Are Deep Fakes Fanfiction?

[Deadman]

Recently came across this story about a singer who is open to sharing profits of AI deep fakes on music.

https://www.forbes.com/sites/martineparis/2023/04/24/grimes-tells-fans-to-deepfake-drake-her-welcomes-collaboration-with-ai/

So now I’m wondering, are deep fakes fanfiction? Technically it’s presenting things that didn’t happen as real. Like an AU story that you can find on this site.

Where do you come down on this?


 

Edited April 25, 2023 by Deadman

[Desiderius Price]

Got no issues with a creator explicitly authorizing AI deep fakes, and I can understand an estate helping a creator posthumously finishing a near-done work.  It’s the others, when brilliant idiots get involved, that I start to get concerned TBH.

So, is AI generated content a derivative work?  yes.  Fanfiction… can AI be considered a fan?

[Deadman]

23 minutes ago, Desiderius Price said:

Got no issues with a creator explicitly authorizing AI deep fakes, and I can understand an estate helping a creator posthumously finishing a near-done work.  It’s the others, when brilliant idiots get involved, that I start to get concerned TBH.

So, is AI generated content a derivative work?  yes.  Fanfiction… can AI be considered a fan?

 

People using the AI can be though.

[Desiderius Price]

7 hours ago, Deadman said:

People using the AI can be though.

Yes, but it’s the AI doing the writing.  Maybe treat it the same as somebody writing fanfiction on commission?  (ie, I’m exploring this question like you are, no definitive yes/no response… yet)

[BronxWench]

I have to say that I’m not a fan of throwing copyright out the window, which is really what deep fake is. If Grimes wants to explicitly allow deep fakes of her voice, that’s her privilege, but I don’t think it’s even remotely alright to assume everyone who creates art, music, or written work is fine with being deep faked.

Since I write for publication, I tend to focus on that, and it makes me extremely uncomfortable that someone would feel entitled to profit off what I’ve created. The majority of writers don’t earn enough to support themselves exclusively by writing. We’re not all NYTimes bestseller listed, or snapped up by the increasingly small number of large presses. We get a small enough slice of the pie as it is without having to worry about someone deciding to use AI to write a book using our characters, or settings, and taking our sliver of market share away from us.

It also begs the question of why anyone would prefer to have AI write for them. For me, much of the joy and satisfaction of writing is putting the words together and creating a phrase, or a sentence, or a paragraph that resonates. I like to find cadence, and play with language. Why would I give that over to an AI program? I’d be ashamed to call myself a writer if I didn’t actually write what I put out there, either for sale of for free on a site like AFF.

[Desiderius Price]

If we use the industrial revolution as guidance…. writing won’t die, there’s going to be artisans for centuries to come.   After all, we still use shovels to dig holes.  Maybe things like user manuals would get better written by AI?

I do see corporate greed getting involved.  Why pay a script writer when AI will do?  Why pay actors when you can deep fake the legends?  Only push-back from audiences would stifle that.  It will also tank the market...at least the AI generated stuff as it’s not copyright protected (for now), so nothing legally stopping me from firing up the latest napster to share it all.

For now, the human written stuff (at least the better stories) will still have heart and soul, something AI can only mimic, so there’s going to be a quality to our generated stuff.  For us, AI will, at most, be a writing aid … say you need a custom “mini-novel” or a fake advertisement within your story, or similar; maybe ai-assisted self-editing?  Editors might be in more peril than writers.

Me?  Enjoyment’s in the creation, especially those first drafts.  I don’t want AI doing that for me… if it’s doing all my thinking too, why do I even exist?  AI has that threat to snuff out the uniqueness to being sentient.

Loving the debate, wouldn’t want AI to simply create an “answer”, otherwise, it’d all be 42.

[BronxWench]

I’m not entirely sure AI could ever replace editors.

Until I went through the editorial process, I had only a rudimentary understanding of what was involved. There’s a huge amount of work beyond the simple grammar and spelling check, and looking for overuse of common words. A really great editor gets into your character’s heads, and will question you about why a character did something, when it seems like they might have been more inclined to do this other thing. Or they will spot chronological issues, or perhaps they just want to comment that they found a particular bit you wrote to be emotionally powerful. Whatever it is, AI can’t do that for you because it requires sentience and emotional involvement, neither of which AI possesses at present.

And the mice would like you to know that it took sentience in their test subjects to arrive at that answer. And fjords.

[Desiderius Price]

I’ve not gone through the editorial process, does seem that AI can definitely assist as you identify.  I’m guessing it’ll make further inroads into the process… full takeover?  Maybe, maybe not.  With time, I suspect that even the better-with-human traits will be bundled into an AI, and the self-pubs will lean on that.  After all, AI doesn’t have to do it better, simply has to do it cheaper.

Mice needed the sentience and fjords for the ultimate question AI … which of course was “what’s six times nine?”  (Which is 42 in base-13.)

[Wilde_Guess]

Hi, BronxWench and all.

1. I’m reasonably sure that that AI can not replace the editor or “beta-reader” process.  In the end, we’re people writing to entertain other people, for the exact reasons you listed.  The current crop of spelling/grammar checkers can’t even get things right on a consistent basis, particularly with any dialect of English.
2. Any time one is creating a “derived work,” by whatever means, they need the permission or license of the original work’s owner and creator in order to sell the work.  So, if the owner/creator of the original work denies permission, it’s game over, and it should be that.
3. The talk about AI, whether in creating prose or music, reminds me of the slogan typically posted in every fourth grade classroom in the United States since I was a fourth grader, I.E. nine years old.  “Just because you can doesn’t mean you should.”

Finally, since AI lacks a soul, the very best it can create is a pale imitation of the original work.  In the case of “continuing” a “story series” by a dead author, even real people with real talent have serious problems continuing a series without inevitably taking “ownership” of the later works in the series.  All the “post-Conan-Doyle” Sherlock Holmes stories are out there, and quite a few of them are great.  But they aren’t the works of Conan-Doyle, and could never pass for such.  Likewise, the post-Flemming James Bond works could never pass for Flemming’s work.  The authors who created the later works were not only not Ian Flemming, they were also so far removed from Flemming’s life experiences that they have no hope of truly creating a good true “James Bond” story.  Flemming set his Bond stories in the time in which he was writing.  That is, in the first near two decades after the end of World War Two.  So, while the Broccolis were able to write (or have written) quite a few good Bond works, they weren’t Flemming’s, and could never pass for it.

In the profoundly rare instances where an author can successfully continue a series work after the original author dies, it’s so much harder than they make it look.  If you are familiar enough with the authors, you can still tell the difference more often than not, even while enjoying both the “original” and the “continued” series. 

The most obvious example here is Robert Goldsborough’s continuation of the Rex Stout Nero Wolfe series.  Goldsborough has written seventeen full length stories in a style and quality near-identical to Stout’s forty-seven book (some books being double or triple novella in one volume) series.  While “both halves” of the Nero Wolfe series are great reads for those who like whodunits, you can tell the difference between Goldsborough and Stout, beyond being familiar enough with the works to know in advance which author wrote each story.  Goldsborough started writing Nero Wolfe for all the right reasons as well as to make money.  He was a very good author before writing Murder in E Minor in 1986.  But the post-1975 Nero Wolfe works, as good as they are, would never stand up to detailed scrutiny if Goldsborough had actually tried to pass them off as Rex Stout’s works.  Goldsborough openly claimed at least for the first stories to be trying to work up to the standard where he could, even while never actually trying to.  All this is from a talented real person, not a bot.  AI will only deny livelihood to real authors in exchange for making truly garbage literature for those who either don’t know any better or are lied to.

As an aside, even without AI/zombie bots in the mix, resurrecting an abandoned work that really existed is incredibly hard.  Harper Lee’s Go Set a Watchman was already in the editing/rewriting process in the late 1950’s when her editors convinced her instead to create To Kill a Mockingbird from expanded versions of the flashbacks within Watchman fleshed out and stitched together.  Mockingbird was a mammoth best seller and Pulitzer Prize winner.  Lee never wrote another book, and for most of her life refused to take up finishing Watchman.  At the very end of her life, a copy of her 1950s manuscript turned up, and she reluctantly gave in (others have claimed her consent was obtained under less honorable conditions) to having the work published, but was well beyond the ability to do any further work on Watchman at all.  Thus, the people “cleaning up” Watchman for publication tried to “fix” it without Lee’s assistance.  And in the end, despite Watchman having good bones, the book still read like what it was, I.E. a first draft that was poorly polished at best, and by people more interested in cashing in than they were providing an entertaining story.  Thus, it received slightly favorable but mixed reviews.  One example of “continuity” glaring enough to indirectly make it into Wikipedia (from the LA Times) was the winning of the trial (in Watchman) that Atticus Finch lost in Mockingbird.

The Watchman we got was better than no Watchman at all.  But it would have been so much better if Harper Lee had truly gone back to writing and polishing Watchman herself ten years or so earlier, when she still had the ability to fix it herself and bring it up the the same standard as Mockingbird.  And even if someone tried to use an AI to “fix” Watchman, the end result would actually have been worse.

In the end, whether we were ultimately created by a Supreme Personage or a Cosmic Dice Roll, we are people, and not machines; and machines can not become people.  Genuine creativity can not be created mechanically.  This ability to either create or sincerely admire true creativity is part of what makes us people instead of machines.

Edited April 26, 2023 by Wilde_Guess
minor grammar corrections

[BronxWench]

1 hour ago, Desiderius Price said:

I’ve not gone through the editorial process, does seem that AI can definitely assist as you identify.  I’m guessing it’ll make further inroads into the process… full takeover?  Maybe, maybe not.  With time, I suspect that even the better-with-human traits will be bundled into an AI, and the self-pubs will lean on that.  After all, AI doesn’t have to do it better, simply has to do it cheaper.

Mice needed the sentience and fjords for the ultimate question AI … which of course was “what’s six times nine?”  (Which is 42 in base-13.)

An AI will never provide the insight of a human editor, who can identify what human readers will want to read. It’s pretty basic, actually. Spell checkers and grammar checkers, no matter how sophisticated, can’t read for content like a human. What you’re thinking about is the proofreading stage, which comes at the very end of the process, after the editor and author have beaten that gruesome draft into something resembling a real book.

(And even then, the author can overrule the proofreader because conversation is not grammatically perfect, after all.)

[Wilde_Guess]

Hello again, all.

I didn’t actually try to answer the original question.  Here’s my answer.

Not just no, but fuck no. 

AI Deep Fakes are the antithesis of Fanfiction.  AIs are not people and thus can’t admire genuine creativity; and many who would use AIs for such a purpose are little better.

In my previous post, I mentioned the works of Robert Goldsborough.  Those works were properly licensed in advance with the Rex Stout Estate, and were successfully published and critically received on their own merits.  Ultimately, though, Goldsborough’s Nero Wolfe was the ultimate pinnacle of fanfiction—one that was so good that it went straight from the typewriter to the commercial editor, and then to the publisher and commercial success.  Without the creativity and love Goldsborough (and his mother, whose love for Stout’s Wolfe inspired his efforts) had for Stout’s works, plus Goldsborough’s own talent and creativity, those seventeen (and counting) books would never have happened.  AIs are neither talented or creative, thus an AI couldn’t possibly create anything good; here or elsewhere.

Thanks.

[BronxWench]

10 minutes ago, Wilde_Guess said:

 

In the end, whether we were ultimately created by a Supreme Personage or a Cosmic Dice Roll, we are people, and not machines; and machines can not become people.  Genuine creativity can not be created mechanically.  This ability to either create or sincerely admire true creativity is part of what makes us people instead of machines.

The above is exactly why I don’t think AI has a place in actual writing, even if it’s fan fiction.  The New York Times did a comparison of AI-generated essays/writing prompts versus those written by actual children.  (https://www.nytimes.com/interactive/2022/12/26/upshot/chatgpt-child-essays.html ) Judy Blume, the author, participated. Give it a try, and see how well you do.

[Desiderius Price]

14 minutes ago, BronxWench said:

An AI will never provide the insight of a human editor, who can identify what human readers will want to read. It’s pretty basic, actually. Spell checkers and grammar checkers, no matter how sophisticated, can’t read for content like a human. What you’re thinking about is the proofreading stage, which comes at the very end of the process, after the editor and author have beaten that gruesome draft into something resembling a real book.

(And even then, the author can overrule the proofreader because conversation is not grammatically perfect, after all.)

I tried grammatically perfect dialogue – it sucks and tends to be overly wordy, unrealistic.

Concur that human editors will remain best, the gold standard, for years to come.  I’m thinking about sleazy/cheap self-pubs or similar trying to cut corners.  The issue is when about everybody starts doing it, degrading the industry as a whole.  When you apply this cost-cutting mentality to any other industry (appliances, tools, whatever), the middle tier gets gutted because the white box products are good enough for 90% of the market; the upper tier stratifies, before they too, are forced (due to finances) to use those white-box products.  All that’s left are the artisans with $5k tables, or the IKEA (& clones) with the $100 tables.

[Wilde_Guess]

Hi, BronxWench, Desiderius Price, and all.

9 minutes ago, BronxWench said:

The above is exactly why I don’t think AI has a place in actual writing, even if it’s fan fiction.  The New York Times did a comparison of AI-generated essays/writing prompts versus those written by actual children.  (https://www.nytimes.com/interactive/2022/12/26/upshot/chatgpt-child-essays.html ) Judy Blume, the author, participated. Give it a try, and see how well you do.

It’s behind the paywall.  I’ll pay for a subscription to the NYT after they pay repartitions for their part and participation in the Holdomor, and no sooner.

Desiderius Price, I agree with you on “grammatically perfect dialogue.”  It’s also one of the things I have to watch in my own writing.  While my grammar is by no means perfect, it is habitually quite good.  So, I have to watch myself (with varying degrees of success) to not have my characters speak above their “linguistic weight.”

Thanks.

[Desiderius Price]

33 minutes ago, Wilde_Guess said:

Desiderius Price, I agree with you on “grammatically perfect dialogue.”  It’s also one of the things I have to watch in my own writing.  While my grammar is by no means perfect, it is habitually quite good.  So, I have to watch myself (with varying degrees of success) to not have my characters speak above their “linguistic weight.”

Something I have to watch out too, seeing that on my current story I’m reworking.  This is a spot where AI could be useful, help you assess “linguistic weight” of the dialogue as you type along and/or revise

 

1 hour ago, Wilde_Guess said:

In the profoundly rare instances where an author can successfully continue a series work after the original author dies, it’s so much harder than they make it look.  If you are familiar enough with the authors, you can still tell the difference more often than not, even while enjoying both the “original” and the “continued” series.

For posthumous… first case I thought of is when an actor dies mid-production, might be able to fill in missing pieces, release the movie.  Depends on the amount needing deepfake, obviously.

[Wilde_Guess]

Hi, Desiderius Price.

12 minutes ago, Desiderius Price said:

Something I have to watch out too, seeing that on my current story I’m reworking.  This is a spot where AI could be useful, help you assess “linguistic weight” of the dialogue as you type along and/or revise

In your example above, AI could be a tool to help an author.  So could a beta reader/editor, with likely better results.

 

12 minutes ago, Desiderius Price said:

For posthumous… first case I thought of is when an actor dies mid-production, might be able to fill in missing pieces, release the movie.  Depends on the amount needing deepfake, obviously.

Some less dubious “trickery” has been used in the past for circumstances like you’ve listed.  However, AI is inherently untrustworthy, since by its very nature it will “taint” that actor/actress’s entire performance in that movie.  It’s far better to edit the screenplay to downplay the dead performer’s character, or reshoot all the scenes that performer appeared in with a different performer.  AI is already being used excessively to the detriment of real people, whether those people are the human creative people or the audience for which those people would otherwise perform.  This says nothing about those cases where real people should/would not perform at all.

Thanks.

[BronxWench]

Apologies for the paywall thing. I forget about that all the damned time, probably because I’m on the far side of middle aged. 

[Desiderius Price]

1 hour ago, Wilde_Guess said:

In your example above, AI could be a tool to help an author.  So could a beta reader/editor, with likely better results.

Even before I give a manuscript to a reader/editor, I want the thing to be as good as I possibly get it.  To have an AI that puts red squiggly marks beneath questionable things… it draws my attention, and so long as I can tell it to “ignore”, I’m fine with that level of AI.  So, having AI assisting in showing where there might be a problem, that I should double check, that’s fine to me.  More comprehensive the AI, the better...at this stage.  Might it eventually replace the need for an editor in all but the most serious of publications?  Well, I’m not hiring out an editor for what are ... web postings.  As AI improves, how “good enough” will it be before it can handle … say magazine or cheap-romance level of editing? 

Ditto for the writing… not saying AI will be better, but when will it be “good enough” for somebody more interested in quantity over quality, but still claim it as their own?

[BronxWench]

1 hour ago, Desiderius Price said:

Ditto for the writing… not saying AI will be better, but when will it be “good enough” for somebody more interested in quantity over quality, but still claim it as their own?

Therein lies the issue for me in my role as moderator. It’s not the work of the posting author, no matter how you spin it, and therefore it’s not allowed on AFF per our Terms of Service. Other fiction archives might have a different opinion, but because we are an age-restricted site, we are likely to come under closer scrutiny in general, and there’s no sense in compounding things by letting people post work they didn’t actually write themselves.

Maybe someone needs to start a fiction archive/website for AI-generated fiction? Bring a little transparency to the matter, so to speak?

[Desiderius Price]

3 minutes ago, BronxWench said:

Maybe someone needs to start a fiction archive/website for AI-generated fiction? Bring a little transparency to the matter, so to speak?

Heard foxnews recently let go one artificial air-head…

 

[BronxWench]

6 minutes ago, Desiderius Price said:

Heard foxnews recently let go one artificial air-head…

 

Do NOT invoke that which you cannot banish!  Sweet merciful gods of my fathers… 

[Deadman]

Well I’m also sorta thinking in terms of deep fake videos.

https://www.forbes.com/sites/danidiplacido/2023/04/25/avengers-director-joe-russo-has-a-bleak-vision-of-ai-generated-movies/

Apparently some people believe that movies will be entirely AI generated in a few years.

But obviously, there’s also adult deepfakes where celebrities get put into sexual situations.
 

[Desiderius Price]

7 minutes ago, Deadman said:

Well I’m also sorta thinking in terms of deep fake videos.
https://www.forbes.com/sites/danidiplacido/2023/04/25/avengers-director-joe-russo-has-a-bleak-vision-of-ai-generated-movies/
Apparently some people believe that movies will be entirely AI generated in a few years.
But obviously, there’s also adult deepfakes where celebrities get put into sexual situations.

While I don’t think we’ll go 100% AI …. yet.  I’d expect to see AI creeping in with fake actors for those “extras”.  Even if the AI costs the same, less problematic, less risk to the schedule, than dealing with the hassles of “people”

As to deep faking sexual situations…. yeah, legal or not, might not be contained to “adult” actors being faked in either.