Let's Talk About AI

By "AI", I am referring to the creation of content using AI generated images, video, etc.  Not the AI used to scan metadata or moderate content on social media or media sharing sites.  Some things should still have human oversight whether a platform uses AI to analyze or moderate content or not.  

Kiss this, Princess.
Also see this creation on this page: House of High Spirits.

Text

First, the text on my platform is mine.  I don't use AI for that.  When it comes to text, there is nothing wrong with a little help with checking spelling, punctuation, capitalization, and to some extent, grammar.  It isn't always accurate, but it can be very helpful.  However, my text is my own.  It's my voice in print.  I've always enjoyed writing, and I've worked hard to write as well as I possibly can.  It doesn't matter what I'm writing about.  What matters is that my writing is my voice and authentic to me.  

Where would something like AI actually be useful when it comes to text?

The proofreading and editing part I mentioned above can't be hurt by a little AI assistance.  People who struggle with spelling and grammar might benefit from it when leaving comments online.  If you need to write a brief search description for a page on a website or blog like this one but you're just not that kind of writer, a little AI can't hurt if it helps you to write descriptions of your own.  If you need something to throw you a bone to come up with ideas based on what a search description should include, having an AI assistant that can scan your content and give you ideas that are relevant to that content, then a little AI assistance can't be too bad.

What I avoid is having AI write all of my text for me.  Let me also add that I prefer to write my own descriptions and will default to that preference no matter what.

Tags

Meta tags and hash tags are an area where an AI assistant can do just that: assist.  SEO (Search Engine Optimization) and platforms in general despise keyword stuffing with a passion and tend to rank content lower if they "sense" keyword stuffing.  Someone like me who tends to be very thorough in all aspects of content online can be vulnerable to unintentionally using too many keywords.  

Oh, sorry.  Keywords are basically like meta tags.  They are used to help search engines determine the content on a web page for indexing.  These are typically what HTML turns into meta tags.  For example, my oldest website is about Anne Rice's Mayfair Witches novels.  Keywords would be things like "Mayfair Witches", "Anne Rice", etc.  Well, without the quotation marks.  Website code puts these into an HTML meta tag that specifies these are part of the site description.  

Hashtags would be #mayfairwitches or #annerice.  These tags are typically for social media like YouTube, Facebook, Instagram and other platforms.  YouTube is itself a search engine as well as a social media platform, but sites like Facebook and Instagram have their own ways of indexing content.  Hashtags are a part of that.  

FYI: on platforms like YouTube, there is another type of tagging, which is usually referred to as"mentions".  These tags begin with a @ symbol.  These are the type of tags that are considered a form of making contact with the individual or site/channel/profile you are tagging.  These are the tags to be especially careful about overusing for many reasons.  Spam tagging, unwanted contact, and other disturbing ways to show a lack of social etiquette. 

Back in 2008, this whole business of keywords and tags was mystifying to me.  I didn't really pay attention to the keyword section of each page of my website on its original host, Tripod Lycos.  This means many pages either had the same keywords or none at all.  In later years, I finally did go over the keywords for each page.  Hashtags and meta tags (in HTML format) weren't allowed in the keyword section.  The keywords did go into the website's code, which means it added the keywords I provided into a meta tag the host generated itself.

Where might we run into problems with keywords and tags that AI can help solve?

If you have an AI assistant scan the content of each page, each watch page individually because you're struggling to decide what keywords, what tags should be used, that's where it can help.  Why each page individually?

To avoid using keywords and tags that are irrelevant to the content on that page.  What is relevant on one page isn't necessarily relevant on another.  If you specifically talk about witches on one page and ghosts on another, then you won't want to have a keyword like "Ghosts" added to the meta tag for the page that only discusses witches.  

That's what search engines will positively screech about.  Because they are trying to index content and errors like that on a website means there will be errors in search results that predictably drive people bonkers when they're trying to find something.

So why not have an artificial secretary review the content and suggest tags to use?  

YouTube was already doing this by suggesting relevant hashtags before AI really took off.  Whatever relevant hashtags a scan finds for you can also be used as meta tags.  People usually add hashtags to titles and video descriptions on YouTube, but video uploads as well as channel settings in Studio have a field to enter tags into.  This is where you'd add meta tags.  This is where you would add tags like "Mayfair Witches" or "Anne Rice".  People do put hashtags in these fields, yes.  You can also use meta tags in that fields like you would keywords in the keyword fields on websites.

So far, I think we can see that AI has some useful utilitarian purposes for content creators.

Content Creation

As for content creation, hey, nothing wrong with that.  I create funsies on NightCafe, yes.  And I have thousands of versions of things generated, mostly due to my persnickety nature when it comes to results being what I want.  Now those are accurately considered AI generated.

  

NightCafe also has a sketch-to-image feature that I've used to render sketches as well as images of my 3D models.  To be clear, though, there is absolutely NOTHING about sketches or my 3D models that are "AI generated".  Not even close.  I found NightCafe when I was searching for a way to render images exported of my 3D models in a photorealistic way.  In fact, many professionals who use 3D modeling software for architecture use rendering engines like Lumion or Enscape for that purpose.  They use them to give clients a visual idea of what their project will be, what it will look like when completed.  Okay, yes.  A sales tool.  

Upper Depot Street 3D by Demilune

The image to the left and below show what a 3D model looks like in SketchUp and what it can look like using a rendering engine.  Needless to say, the results are astounding.

The problem is my old fart of a device cannot handle the enormous amount of resources necessary to run these rendering engines.  These rendering engines are also expensive.  

Are we surprised?

Upper Depot Street 3D Render Artist Unknown
The buildings from Demilune's model
are on the far right of this image

No, I wasn't surprised, either.

This is what led me to finally be willing to look at AI generators as a possible means of rendering my 3D models and even sketches to be photorealistic.

Render versus Generate

This is the part where, if you use AI like NightCafe as a rendering engine, you would say "AI modified" or "AI render".  I typically indicate those types of images as being "AI rendered" when sharing them on platforms like YouTube.  Since I am very particular about making sure my content is not mischaracterized, I began to use the tag section to indicate what exactly I use to create the content.  If it is a render I used NightCafe to make from an image of my 3D models, I will indicate that it is an AI render, that NightCafe was used to render, and that the model itself was made in SketchUp.  

So, the tags would go like this: AI Render, NightCafe, SketchUp, 3D Model.  Etcetera.  

We know good and well AI can make mistakes.  It really is in its early stages.  The idea behind using tags on YouTube as a means of accurately characterizing how content was made is to help provide platforms with ways to accurately describe and characterize that content without mislabeling something as "AI generated" or even "AI modified" when it is not.

Recently, I came across an image that was used as a promotional item for AMC's Mayfair Witches series.  It does show up sometimes with the label, "AI modified" on Pinterest, but it is unlikely that this is the case.  A local artist in New Orleans, Pompo Bresciani created the image (please click his name to see his portfolio), and given its age, I'm 99.9% certain it is NOT AI modified.  

The opening credits of the same show were created by a group of artists who have their work on Behance, which is owned by Adobe.  Most likely, the graphics of the promotional material that came out just before the show premiered in January 2023 were created using Adobe After Effects.

AI and CGI

If I recall, film and television in general have used what has been referred to as CGI for many years.  When you think about it, CGI might just be something of an inspiration or even a precursor to the idea of AI generation.  

The best example of the use of CGI in films that would be even somewhat comparable to how AI could be used in content creation now is the 1997 film Titanic.  Look at the scenes where the ship's stern is rising from the water with people suddenly losing their grip on deck rails, capstans and whatever else people grabbed at to try to keep from falling.  For those people who actually were people, yes, this massive set of the ship's aft deck had things like those massive capstans made of softer stuff than what capstans were really made of (metal).  But many of those people were added to the scenes by CGI.

Why?

Well, because this wasn't 1925 on the Middle Fork of the Coquille River.  The risk of actual people being killed during scenes that have them tumbling down or overboard is one that filmmakers, producers, studios, etc. just were not willing to take.

1925?  Say wut, now?

I'll explain later.  Maybe in a later post.

AI and Intellectual Property

For now, let me say this about AI as a creative tool.  There is nothing wrong with having a little fun with it as long as it is used ethically and responsibly.  I am aware of the fact that many artists across all means of artistic expression are NOT happy about their work being used to train AI modules without their consent or even their knowledge.  It is one reason why, if it's my own content, I do give permission for companies whose platforms I use to, in turn, use my content to train AI modules.  I just don't use generative AI beyond a certain point.  Like text.  Or music.  Or even videos, unless I'm animating a face with my composite character face that I made for my use in my own creations.  

My character composite does make use of my own likeness.  Why?  It's very simple.  Legal reasons.  I just don't like the idea of using the face of someone else who is alive, especially if they are well known.  It's to avoid that resemblance as much as possible.  Sometimes, a result might bear some resemblance, but it is unintentional.  

My Character Composite by ME

I often use these types of creations of mine as avatars.  And yes, I do have characters of my own.  I showed one creation I made recently that is inspired by a character of mine that predates The Witching Hour.  The avatars represent me, but also express my interests, the topic of discussion, or some combination.  They are a creative way of protecting my identity by keeping actual photos of me less visible and vulnerable to bad actors.

In fact, a lot of my NightCafe generations are created with the idea in mind that they might be used to compliment other content when combined with it.  They are created with the idea of their use as digital graphics.  Not just personalized avatars, but things like banners and decorative images.  I'm not crazy about using stock photography that is far more generic.  A generator can help content creators who don't necessarily have the ability to hire someone to design their "brand" for them.  Especially if it's more a personal brand than a business brand.     

AI Controversy 

To close, let me say this.  Television producers, filmmakers, professional photographers, artists, recording artists, musicians, novelists, journalists and many more typically do not care for AI as a means of replacing craftsmanship in artistic expression.  Why should they?  These are people who worked very hard to learn and expand their craft, their art.  This human authorship should not be replaced by machines any more than human oversight should be.  What really irks them is when their work is used by AI modules without their permission or even their knowledge.  

So, what do we do about this?  Where do we draw the line?  At what point does it become exploitation or other forms of misuse?  How will the question of whether or not something produced by a generator is the result of an AI module trained specifically on their work be answered?  How would intellectual property owners go about proving this?  It's one thing if it's blatantly obvious, but what if it's not so obvious?  What if the generated result can be determined to have been created by an AI generator trained on many things besides protected works?  What if the result bears little resemblance to the works that the AI module that produced it might or might not have been trained on to begin with?

It's far easier to spot extreme bias in a response from AI chatbots like ChatGPT.  Asking it to analyze a statement made by an actual human but including specific words in the question that indicate the one asking already has a negative perception or opinion of the statement being asked about is what I mean by bias.  You know the phrase, "Play stupid games, win stupid prizes."  Except in this case, it would be more like, "Ask a smarta** question, get a smarta** answer."

In other forms of AI generation, will it be as easy to spot intellectual property used to train the module as it is to spot biased answers that clearly result from being asked biased questions?

I don't know.  But it is obviously past time to find out.