|
|
Artificial Intelligence or Actual Ignorance? |
I have to admit that in 2025, I turned to YouTube for many things other than
sewing ideas. (As a matter of fact, about half of what I'm subscribed to these
days has nothing to do with sewing/crafting; they are channels run by rational
thinking, ordinary human beings disseminating current events that should be
important to all of us.)
For the fact that I've been spending more time on YouTube videos, I can't
help noticing that more things are catching my eye from the sidebar
recommendations.
Most notably, I recently became aware that crafting channels are employing
AI to create the thumbnail images that appear on their title screens. My
first encounter with this phenomena was fairly innocuous: I clicked on a
video by a crafter who had showed off a group of small owls made out of
old denim. When I got to the end, I thought the finished item was cute,
but couldn't understand why I felt less enthused about it than at the
beginning. Scrolling through the comments, the reason why became obvious.
Someone called her out for encouraging click-throughs by using an AI
generated image of the project, which did not resemble the actual finished
item. Once I compared them myself, that was unquestionably true.
 |
| Is it a matter of ethics? |
Now, this is probably one of the least egregious bad outcomes that can
happen with AI. This was a free tutorial for a craft that ultimately
looked similar. One could say that the most controversial part of
it is the idea of trying to get clicks by featuring an image that
doesn't accurately represent the finished item. But then this has been
happening since before AI; one channel that I've known about for years
has consistently used photos – without attribution – of other designers'
bags to promote her DIY versions of them.
Being called out won't stop these people from continuing the practice;
about that much, I'm sure.
If you're not familiar with what I'm talking about, here's an entirely
different example of yet another "denim owl" project that uses AI to
attract views on YouTube. (Obviously, the crafter didn't make eighteen of
these fabulous, perfectly identical, denim owls.)
|
| Another "denim owl" themed project on YouTube that uses AI-generated images... |
When one comes across something like this, it would be natural to be
curious to see how they're made, right?
Not wanting to be "gamed" this time, however, I clicked directly to the
end of the video. This is what the finished owl(s) that was made during
the tutorial actually looks like.
|
| Finished owl keychain project by Ziba Trix on YouTube... |
Gotta say, it lacks the "ooh, I gotta make that" feeling that the first
image evoked. (Not to pass any judgement on the project itself, but I
personally wouldn't have clicked on the video had this image been shown up
front. By the way, I think a YouTube video has to engage the viewer for at
least thirty seconds in order for it to count as a "view;" therefore my
peek at this one most likely didn't count.)
I must admit that the level of adoption of AI by YouTube content creators
came as a bit of a surprise to me. (It's probably because I had yet to use
any AI tool myself.) Also, it's almost like I hold those "around me" to a
higher standard and it's a bit of a letdown to see them succumb to a
ubiquitous practice just because everyone else is doing it. (Or perhaps
the actual problem is that I'm being a luddite!) In any case, when I saw
evidence of it on one of my subscribed channels, I was oddly disappointed.
Ulyana of Estadistica Datos del Mundo has built
herself up a following of over one and a quarter million followers since her
early 2020 debut on YouTube. It would appear that she has already mastered
the "secret" to success on that platform, without the need for AI. But it's
clearly being used on her channel, with more recent thumbnails featuring
many of whatever the project is. (Now, maybe she did make
seven, eight, or fourteen identical whatevers, but I suspect not.)
|
| Obvious AI-generated image of project... |
The little heart-shaped pouch that's shown above shows how creators like her
are plumbing the depths of AI to come up with project ideas that they then
attempt to reproduce. As might be expected, I do believe that the AI created
image shows a nicer looking pouch than the actual finished item (see below).
The AI pouch has a larger, curved "flap" that mirrors the rounded outline
of the top part of the heart, as well as an additional opposing
(non-functional) decorative snap. The real pouch features a simple tab
situated along the straight side of the heart. Sewing-wise, the AI version
would be difficult to pull off. None of the fourteen pouches in the
AI generated image shows the top right side of the pouch from an angle that
would reveal how the flap is attached to the base.
|
| Picture of actual project... |
There is a common take on the whole "AI thing" that since we can't do
anything to stop it, we may as well get on board and take advantage
of it ourselves. I'm not sure what my opinion is on the whole matter, but it
does muddy the water for everyone. (Spoonflower design
competitions, for example, specifically state that AI designs are
ineligible.) In cases where no real scam – more on actual scams later – is taking place, raging and
stomping one's feet against it can be seen as a bit of an over-reaction.
Over the course of writing up this blog post, I had to – obviously –
"create" my first AI image. My prompt was, "I want an image of an owl made out of old denim that is small enough to
be used as a decorative keychain."
|
| Denim owl as keychain generated by Artlist... |
Not bad, huh? At least this one looks like it was sewn. The sewing
might be challenging, but it wouldn't be an insurmountable task. I think that could be a piece of leather being used for the beak, and the claws could certainly be made out of brass coloured
beads. Doable. But you know what my next thought was?
As the YouTube commenter above pointed out, this form of AI becomes AI by learning from everything that's out there on
the internet. (Which is why my graphic at the top of this post is titled
"Artificial Intelligence or Actual Ignorance?") I was at a party last
November where a themed version of Jeopardy was being played. When one of
my tablemates – a young woman in her early thirties, I suspect – said that
she found the (ultimately wrong) answer on ChatGTP, she looked at me like I was from another planet when I told her,
"Well, that just goes to prove that ChatGTP is as ignorant as the rest of
the internet." It didn't seem to occur to her that online AI wasn't the be-all and end-all
of all knowledge.
But back to my AI image... isn't it entirely possible that this might be
strongly representative of someone else's intellectual property? (I know;
"strongly representative" does not translate into copyright infringement.)
Could it be a complete duplicate? Does that ever happen? (I
did run it by
Google Lens
and in this case, did not find any similar matches, never mind exact
matches. So perhaps Artlist did a good job here.) All I'm
saying is that we just never know if someone's chosen AI tool will decide that
an exact copy of something that already exists is the best match for the prompt that it was given.
Dare I say that anyone relying on AI for things and not checking its results is showing real "opposite of intelligence?"
Of course, YouTube isn't the only place where AI is tripping out.
Literally every social media platform is rife with AI content these days.
Manufactured videos have long been popular on Facebook, for example, with many of them spreading misinformation like virtual
wildfire.
But I do take issue with the rapidly increasing use of deep fakes to gain
clicks. A few days after the death of Rob Reiner, a video showed up in my
YouTube feed featuring Billy Crystal talking about how it "should have
been him" instead of his friend. Now, I don't know much about Billy
Crystal, but what I do know is that he is not a publicity hound
and that he and Reiner have a history going back over fifty years.
It seemed unlikely that he would come out and speak about this horrific
event in such a manner. A quick glance at the comments confirmed that
people were chastising this YouTuber for using AI to produce a money
grubbing fake video.
My husband likes to say that he can tell when a person is AI
generated/deep faked. I think, however, that until/unless one
suspects that something is fake, one may not be so astute or
discerning. The problem is that we are being forced to look closely at
everything to see if it's real or not. I mean, yes to employing
good habits to reduce the potential of being scammed (i.e., I didn't think
that my prime minister would endorse investments of any kind,
but apparently some trusting seniors did) but having to take a magnifying
glass to a screen every time you see someone saying something? Anything?
That's too much and I don't believe that the vast majority of internet
users are up to the task. Unfortunately, our world is quickly descending
into a "fake until proven true" reality... and sadly, that just might be
the least dangerous part of AI.
In the meantime, what do you think about YouTube videos that entice you
to click, based on AI manufactured images? Is it dishonest when the
final outcome is not what the image promised? Is it fair play for
creators on YouTube to gain views – which boosts their income – because of their
"crafty" ability to use AI?
Will this discussion now prompt you to jump to the end of a video as
soon as you open it up, just to see if you're being led around by the
nose?
Practice smart viewing, I say! (And seriously — keep an eye peeled for real scams. It's better to be skeptical these days.)
'Til next... 🇨🇦🍁
