An analogy is a rhetorical device that can
be useful in illustrating a point efficiently. Saying this simpler, more well-trodden, or
more familiar idea is similar to this new idea can have a lot of persuasive power. Over the past few months I’ve heard a lot
of analogies in reference to AI Art. Many of these analogies have been quite bad,
analogies that seem useful at first but completely fall apart because their cores are fundamentally
broken. Of course a literal perfect analogy would
not be an analogy but I’m not talking about a few mismatches I’m talking about comparisons
that don’t even make sense. With the ever looming threat of AI taking
thousands of jobs, and those who insist these fears are exaggerated, or insist this will
not happen, or in fact if it does happen its a good thing, a lot of emotions have been
charged and a lot of online arguments have taken place. So let’s start with the best analogies and
move to the worst and scrutinize their merits and demerits along the way. AI art is just like photography. The idea with this is that just like with
the invention of photography, AI Art is a new way of making visual representations far
quicker and with less skill and that artists were up in arms about the “death of art”
when it was announced too, but no such great replacement ever happened. While there were artists and critics who pushed
back against photography when it first appeared, there’s this one quote from a very prominent
artist at the time, and this quote from critic Charles Baudilaire for example, I haven’t
found enough evidence there was some widespread and pervasive movement against photography
from artists. Consider that when photography first came
onto the scene it was far more laborious than taking photos today, it was slow to catch
on, and everything was in black and white. There were of course comments from critics
about how photography was not “real art” which certainly has parallels with AI art
today. Another problem with this comparison is that
photography captures an image of what you put in front of the camera, which is limited
to whatever you have access to in the real world at the time. AI art is a computational hallucination informed
by millions and millions of artworks. You could say the skill level and time involved
with each of them is comparable but the inputs for these outputs are not. This is currently the best analogy for AI
art I’ve encountered despite its flaws and it’s the only argument I find even mildly
convincing as a rebuttal to the concern that AI art will replace artists. Throughout all the art history classes I’ve
taken whenever the invention of photography is brought up the emphasis is that photography,
over many years, changed art, it did not replace it. It is possible that AI Art will simply change
how art is made and not diminish the quality of it or amount of people making it. But it’s within the flaws with this comparison
that I think a less optimistic picture is painted. In the rapidly changing age of AI it’s important
people stay informed, however, with the plague of media clickbait and outright misinformation
on the internet this is often easier said than done. That’s why I’m excited to tell you about
an app and website I’ve been using that is working to solve this issue. Ground News is a small independent team that
is working to add context and transparency into the information we consume. For example, when I was researching I came
across this story on reports of Europe wanting a label to accompany any AI generated media. On the website is an intuitive visual breakdown
of the news outlets covering it, we can see the total number of sources reporting on this
and the bias distribution of coverage- in this case it’s pretty evenly covered by
the left and right with the majority of sources from center leaning. For each article you can see who owns the
outlet, the factuality of the sources and where they lean politically. You can also compare headline and summary
coverage. This kind of quick overview can be very helpful
for my research and for spotting trustworthy information. I particularly like that we can follow specific
topics such as Artificial Intelligence– to get notifications on breaking news and
even ‘blindspot’ stories that are over or under reported by one particular side. In this way, it’s easy to stay fully informed
and find information you might otherwise miss. If you find this as useful as I do, go to
ground dot news forward slash solarsands. Sign up for free or subscribe through my link
before July 1st for 30% off unlimited access. I believe this is a crucial service to combat
media bias and I would like to thank Ground News for sponsoring this video. AI art is like Chess being played by computers. I think the idea with this is that Chess is
a sports-like, okay I guess it’s a sport—and so computers out performing chess grandmasters
had no real effect on the sport. So there will be no real negative effects
with human artists being outperformed by computers. Chess didn’t die, chess players didn’t
diminish so why will artists? An alright comparison to make the point that
a lot of people like to consume art for art’s sake. The problem is that art has plenty of utility
beyond competition. Sports are for entertainment, the point is
to watch human beings compete. Some art realms have aspects of competition,
a ferocious pursuit of beauty, even pushing the limits of human achievement but we need
to be honest with ourselves... Most art is a product. People don’t like to be told this because
we have all these notions about individual expression and “soul”--all this mushy
emotional stuff that comes with exalted mantle of “artist”—but I think we tend to overstate
these notions. That’s only part of the reason people make
art. Contrary to the popular saying, you can’t
make art if you’re actually starving. Art has a price tag, people can sell it, people
can buy it, it’s an economic asset, it’s a means to an end. It’s dictated by the same calculus of risk
and reward/supply and demand as anything else. Art is also everywhere, it’s not just paintings
that sell for ludicrous amount of money in auction houses, or the Instagram artists or
the landscapes your grandma makes. It’s in, packaging, stock illustration,
posters, billboards, commercials, user interface design, industrial design, architecture, video
game design, movie concept design. Art keeps people alive not just spiritually
but economically. It can be good therapy but it can also be
the cause for therapy. It can be fun but it is also a JOB. ART is a JOB. Tons and tons of what we consider “fine
art” today was initially made for money because it was the artist’s job. A lot of artists make artworks not just because
they enjoy it but to gain followings to gain influence, to build brands, to make money
because it is their job. The ideal in a healthy society is that people
get to fulfill their artistic vision and not be influenced solely by the pursuit of profit,
but without the expectation of making money a lot of art simply would not exist. If you sell your art or sell the idea of your
art, your art is a product. I don’t think that devalues it by default,
I don’t think that makes it “not art”. But here is the economic reality: when a company
needs to produce 1000 unique concepts so that they can get a good idea for something new
in their product line are they gonna pay a team do it? Or are they gonna pay the far less expensive
monthly subscription to Midjourney version 17? Or better yet they’ll produce their own
personalized AI model trained on the companies archives, infinite concepts for a fraction
of the price. So when people say things like “If AI is
capable of taking your job you deserved to be replaced” I’m really left basking in
the callousness of that statement. I’m sorry you think so low of artists who
are not at the cutting edge of creative development, a position that has only been occupied by
a rare few in art history. I’m sorry you think so low of the people
who are out there designing your world for you. There may be a point that AI art gets so good
that all I have to do to make the visuals for my videos is input some text and do some
post editing, now I doubt I would do this if AI continued to train on art without permission,
but if there was a theoretical machine that used just public domain works, I mean...I’m
not just going to be able to save money I save a lot of time too. That’s less time drawing out references
for an artist to go by. If I wanted a custom visual for a video I
would have had to pay an artist, now that artist would not get hired. Those are jobs being lost. I don’t feel very good about that, but what
if everyone on YouTube does it that way as well? What if it makes my videos even better? AI Art is like NFTs. This is one of those comparisons where I’m
like, y’know your heart is in the right place but your way off mark. Yes NFTs and AI Art both have scummy tech
bro scammers in them but NFTs are completely useless. AI Art is incredibly useful. The ability to sort of but not really collect
digital artworks is silly, the ability to write a sentence and get a more often than
not well rendered approximation of what you were thinking instantaneously is revolutionary. The first AI art video that genuinely made
me feel something...the kind of feeling that is difficult for even human artists to evoke
in me was this Dark Souls reimagined as a dark 80s fantasy film video, the concept has
been done nearly to death at this point but when I first saw it I was floored. It’s such a natural marriage of two things
I like yet I couldn’t have imagined it for myself, much less render it. It’s one of those things that made me think
“oh we’re in trouble.” I’ve seen AI art being used in the backgrounds
of videos, in advertisements, on the news recently, and in random spots every now and
then, slowly seeping in. I saw people in my design and architecture
school using Midjourney to generate concepts to base their projects off of and this was
common. NFTs wish they could be as useful as that. Techbros don’t need to exaggerate and shill
about the impressiveness of AI it’s right in front of your face. Some may make the claim that NFTs and AI art
are similar because they’re both fads, right? No, AI art is not a ‘fad’, it’s here
to stay. We have to either adapt with it or adapt around
it. The last episode emphasized how we should
set a precedent for asking permission to train on copyrighted works and I still believe that. I’m not even interested in past AI models
still using copyrighted works as long as AI models of the future use works they get permission
for and public domain, because I’m sure as soon as those are out, models of the past
will be obsolete. If that ever happens I don’t have any arguments
against AI. That’s it really, I can’t think of any
ethical tweaks beyond things like bias, that can and/or should be made to these models. Since it’s been talked about so much already
the shock may have worn off but this technology is still impressive and it’s not going away. Now I was criticized for not acknowledging
the limitations of these systems. So let’s talk about the limitations of these
systems...of which there are now fewer than there were the last time I made a video on
AI art. Hands used to be something that AI models
constantly struggled with, now that problem has been reduced. Letters used to be complete gibberish, barely
letters really, now less so. AI generated video is now even more impressive
and consistent. There are still plenty of problems that current
AI diffusion models really don’t have a good grasp on, very specific aesthetics can
be a struggle, any sense of 3 dimensional space is still severely lacking, consistency
is better but still poor, getting anything that is put together in a world that obeys
physics is quite difficult if not impossible, unique compositions are rare and a lot of
AI generated art has a recognizable generic sheen to it. However, while some of these may be hard limitations
based on the technique the models use, a lot of these can be viewed as simply challenges
to overcome. People will look at tweets like these, with
videos demonstrating an AI filter and pick apart all the ways that the AI failed, note
all the constantly changing elements, give a hearty laugh, pat themselves on the back
for being oh so clever and rest assure that AI is just a “cheap parlor trick” because
this random AI video is not perfect. All of you need to start looking at the bigger
picture. The fact that an AI anime filter can even
keep this amount of consistency is pretty alarming. The fact that with a glance AI art can pass
for something created by a human is pretty alarming. And that time period where you have to stare
at an AI image to realize it’s AI generated is going to get longer and longer until you
just can’t. Maybe I’m not as observant as I used to
be but I’ve been tricked a few times already that something wasn’t AI when it was, I’ve
also been suspicious that some things were AI when they weren’t. Now I find myself zooming in on every illustration
I see wondering for a few seconds whether I can trust what I am seeing and ultimately
it’s just a massive headache. No one can be expected to be a detective every
time they see an image online, which is often. Enough with the Never statements. AI art models, specifically diffusion models,
are unlikely to do some things, but can possibly do a lot of other things, I don’t think
we’ve seen their full potential yet whatsoever. Even if there are plenty of things they can
never do I certainly don’t think even this brand of AI models are in their final form,
I seriously doubt that copying and pasting these long keyword paragraphs that look like
the Vietnam war memorial—will remain the most efficient method in getting what you
want out of the machine. I could easily see someone creating interfaces
where you could adjust the images in real time, or I could easily see some sort of mind
reading apparatus being used to visualize what you have in your mind combined with some
form of automatic optimization and I don’t even think that’s that far away. People are a lot more clever than you think
when it comes to getting what they want, just because you couldn’t get some specific thing
from the machine by tinkering with it for a few hours doesn’t mean that is now the
boundary of its power. Simply consider what is possible when you
use other tools in addition to these models: AI art is like collage or photobashing, now
to call AI Art a collage or photobashing tool is inaccurate technically speaking. Collage and photobashing are art forms in
their own right, but even then there are disputes about how much you can’t change in a collage
until its copyright infringement. AI art is different from that in that the
artworks it is trained on inform what the outputs will create, not cut out pieces, there
is an argument to be made that this is fair use, or that this isn’t fair use, like I’ve
said I think there’s certain stipulations that make me lean towards not being fair use
but the law could see it either way, which is why I have my doubts that litigation is
the only way or most effective way to solve this problem. I’m just saying if any of the lawsuits going
around fail well...don’t be too surprised. The fact that AI art models have the potential
to cause such catastrophic damage to the art making ecosystem makes me believe that the
companies that made these models really should have just asked permission first. But the tech is out there so really even if
the big names get taken down people can still use their own models on their personal computers. AI Art is like CGI, CGI caused a revolution
in making animated films and the computer does a lot of the work there too, just like
AI. Only in so far that it renders some amount
of detail by itself, other than that no, CGI is not like AI art at all. CGI is it’s own art form that requires lots
and lots of decision making that the computer is not capable of. Even a basic understanding of the sheer amount
of time that goes into CGI animation shows the extreme differences between CGI art and
AI text prompting art. AI Art is like digital painting/photoshop. I guess the idea behind this is that digital
painting was an advancement in giving people more accessibility to painting and drawing
but other than that this comparison is almost completely useless. Photoshop doesn’t automatically create content
there’s no text bar where you can type in what you want and it makes it, there’s no
button that you can press and it generates an image, that’s what ignorant people think
digital painting is like but that’s just not the reality at all. It turns putting pigments on paper into pixels
on a digital canvas plus a few time saving extras, that’s it. There’s an irony here in that all of these
things AI is being compared to...AI will and is seeping into. There’s a sentiment going around that AI
art requires little to no effort to make, and while that may be true for most users
of the tech it’s simply not true for anyone who takes it seriously. When I spoke to AI artists there’s typically
a lot more going on than typing in a prompt and calling it a day, there’s a fair amount
of editing, post processing, and messing with several tools at once to get what you want,
and this process takes a real effort. Now to say this takes as much skill or is
even close to the same process as making art traditionally is ridiculous and to argue it
takes the same time and effort seems self-evidently absurd, AI art is useful because it takes
less time and effort, that’s the main reason people use it. Let’s not confuse our concepts here calling
something “fanart” or not tagging something with “Ai art or ai generated” when posting
an AI work, as if it’s on the same level as regular art is pretty misleading. Just like with the Art Station fiasco I’m
not concerned with whether it’s “real art” there is simply a clear and obvious
difference between the process of an AI artist and a traditional artist. Let’s not start mixing the two because they
categorically are not the same. The biggest flaw with all of these analogies
the largest deviation from any other technology we’ve developed is that AI Art models make
creative decisions. It may not be creative in the same way humans
are and it may still have lots of limitations, but when I type in “steam locomotive with
spider legs” into Midjourney I didn’t ask it ask it to keep the train tracks, a
more intelligent system might figure out that a train with spider legs wouldn’t need train
tracks anymore--it figured their was a high probability a train would be accompanied by
train tracks and it put them there. But I also didn’t ask it to turn the metal
hosing into this creepy hair, or put this organic housing underneath that seems to morph
into machine, or make the legs this weird combination of train wheels and spider leg
parts. Those are things it added, things I think
are actually kind of clever, things I wouldn’t have thought of. Hell this was a random idea I’ve had for
a while and this just brought it into fruition, and I think the results are pretty impressive,
with some tweaking, I would actually use this. Sure one could say it arrived at these ideas
by accident, that is has no understanding of the things it does, but that doesn’t
change the results. This is the closest we’ve ever gotten to
a machine producing art on its own. No camera, no software, no other art making
tool holds a candle to it. Most times I pick up a pen and doodle for
the sake of doodiling, but sometimes I pick up a pen with an idea and I think to myself
“what’s the point? I can ask an AI to generate it, it will be
roughly what I want and I can just do some editing, and if that’s not the case now
it will be soon.” Needless to say I’ve been drawing less recently. This sentiment is of course shared by others,
it’s an existential crisis--what reason do most artists have to continue cultivating
a skill, especially when its a skill like drawing that takes a lot of effort and time
and personal sacrifice to master? Maybe somehow there’s enough of a social
stigma against AI art that most people won’t use it because they believe they want to maintain
some form of artistic integrity. But there’s just some things where that
doesn’t matter. Come on, think about it, what is the one type
of artwork where its main consumers have absolutely no care at all for the artistic integrity
of the work? Pornography. Consider that the consumption of porn has
in several instances been a driving force behind technology such as VHS and of course,
the internet itself, I don’t think this is any different. NSFW artists may seriously be in trouble. Although AI models are difficult to get explicit
images out of, there are ways of getting around it, and there have been some posts going around
that have shown, how do I put this delicately…“convincing” results. So is AI replacing art jobs right now? Well the evidence so far has been just bits
and pieces. There’s been one random article of AI taking
the jobs of game illustrators in china, with one illustrator saying she simply tweaks AI
images for one tenth of her original rate. Netflix Japan put out a tweet that cited a
“labor shortage” as one of the reasons to produce a short with AI assistance for
background images. Now the short itself...is okay. The backgrounds are passable if you don’t
look too hard at them but they certainly are not good enough for me to think background
artists are going to be replaced soon. One twitter user found a guide that details
how to commission an artist to make a colorless sketch at a cheaper price while using stable
diffusion to render the rest completely. Freelance and self-employed people will certainly
have to navigate a now even more difficult world. I kind of expected this bits and pieces implementation
as AI art is currently surrounded by controversy and instability, with likely the biggest unknown
being whether AI art can even be copyrighted, so people are only left to speculate. But the thing that made everybody lose their
minds was the Corridor Crew short. A video in which they used the stable diffusion
model and a dataset consisting of screenshots from Vampire Hunter D Bloodlust (a fact they
disclose as clearly as possible) to create a short video that has the appearance of a
hand-drawn rotoscoped animation. This video is pretty amusing and although
it does not look incredible and there are plenty of things you can pick apart about
it, it’s an interesting look into what AI is capable of. The art community did not handle this well. Yeah Corridor Crew could have been a little
bit more delicate and could have mentioned some of the ethical concerns people have. But I’m sorry this was not worth being this
outraged. This isn’t the company that made the AI,
this is a VFX demo by a YouTube channel which specializes in VFX demos. This short was hours and hours of work with
an AI filter on top of it, to the point where the AI filter is kind of the icing, a lot
of the entertainment of this video was the editing, acting, and post production. There was also this strange mob of some reasonable
criticisms mixed with confusing criticisms. People saying “its not animation its just
rotoscoping” as if animation has not used rotoscoping for nearly as long as animation
has been a thing? “Why didn’t you hire actual animators?” Because the whole point is a tech VFX demo
to show what’s possible without people drawing frame by frame animation. This isn’t jobs being taken away from animators,
if they didn’t use AI this video just wouldn’t have been made. “Look I paused on a frame where your experiment
failed boom AI art destroyed--video is bad.” By the way they did bring animators on to
react to their work, which the animators made critiques of….because it’s an experiment. “Pick up a pencil” as if the entire point
is to see what can be made while avoiding the, let’s face it, incredibly laborious
process of frame by frame animation using studios full of people. The intention of these videos was clearly
and obviously to explore a different type of animation to potentially help animators. The reality of these tools in helping rather
than replacing animators remains to be seen, but I think their intentions should be considered
here, and the ethical concerns people had should have been explained to them, y’know,
not relentlessly shot at them through a screen of petty insults. The message I was getting from all this vicious
criticism was not “you should have talked about the ethical implications of the data-sets”
the more reasonable stance, it was “AI Art is always evil and you are bad people for
using it” the far more extreme stance. Anti-AI activists have to get around the hurdle
that to the average person AI art is cool, okay? Yes it’s kind of made of art taken without
permission, but it’s cool. Good luck bullying, and yes that’s what
this is, bullying, every single person who uses AI, seriously. I really really doubt that harrassing people
about their AI art pet projects is the winning strategy here. Especially when you fail to consider their
actual intentions and immediately paint them as villains. I need everyone who is Anti-AI to understand
that the ethical concerns about AI are not universal especially not to the average person,
just because lots and lots of people get angry on twitter doesn’t mean the same sentiment
is shared by the general public, it doesn’t mean it’s a settled issue, it doesn’t
mean being as viscous as possible means people start agreeing with you, yes even if some
of the enemy is being viscous with you as well. There’s a cognitive bias going on here that
lots of movements fall into, people get into echo chambers, they lose sight of what they’re
actually supposed to be fighting for and it just becomes a witch hunt every few days. When an experienced level headed animator
makes a reaction video to the Corridor Crew video what does he say? “Oh wow that’s cool” “but please let’s
get permission before putting artists work into AI” That’s it. So much more graceful than everybody else
who went completely hysterical over this. So much better than calling it “degeneracy”
Like seriously what the hell? You have to actually convince people who are
not on your side why AI art is cause for concern. You want to pressure someone pressure the
companies that will have a far more tangible impact on the economic and ethical realities
of this technology. Not some random internet users posting their
AI art pics. If you want to criticize a wildly popular
and well-liked YouTube channel for using AI make your critiques clear, strategic, and
diplomatic. I don’t know, act like adults. What’s worse it that Pro-AI people have
figured out the outrage game. Take this post for example; a decently impressive
but still obviously uncanny AI creation with an exaggerated caption “oh look how good
this AI generated thing is it’s going to replace you!!!” then the poster goes into
the replies and baits people, saying “hurr durr this video is actually perfect” then
hundreds of people engage with the tweet pointing out the obvious flaws in the video, not realizing
the original poster is pretending to have blind devotion and acting stupid on purpose,
and bing bada boom they shill whatever they want on top of advertising the video now with
a much bigger audience all thanks to people angrily quote tweeting it, and these tweets
get millions of views off of this little game. Do you know how many views the guy who made
this video got with their tweet? 680,000
How much does the outrage tweet have? 39 Million. Stop falling for this crap you morons. Hate clicks are still clicks. I’m worried that this Anti-AI movement is
falling into the same pit as other movements before it. Websites like twitter encourage this kind
of ridiculous emotional reaction, and I fear everything is getting way too accusatory and
way too stupid to the point where a lot more damage is going to be done before anything
gets better. People need to lay off the bickering no matter
how tempting and focus on the substantive plays. By constantly dunking on random AI projects
at best you might be making a small amount of progress on the optical front, at worst
you are playing directly into their hands. I believe there are sides to this, and this
is something of a war, and I’m telling you, you need better weapons and you need to pick
your targets wisely. At this time I can’t bring up a statistic
that says “this amount of people have been or will be replaced by AI” I don’t think
we’ll see reports like that for a while. Especially when you have tech bros embarrassing
themselves with fantasies of replacing an entire industry with AI without considering
the nuances of that industry even for a second. I don’t see it replacing most jobs in the
near future. Despite this AI is already present in low
effort scams and social media content, other more respectable jobs like being an illustrator
who takes commissions for small businesses, graphic designers, advertisers, coloring book
illustrators, that stuff may be at risk. But big name artists who have a brand and
a personality that people admire on top of their artwork, artists who work almost exclusively
in traditional mediums like oil paint or graphite, artists who make things purely for their emotional
content will probably be fine for the foreseeable future. Any artist who has to do something that requires
even a little bit more complexity than illustrating a vague concept will be probably be fine. But I have to put an asterisk next to this
that says “for now”. My honest advice for artists would probably
be to start pursuing a second form of income on the side because it really does seem like
a lot of things are up in the air and I don’t know what’s going to come down. Make a YouTube channel, post videos of your
process, set up paywalls for some of your work, learn some other physical art making
skills. Make sure you stay informed about these systems
and the politics behind it. And most of all fight for your rights. Corridor Crew are not the people who will
usher in the AI dystopia---it’s those in power that insist on asking for forgiveness
rather than permission, it’s the average person who doesn’t do their due diligence,
and it’s the systems that are designed to keep us hostile and divided.