experience when
we interact with technology.
Algorithms and machine
learning used in artificial intelligence models are used to provide you with a
personalised music playlist or to surface the next film that you might want to
watch on Netflix, and AI is extensively used in medical research and healthcare
where it has already saved countless lives, and of course it drives those
social media algorithms that in turn drive us all crazy. Without it, it
wouldn’t just be humans that would be less efficient, the infrastructure that
essentially drives the world we live in would be too. We might not know it, but
we have already become addicted to AI.
So AI really does play an
important role in the fabric of our everyday lives whether we realise we depend
on it or not. Now AI has exposed itself to the world on platforms such as
ChatGPT and the forthcoming Bard from Google, essentially putting the ability
to drive it, firmly in the hands of everyone, good, bad or indifferent. That’s
where it becomes infinitely more useful in our everyday lives, but that’s also
where it becomes exponentially more concerning at the same time.
 |
| Endless Path by Mark Taylor – another abstract inspired by the iconography of the 80s and 90s. |
If the bad actors decide to
grasp it with both hands, and they’re already reaching out in droves to do just
that, then we begin to have major issues in telling the difference between
what’s real and what’s not. If we break that down further, what we will see,
and there is very little doubt that we won’t, is that public access to these AI
platforms will become increasingly weaponised. If we step back and consciously
think about that for a moment we have already seen this happen through
democratic processes, social media and even government and state propaganda, and
when machines can do the work of thousands of people to weaponise content, it
becomes massively more cost effective to fill the internet with bias,
prejudice, and intolerance or just outright lies.
The world is already blighted
by scammers but generally, the more elaborate and most successful scams are
expensive operations to run. Mostly the scammers have a couple of options when
it comes to funding their operations, either they will have an army of social
media trolls writing copy which is often grammatically incorrect making it
slightly easier to spot, but even when paying below minimum wage this is an
expensive option. Or, they can now ask an AI platform to spill out copious
amounts of copy saying the same thing in many, many different ways to give it a
more authentic and authoritative feel.
You don’t necessarily need an
army of people at this point, just a person, a laptop and some intent. People
are expensive and for things like this, there’s not much in the world of
enforcement via other AI instances that can easily police this. It’s also
something that causes some contention within social media circles, this type of
content drives division and engagement and those are the two very things that
also drive ad-revenue so you really have to ask a very simple question around
whether or not social media companies really, on a deeper level, want to do
anything over and above what they already do to stop it.
Cynical me might suggest that
it would make zero financial sense to counter the negative aspects of AI driven
content on social platforms, and there’s little doubt it would be both complex
and expensive, but longer-term, I’m confident that tackling it now would make
for a more sustainable future. If they don’t begin to tackle it better than
they are, I’m reasonably confident that the blunt force of regulation might
come into play and I’m not convinced that’s entirely healthy for any of us either.
Whether it’s convenient for
social platforms to not accelerate the policing of AI generated content is a
question for another day, maybe another blog by another author, but until there
is a mechanism to fund and monetise mainstream social media access more
effectively than ads, and a mechanism of funding that’s adopted by the many,
well, it does make one wonder if the real issue around the toxicity of social
media will ever be reduced whilst the price of entry is free.
We are now at a point in time
where it’s increasingly and often impossible to tell the difference between
good information and bad. AI is so much more efficient and does a more
consistent job of writing copy than an army of people can and of course, like I
said, more importantly it’s nowhere near as expensive.
Add to that the role of AI in
generating new users and we take a leap into a world where it becomes a
perpetual cycle where accounts are shuttered and within five minutes a hundred
more are created. Social media’s problem is that they didn’t start charging
people from the off, and that’s going to be challenging for them to now
overcome.
 |
| Industrial by Mark Taylor – another new geometric abstract inspired by the brutalism style of architecture which was predominant during the 70s and 80s. |
Whilst there are some
safeguards in place with these platforms, it’s usually the job of other AI
models to sort out the good and bad. Chat GPT for example will remember
conversations that people have with it, it allows a user to provide follow up
corrections (as in change the output) and it is trained to decline
inappropriate requests, but from experience it’s how you generally ask what you
ask where this particular safety net can live or die.
ChatGPT specifically calls out
some of its limitations as in, it may occasionally generate incorrect
information, it may occasionally produce harmful or biased content, and it has
limited knowledge of the world and events after 2021. But AI models are
designed to always learn so it’s only a matter of time really as to how quickly
some of these limitations will be overcome.
Bear in mind, AI is fallible,
it’s also often biased, AI only knows what it knows, and in the case of
ChatGPT, barely nothing about life post 2021, and sometimes, well, it kind of really
does make things up. Is it sentient you might ask, well I asked that very
question of ChatGPT and honestly, I think it lied when it told me it wasn’t.
But these platforms and
ChatGPT more specifically, can be useful as a business tool, equally they can
also be problematic, so let’s see if we can figure out what it is good for and
more importantly, will it really help us to sell more art?
Well, here’s the first
elephant in the room that we probably need to look at, can AI produce original
art. The answer cuts far deeper than the question implies. ChatGPT can’t, it’s
only good for text based output, but there are neural network engines that are
driven by AI models that are really super-efficient at creating images and
there are AI models that can create deep fakes, including perfect photographs
of people who don’t actually exist.
While AI-generated art has the
potential to revolutionise the creative industry, there are several ethical,
technical, and philosophical issues associated with its use. Here are some of
the most significant:
One
of the primary issues with AI-generated art is that it raises questions about
originality and creativity. Critics argue that machines are incapable of true
creativity because they rely on pre-programmed algorithms and data inputs. As a
result, some people question whether AI-generated art can ever be considered
truly original.
Another issue with AI-generated art is that it may perpetuate existing biases
and lack of diversity. Since the algorithms are often trained on historical
data, they can inadvertently replicate and even amplify existing prejudices and
inequalities. This can result in AI-generated art that is sexist, racist, or
discriminatory in other ways.
There is also an ongoing debate about who owns the intellectual property rights
to AI-generated art. Since the art is created by machines, it’s not always
clear who should be credited as the creator nor is it clear from the output who
owns the copyright but generally it is without any shadow of doubt, using your
art and my art and the art of millions of other artists to do its thing. It’s
possible that if you have ever uploaded a piece of your own artwork to an
online space that an AI model has already learnt from it.
Some
critics argue that AI-generated art is not really art because it lacks the
human touch. They believe that true art requires human emotion, intuition, and
expression, and that machines can never fully replicate these qualities. The
counter to that could be that it is using your art and my art and the art of millions
of other artists to do its thing and there are plenty of emotions right there.
Finally, there are concerns that AI-generated art could
lead to mass production and replication, which could devalue the art and reduce
its uniqueness and cultural significance. This could lead to a situation where
art is created solely for commercial purposes, rather than for its intrinsic
value as a form of creative expression. Flooding a market only ever serves to
water it down.
As a
creator of art, as far as AI goes, and as much as I live for technology, I’m
not a fan. I was never really a fan of the neural network art that was being
generated about five or six years ago because very quickly we were seeing
pretty much the same art over and over. There were a number of apps that sprung
up on the various App Stores and when lots of people figured that art (can we
even call it that?) could be created with a single click of a pixelated button,
lots of people suddenly thought of themselves as artists.
The markets were flooded at
the time with similar images from the same app and virtually everyone who had
little to no previous experience in creating artworks thought that what had
been created was somehow unique to them. Many of those apps used very confined
and linear AI models so the results were almost always generated from one of a
small number of presets.
But, it really comes down to
the single bottom line that however AI is presented, it takes our art and does
something with it that we don’t particularly want it to do, and at this point,
it could be argued that this will be the time to consider an alternative career
outside of creating art because I’m not even remotely convinced there will be a
way to monetise your work if AI continues to take it off us without permission
or recompense.
 |
| Polybius Redux by Mark Taylor – the second work to be inspired by the mythical Polybius arcade game which was said to be released in a video arcade in Orgon sometime in 1982. |
I don’t think that any of us
will have the option to ever ignore AI and I’m confident that it will become
significantly smarter over the next decade. Its misuse will become an even
greater problem so the hope is that there are plenty of smart folks trying to
figure out ways to minimise the risk of this happening, but, I fear that even
with it’s limitations today it may already be too late. The juggernaut is well
and truly coming down the road.
Like I said just now, I’m not
a big fan of AI in art, but I have to admit I am a fan of some of the benefits
that platforms such as ChatGPT can bring in terms of efficiencies that we all
need to find when attempting to run a small business.
When AI plays a role in
Generative art and Fractal art, there is a significant difference in the generated
output in that the art is created using algorithms and fractals, rather than
from taking the work of others. This kind of art can be unique, especially if
the artist creates their own algorithms.
But whichever way you cut it,
there are plenty of unanswered questions that I don’t think we’re even close to
answering anytime soon. How will we prevent certain populations from being
discriminated and marginalised by AI, how can we get AI to take human context
into account, can we ever get to the point where we are able to teach a machine
empathy, compassion or to be ethical, and maybe one that is often forgotten,
how we address the balance of access to available data sets?
Just like our art gets taken by
some AI systems to generate a work, our data which is collected by the tech giants,
big corporations, governments and advertisers can be used to train these models
and we don’t always have a say whether our data should be used in this way.
Either way, that’s the kind of data that has a real commercial value to those
organisations which puts them at a huge advantage.
Another question that seems to
be challenging the overarching concept of data driven AI modelling is how do we
even begin to control or at least rein in what is seen by many people as an
invasion of privacy. My fear beyond using the data in AI models is that the
clamber for personal data exponentially increases beyond where we are even
today.
The huge amounts of what is
essentially surveillance data that links us to people, places, times, and what
we look at online, is one of modern histories greatest mass surveillance programs
that largely gets ignored for the most part by the masses, but this could
eventually lead us on to a path of social oppression.
Before AI takes a more
complete stranglehold of our everyday lives then I think we also need to start
answering some of those really difficult questions, especially as the models
evolve, the systems further develop and we become even more addicted to its perceived
benefits.
I asked ChatGPT a number of
questions about running a successful art business and the responses could have
appeared on the pages of many of the blogs that like the one I have created
here, offer hopefully, useful advice to artists on running a business and
raising your profile as an artist. What I came away with though was plenty of
generic answers without any real context.
When I write I like to provide
the context and back the answers up with the scars of experience of life as an
artist in what has, especially over the past decade become a very complex
industry. ChatGPT identified some key areas to focus on when I asked it what I
would need to consider when creating art professionally. It covered quality of
work, making sure you have business and marketing skills, it also covered
personal branding and it focussed pretty heavily on making sure that you
develop networks and relationships. It made no mention of making sure you were
on top of your taxes, and just like the rest of us in the UK, it gave up trying
to figure out post-Brexit import and export paperwork.
The advice was almost there, it
just felt clinical and made everything sound way easier than it is in real life.
It offered no real world examples and certainly didn’t proffer any kind of
guidance as to how an artist might take the journey from A to B.
It did suggest that
professional artists need to be persistent and resilient, so at least AI sort
of recognises that the art world is a challenge, yet despite this being its
strongest response, it was still bereft of any practical guidance that would
guide me through some of the challenges that professional artists come across
every day. That said, it may be the way I asked the question and we’ll come to
that a little later.
With some answers in the bag I
was keen to find out more. I asked a question that I thought might be more
relevant during the current economic climate, how do I produce art more cost
effectively. I thought this was going to make it struggle with a response, but
I have to say that given the rising prices of art supplies I was eager to see
something new that I hadn’t already thought of.
The answers were once again,
relatively generic. Use recycled materials, use alternative surfaces, create
DIY art supplies, and one that I can categorically say without any shadow of
doubt is no longer cost effective, try digital art. As I mentioned a few weeks
back, the cost of creating digital art has massively increased over the past
year or so which tells me that it’s responses are at times out of date.
It also suggested that I join
an Art Co-Op, and I’ve mentioned this many times over the years on these very
pages. Where it fell a little flat was in underestimating just how many
communities have these in place already, and it woefully underestimated just
how many professional artists want to come out of their silos to work with
others, that’s always been a problem in recent art history at least.
 |
| Bootleg Gotchi by Mark Taylor – inspired by the many knock-offs of the infamous Tamagotchi toy that became popular in the 90s. This was hand drawn with the detail around the edges of the plastic wrapper taking just over 11 hours to create, but I think it looks realistic! |
Let’s be honest, this is
probably the single most asked question ever, it’s the one question that new
artists and long-time artists alike want to know, as if there is some kind of
golden ratio for pricing that none of us have ever found. Surprise, AI didn’t
find the golden ratio for pricing either.
It suggested I consider the
cost of materials, my time, the size and medium, market demand and competition,
along with artist reputation and experience. All the stuff we already know but
then tend to do nothing about, instead we revert to making a price up and spray
and pray the work all over social media.
The advice here was solid, it
provided all of the data points you need to find the data on, although still
lacking any real context, but I think that this is a question that every artist
asks in the expectation that someone will confirm that their work is worth a
million bucks per square inch. Look, even AI can’t figure this one out on its
own so I think the only remaining option is to do exactly what it tells us to
do and go figure that bit out for ourselves using the very data points it told
us to explore.
Where AI will have an impact
here is in making sense of all of the data that you collect so that you can
confidently work out how much to charge, but if you’re after a straight answer,
you’re not going to find it here.
I had to ask didn’t I? Well,
here’s the response:
Making one million dollars
from creating art can be a challenging task, as it requires a combination of
talent, dedication, business acumen, and a bit of luck.
Again, the answer was somewhat
disappointing. My hopes and dreams were doomed once again. A bit of luck seemed
a little on the optimistic side but it did suggest that any artist with a goal of
making a million from their work might also want to consider diversifying
beyond creating art.
Wise words indeed but there
was no real context around what the diversification ought to be. I’m sure it
could have been pressed further than I pushed it to provide better answers and
again, I’ll touch on that in a moment. If you are wondering whether it is
possible to make a million dollars from creating art, then yes, it is, but the
probability of doing that quickly is very unlikely.
How do I sell more art, and, What
kind of art sells best. The answers were once again fairly generic, but
so was the question in all fairness. I got the usual stock response to, how do
I make more sales, the answers pointed me towards covering social media,
gaining representation from a gallery,
and to network with other artists, and from experience, these tend to be
the bare basics that you need to have in place. It failed to mention that
market research, business acumen, and an entrepreneurial attitude were perhaps
more critical than even artistic talent. I’ve said it before, even bad art
sells with great marketing.
As for what kind of art sells
best, figurative, landscape, abstract, street art and pop art were the answers,
but bear in mind that these would be the same answers that anyone could give
after spending a few minutes looking through online galleries. That’s because
the answer you will get from this question is all to do with the numbers, there
are simply more artists creating art within these genres so the volume of sales
will be higher overall. You could say that those subjects are collective best
sellers across an industry, I’ll never be convinced that they’re best sellers
for an individual artist.
If ten artists create
landscapes and sell a hundred works between them, the overall numbers look good
for those who sold more than other artists, but the numbers wouldn’t be as good
as the one artist working in a niche who sells twenty or a hundred works. The
subjects are popular because there’s a wide choice, not necessarily because
these genres are financially the best to create. Again, this is why you will
always need experience working alongside AI.
 |
| Video 2000 by Mark Taylor – this VCR was released in the UK and Europe but it didn’t set the world alight. It was intended to compete with VHS and had some great features like double sided recording on the video cassettes and the tape was completely internal. Possibly the most technically challenging retro inspired work I’ve created over the past several years, in part because there’s not a lot of reference machines available! |
The example I’ve been using to
show you some of AIs capability is ChatGPT, it’s very much a work in progress
but it is about as accessible as the public can get to using what can be an
extremely capable AI tool, and it’s also the closest that the public can get to
using a completely incapable AI tool.
When I say incapable, that
might make it sound as if ChatGPT or AI in general is just a novelty that’s
unable to be used for anything serious and that’s definitely not the case. What
makes it incapable isn’t necessarily the technology, it’s that to get anything
useful out of it the user has to be just as capable in framing their requests.
The user also has to be very specific in what they want from it, more than
that, you have to be very, very, very, specific and you might need to reframe
any follow up questions to push it even further so that you get better and
sometimes more credible answers.
Can a tool such as ChatGPT
write an article for a website like this, yes, but it’s not as entirely as
clear cut as that. I personally wouldn’t rely on it as an automated tool to
provide me with the kind of words I spend hours pouring over every time I
publish a new article, simply because readers who read what I write visit this
site to read what “I” write. I hope…
Without some real input from a
human, it’s not overly great at providing an article with plenty of contextual
examples. For it to do that you would need to take the output and then break it
down even further by asking a series of further questions and then, you might
be somewhere a little closer to something that’s a better read but it’s worth
remembering that AI will never be you, well at least not just yet.
As it is, without spending
just as many hours on refining the output as you would to write an article in
the first place, this kind of more generalised, publicly accessible AI is too
general and much too vague in its responses to give you something that you can completely
rely on to give you something that is close to being print ready. You have to
be careful not to rely on it to provide you with facts, some of the output will
be out of date, and it can generate responses that while factual, can be
misleading without any accompanying context.
As a tool, AI in a ChatGPT
sense shouldn’t be feared, as long as it is used responsibly. It can be really
beneficial in saving you a heap of time if you spend a little time figuring out
how to work with it. You have to look at it as a slightly smarter version of
the virtual assistant that you might use to turn up your heating, ask for a
weather report or turn on the lights, but give it time to mature and I’m
confident that these platforms will become increasingly more complex, and
increasingly more simple to use.
AI more broadly beyond the kind
if AI as found within platforms such as ChatGPT is much more refined because it
is developed with one very specific output in mind. If it’s used for scanning
medical data then the models used to provide the machine learning are going to
be much more specific and detailed than the more general learning that’s been
applied to bring you something like ChatGPT.
That’s not intended to be
detrimental to ChatGPT, it’s models are without doubt complex, but it’s models
are also far too broad and wide reaching to be as perfect as some of the more
specialist AI systems are. It needs much more time and development to mature
but it is being rapidly developed.
That said, tools such as
ChatGPT can indeed help you to be more efficient and there’s little doubt that
these tools can take some of the pain away from carrying out the mundane tasks
that we all have to perform as artists. I asked it to come up with some artwork
titles by providing an in-depth description of the art, as in, I explained it
in a way that I would explain it to a blind person and the titles were much
better than anything I could usually come up with.
I also had success in asking
it for a list of metadata labels that could be used to search for the work and
again, with a little effort on my part it was able to generate a decent list of
tags and labels that could accompany my uploads to sites such as Fine Art
America. They still needed a little human input but I think I saved about an
hours work from using ChatGPT and that’s not to be sniffed at.
I also had some success in
taking the same description and then asking it to generate a marketing
description that was SEO friendly, and again, it took just a little refining
and I had saved at least another hours work. That’s considerable when you find yourself
playing catch up and uploading multiple works, the time involved in coming up
with meta-tag labels, titles and descriptions isn’t insignificant. I spend at
least 8-12 hours each work just coming up with words when I have multiple works
to upload. With ChatGPT I saved at least half of this.
Where I found real time gains
was in asking it to synthesise data. I used it to generate a simple report
around my client demographics, historic sales, and success from social media
activity alongside a bunch of other metrics I use to keep track of where my
business needs to be and it generated a really simple report with key
highlights just from parsing the data I had copied and pasted in.
Usually I would use a platform
such as PowerBI to create reports, add the data to a data lake and hope for the
best, (I’m not a power, PowerBI user by any stretch) but this meant that I
didn’t have to fiddle around creating new report templates and I got the key
information back within minutes so that I could run side by side comparisons
with previous quarters.
 |
| Minidisc by Mark Taylor – I still have mine from the 90s, the batteries were still working after almost a decade when I eventually turned it on, they hadn’t even leaked which was a miracle! |
There was a monetary saving
here, the results were better than I could get from PowerBI and PowerBI is an
expense I can do without, although I expect that wouldn’t be the case if my
data lake was more substantial, if it were though, I might be inclined to still
use AI in my workflow and I’m sure it will be integrated more with tools such
as Power BI in the future. As it stands in comparison with large businesses, my
data lake is merely a small pond so I can scratch at least one outgoing cost
for now.
It’s also really good for
creating content outlines, if I ask it to put together three content outlines
for a blog post about X, Y, and Z, it has no problem doing that and again, this
saves a heap of post-production time because I have to spend less time thinking
about the small stuff that generally takes the longest time. It’s a solid base
on which to research topics that I’m less familiar with too.
For me, when it comes to topic
outlines, writing descriptors and meta tags, I usually treat this
non-productive time. Of course it is productive, it’s as critical as the
artwork and my writing itself in many respects, but it’s a detail that whilst
benefiting a client or a reader, doesn’t really move the needle on the bottom
line quite as much as finishing off a commission or creating a new piece of
work, or meeting with a potential client and if I’m completely honest, it’s the
bit of the job I like the least.
Arguably, SEO, artwork
descriptions and tags, are the essential ingredients in the sale of art, but if
some of this can be handed over to AI under at least a small degree of
supervision, then it free’s up more time to be able to focus on work that moves
the needle on the bottom line by allowing you to focus on higher value outcomes
and brings back the time you need to do the stuff you can really enjoy. Think
of AI as a new apprentice, an apprentice that still needs some hand-holding but
that can be sent off to do the jobs that don’t necessarily need to be
undertaken by the CEO.
 |
| Y2K2038 by Mark Taylor – yes, we’re due another Y2K event in 2038. Don’t worry, I made a bespoke sticker which you can buy alongside this print on my Pixels and Fine Art America store so you can place it proudly on the lid of your laptop and freak other people out! |
When it comes to integrating
AI in your workflow as an artist, it can be useful, even ChatGPT can save you
both time and money, more than that, it can take some of the mundane tedium out
of your workflow, at least to some extent, but, if you are thinking that AI is
some golden panacea that will allow you to completely take a back seat and
relax on a beach while it does all of the work, that’s still very much in the
realms of fantasy. At least it is for now.
To get the most out of it
today, you have to understand its limitations. You also have to take some of
its responses with a very large pinch of salt and be prepared to take
corrective action. Like a small child near a swimming pool, I’m not entirely
sure it’s wise to leave it unattended and then rely on it to do the sensible
thing, or indeed, rely on it always giving you a sensible answer. It’s as good
as the programmers who wrote the code and it’s as good as the knowledge base it
has learnt from, but it’s not HAL folks, not quite yet anyway.
You need to give it very good,
specific, and direct inputs to get very good, specific and direct results. It’s
that old adage, rubbish in, garbage out, and it takes practice. If you are
thinking that in the next five minutes you will be relaxing with a glass of
Pinot Grigio while it does its thing, I would advise that you don’t drink the
entire bottle.
AI delivered in the guise of
something like ChatGPT, in the publics hands is five, maybe even ten years away
from where it needs to be to do the kind of things that you imagine it will do
already. But if you set your expectations somewhere in the region of it being a
really useful tool that has the potential to save you anywhere between a
miniscule and a heap of time, particularly if you need to build some
foundations which can help you overcome that initial creative block when you’re
writing or researching, then I think it might already be where it needs to be.
Should you shy away from it or
embrace it, well, I don’t think we will ever be in a position to hide away from
it. My advice is to embrace it for the benefits that it can bring and take some
time to understand its limitations. Learn to work with it, and watch it grow,
because at some point in the future you will be relying on it even more than
you didn’t realise you already do!
Mark is an artist who
specialises in retro and vintage inspired works featuring technology and other popular
culture from the 70s, 80s, and 90s. He has been creating professional digital
work since the 1980s.
You can purchase Mark’s work
through Fine Art America or his Pixels site here: https://10-mark-taylor.pixels.com You
can also purchase prints and originals directly. You can also view Mark’s
portfolio website at https://beechhousemedia.com
Join the conversation on
Facebook at: https://facebook.com/beechhousemedia
connect on Twitter @beechhouseart or waste hours of your life on Pinterest
right here: https://pinterest.com/beechhousemedia
