#forgotten snack | Explore Tumblr posts and blogs

britcision · 2 years ago

Text

Happy WIP Wednesday friends! Not a huge deal for you today, but I figured I’d drop the Flashback and give you the last piece in the “Bruce Puts His Head In His Butt” for the night!

(Bruce is tranq’ed by Alfred minutes after the call ends and is put to bed. In my heart. He might actually walk himself up but we all know it’s Alfred’s glare that makes it happen)

Just a taste of chapter 14 of Dead and Loving It, you can find the fic on AO3 or from my pinned post which is the latest chapter, but links to the first and all subsequent chapters are in each post!

———————

A Good Excuse To Be A Bad Influence

Jason was actually on his way to bed on time for once in his life, the early end to patrol and lack of crime lord duties giving him a chance to get a full five hours sleep.

He should have known he wouldn’t get lucky two nights in a row; Constantine wasn’t around to distract Bruce anymore.

He’d contemplated not answering. Contemplated trying not to shoot Bruce in half an hour if the fucker showed up at his window.

The pit growled.

It was the worst thing he’d ever heard. The worst thing he’d ever felt. And he did feel it, vibrating in his very bones.

It sent shivers creeping up and down, muscles tensing as if to run away from something inside him.

He answered the call, hoping it wouldn’t show in his voice.

“What.” Flat, unfriendly. Not encouraging conversation.

“You didn’t come to the cave.” B’s voice was equally flat, but in his case it sounded like a condemnation. An accusation.

Jason gritted his teeth.

“I have shit to do in the morning. Make it quick,” he snapped, giving his bed a glare it definitely didn’t deserve.

His pillows had never done anything to hurt him.

There was a momentary pause before B audibly decided not to push it.

Good.

Jason was in a mood to bite.

“We have intel on the Infinite Realms. I’ve sent the report. You need to stay away from Danny Fenton, for your health,” B said, still cold, still clinical.

Like he didn’t care. Like what Jason wanted didn’t matter.

Jason’s grip tightened and the phone case cracked.

“Yeah, no. Fuck off.” He spat the words, adding “get new phone” to his list of chores for the morning.

He’d been doing so well with this one. Of course Bruce had to ruin it.

At least the old man didn’t seem surprised by his reaction.

“Jason. It… he. His abilities may affect your condition,” he said slowly, sounding tired. Old.

The pit snarled, sensing weakness, and Jason kinda wished he was still lost in its rage. Back when he was, it was easy just to hate those moments.

B showing signs of humanity fucking hurt.

“He is. He’s making it better,” he shot back, brooking no argument.

“We don’t know that, Jason. Please, just… just for a few days. Until we can talk to the League, understand what he’s doing to you.”

Was.

Was that Bruce begging?

It froze something small and soft in Jason’s chest, stuck him in place. And did nothing to stop the flood of icy rage from filling him up.

Filling his chest, crushing his lungs, making it hard to breathe. Because of course, anyone and everyone else’s judgement was worth more to the man than Jason’s.

Begging Jason to listen to him, when he would never, ever, fucking ever listen to Jason. When it didn’t fucking matter if Jason begged.

“And why the fuck would the League know better than a doctor from the Realms?” He finally snapped, ignoring the way his throat tightened.

There was a long silence.

“A doctor?” Bruce asked softly, his voice still so flat and emotionless that only his kids could have read the confusion. Jason rolled his eyes.

“Danny brought me to a doctor. I’m gonna be fine,” he ground out reluctantly, part of him resenting Bruce’s constant insistence on knowing everything.

But… well. If it got the guy off his fucking back.

There was a long silence, one that Jason was fully aware B was likely spending working this new information into his latest paranoid fantasy.

Jason seriously considered just hanging up and going to bed. He was about to do it when Bruce spoke again.

“Would this doctor be willing to speak to the League?” And there it was again, Batman voice, clinical and distant and always, always fucking suspicious.

Jason rolled his eyes harder. With emphasis. Willing to be interrogated by first the Justice League and then separately also goddamn Batman.

Actually, now that he thought about it, he was pretty sure B wouldn’t get anywhere with Frostbite. Frostbite took his work seriously and was, yeah, king of a full realm of yetis.

None of Bruce’s pointed silences, menacing looming, or vague growls would bug the guy who got Danny through Fucked Up Ghost Puberty.

(And would probably be helping Jason through his own Fucked Up Ghost Puberty… joy of joys.)

It might actually be fun to see him try. If just being here wouldn’t put Frostbite in danger, because hell fucking no that wasn’t happening. The guy may not be his king but Jason would still die first.

But of course, in all his paranoid bullshit about the Realms influencing Gotham, B had somehow conveniently missed what America was doing to the Realms.

Like Jason hadn’t even done the full write up.

“Not while the fucking League are required to hand him right to the US government for torture and experimentation. Which, by the way, did you read my report on the Anti Ecto Acts?” Jason asked sarcastically, doing his very worst fake concern.

And again he was met with silence. Fuck, maybe Bruce hadn’t read it. Jason had dropped it in the day before all this gala bullshit had started, and it had been a busy two days since.

Maybe B deadass hadn’t put the pieces together. Might as well hammer it home for him.

“You know, the one that says you, me, Cass, and Damian are all non-sentient because we’ve been exposed to the pits?” Jason added, eyes narrowing.

Which wasn’t technically true, since it was the resulting liminality and ability to process ectoplasm that made them count, but Bruce didn’t need to know that yet.

Finally he spoke again, voice gruff and clipped.

“I’m looking into it. But for now, Jason, please-” he said again, the cover of Batman beginning to slip.

But Jason was done. No fucking chance Bruce was giving him orders when he hadn’t even bothered asking for Jason’s opinion.

He wanted to spout off about dangers of the Infinite Realms after talking to some wet paper bag of a man who hawked his soul like it was a pokemon card. Hard pass.

And even after hearing that Jason knew what was going on a damn sight better than Bruce did, he still wanted to push him around?

Fuck that.

“Sorry B, legally non-sentient, guess I can’t be blamed for my actions,” he drawled, then turned his phone off and dropped into bed.

He had a lot of shit to do before picking Danny up in the morning.

——————-

Jason will be using “legally non-sentient” as an excuse long after the laws themselves are repealed, and just you fucking wait until Damian hears he can try it too 😏

Sorry Bruce, Damian can’t socialize today, he’s legally non-sentient and can’t be blamed if he bites someone

Tag List: @welcometosasakiworld @kyrianclawraith @someonebored0100 @stealingyourbones @starkcravingmad @frostedthroughghost @akikkobara @rainbowbunny0159 @littlefeather345 @violet-catsarelife @serasvictoria02 @wolfjackle @blacksea21090 @secretdestinywerewolf @anime-hipster-the-amazing @undead-essence @skitscratched @blackroserelina @snoodly-boop @trickerdi @mayoota-blog @xysidhe @idkmrpianoman @little-apricot-the-writer @chaoticmistake @the-legal-shipper @bun-fish @aroranorth-west @demon-cat-goes-woof @perfectwastelandcreation @onyxlightdragon @larks-and-katydids @peachesandcreamfemboy @jesus-camp-the-sequel @may-rbi @mothman-the-mothman87 @viyatrix @stargirl1331 @idfk-man10 @thedepressedrobin @skulld3mort-1fan @rootsmudge @ravenshadow17 @cankoking @phantom-dc @mentalcarebear @magic-pincushion @redamancyardor @lyra689 @itsparadoxlacuna @alcorbearson @asphyxia778

#danny fenton dead and loving it #dp x dc #wip wednesday #chapter 14 for sure this time #probably #tentative title: a good excuse to be a bad influence #just a lil snack of angst from jason and bruce #the thing about requests is that sometimes the answer is ‘no’#and you have to just accept that #bruce may have forgotten harley is still in town but I HAVE NOT 😈😈😈#harley ain’t vlad we remember she exists

321 notes · View notes

yvesdot · 2 years ago

Text

Can ChatGPT Do My Job? Initial Musings on AI

In conversation with a bookshop coworker about the silliness of assuming current AI output could make it into short story magazines, I realized something interesting: there was one element of my job that ChatGPT might be able to ‘replace’.

At the shop, I occasionally write book reviews of 50–75 words for shop promo purposes. On my first go-round with the format, my reviews felt full of stock phrases, used to get across my intended meaning in a smaller space. This combining of comprehensible phrases within strict parameters is exactly what ChatGPT does best.

So, could ChatGPT write my book reviews for me?

Some samples of my book reviews, all available on my GoodReads:

Detransition, Baby by Torrey Peters

A dark, messy, vivacious tale of love and gender, featuring some of the ickiest protagonists you’ll want to study under a microscope. Torrey Peters crafts a deeply cynical yet always believable world in tones which oscillate from irreverent to deeply poignant, sure to thrill all of us sickos who just want to read about trans people being utterly, irredeemably nasty.

Big Tree by Brian Selznick

Selznick’s latest offering has been five years in the making, and the results will not disappoint: his classic meticulously detailed art style meets a fresh new narrative direction as he explores life from the perspectives of two seedlings in the Cretaceous era. Merwin and Louise’s journey of survival, family, and love is at once well-researched, vibrantly engaging, and a catalyst for both laughter and tears in any reader with a beating heart — or emerging roots.

We Do What We Do in the Dark by Michelle Hart

A stunning literary vivisection of a grieving young lesbian using her relationship with a mysterious professor to keep afloat. Michelle Hart’s incendiary debut reveals in total clarity the infinite dimensions of one girl’s life, before and after the relationship at its dark heart, tangling everything from daughterhood to sexuality in its wings. A glittering, underrated must-read.

White Cat, Black Dog: Stories by Kelly Link

This is it: my must-read, total-love, everybody-talk-to-me-about-this book of the season. Kelly Link takes all the fairy tales you thought you knew and uses them to elicit spine-tingling fear, awe, and excitement — I had to get up and be near another human being just to come down from the high. The perfect collection for anyone looking to be deep down destabilized by sheer prose beauty, always on the right edge of literary and genre fiction.

(For the record, I now feel quite satisfied with these reviews.)

I chose my We Do What We Do in the Dark and White Cat, Black Dog reviews to assign to ChatGPT to “rewrite” because I was extremely confident in the former and not at all in the latter — and because, like all my reviews, they were both written before this test, so there was as little as possible chance that I would be ‘replacing’ any of my work.

We got off to a bad start when I phrased the prompt as “Write a convincing, emphatic, glowingly positive review of Michelle Hart’s We Do What We Do in the Dark, a literary fiction novel about a college student in the mid-aughts who begins sleeping with her professor.” Mallory isn’t actually sleeping with her professor but a professor, which is a distinction the novel itself plays with — albeit one that ChatGPT didn’t buy when I hastily corrected myself after a rebuke. Ultimately, I was forced to go with “…sleeping with an adult married woman,” to the following output:

Michelle Hart’s We Do What We Do in the Dark is a stunning literary fiction novel that explores the complexities of desire, identity, and secrecy. Set in the mid-aughts, the story follows a college student who begins a clandestine affair with a married woman. Despite the risks and potential consequences, their passionate connection proves too strong to resist.

Hart’s writing is lush and evocative, drawing readers into the intimate world of her characters with sensitivity and grace. Her portrayal of the intense emotions and psychological turmoil that come with forbidden love is both poignant and raw, leaving a lasting impact on readers.

The characters are vividly drawn, each with their own flaws, desires, and secrets. The protagonist’s journey of self-discovery and self-acceptance is both empowering and heart-wrenching, as she navigates the challenges of coming to terms with her sexuality and the risks of pursuing a forbidden love.

Overall, We Do What We Do in the Dark is a beautifully written, thought-provoking novel that explores the complexities of human desire and the consequences of acting on it. Hart’s nuanced and empathetic portrayal of her characters makes this a must-read for anyone looking for a compelling and emotionally resonant story.

When asked to rewrite its work between 50–75 words:

Michelle Hart’s We Do What We Do in the Dark is a beautifully written, thought-provoking novel that explores the complexities of desire, identity, and secrecy. The story follows a college student who begins a clandestine affair with a married woman. Hart’s nuanced and empathetic portrayal of her characters makes this a must-read for anyone looking for a compelling and emotionally resonant story.

Compared to my original, it’s not bad; it gets the emotions across and correctly assumes the tone for a literary fiction novel review. Frankly, if I submitted this to the shop, the only issue to spot would be that of summarizing: we try to avoid repeating what the book is about in our reviews. Cutting that line, it’s a nice, if short, review of a little under 50 words. (Though I miss my vivisection.)

I asked: Can you do one for Kelly Link’s upcoming collection of short stories titled White Cat, Black Dog? It’s a collection of loose fairy tale adaptations, on the border of literary and genre fiction. Each story is illustrated by Shaun Tan. Please also keep it between 50 and 75 words :)

ChatGPT:

Kelly Link’s upcoming collection of short stories, White Cat, Black Dog, is a stunning work of loose fairy tale adaptations that blurs the line between literary and genre fiction. Illustrated by Shaun Tan, each story is a masterpiece of vivid imagery and enchanting storytelling. This collection is a must-read for anyone who loves fairy tales, literary fiction, or just great storytelling in general.

Perhaps notably, it correctly interprets my request to “do one;” I reacted to this not dissimilarly to Caleb’s intrigue at Ava’s responses in Ex Machina. It understands ambiguous phrasing! How novel!

And, of course, it did a great job writing its review. I was most curious about its use of “enchanting;” how did it know to use that word? So I asked for its “thought process.” ChatGPT told me:

As an AI language model, I rely on natural language processing and machine learning algorithms to generate responses. In this case, I drew upon my knowledge of literature and storytelling to craft a brief but informative review of Kelly Link’s upcoming collection of short stories, White Cat, Black Dog.

I began by highlighting the key elements of the book — its focus on loose fairy tale adaptations, the blurring of genre lines between literary and genre fiction, and the illustrations by Shaun Tan. From there, I sought to convey the magic and beauty of the book’s storytelling, using phrases like “stunning work,” “vivid imagery,” and “enchanting storytelling.” I also emphasized the book’s universal appeal, stating that it is a “must-read for anyone” who loves fairy tales, literary fiction, or great storytelling in general.

Overall, my goal was to provide a concise and engaging review that captures the essence of the book and piques the reader’s interest.

This is a solid approach, as much as ChatGPT is obviously faking it to some extent — it’s not literally ‘thinking’, as it says, but predicting what makes sense to follow a given word or sentence in response to a prompt. Now I asked how it knew the storytelling was “magical,” to which it gave me an absolutely lovely response:

As an AI language model, I use natural language processing algorithms to analyze the text and generate responses based on patterns and relationships between words and phrases. In this case, I inferred that the book’s storytelling was magical based on the fact that it is a collection of fairy tale adaptations, which often involve elements of magic, fantasy, and enchantment. Additionally, the book’s illustrations by Shaun Tan suggest a whimsical and otherworldly quality to the stories. Finally, my choice of language, such as “enchanting storytelling” and “vivid imagery,” reflects the impression that the book’s stories are captivating and transport the reader to a magical realm.

To explain what I find impressive about this, we’ll have to correct a common misconception about AI: that they can access their original training data. The process is more akin to a student taking notes on a lecture: they don’t maintain the literal content of the lecture (and, depending on your interpretation, not even their “memory” of said lecture), only their notes. This is what they work from.

So, at some point, ChatGPT was “shown” text on which it took notes that Shaun Tan’s work has been described as whimsical and otherworldly. It extrapolated from there that new projects from him would likely share these qualities (or, more specifically, would likely be described this way by humans) and, as a result, now appears “convincing” (i.e. “like a human”) to me.

Where does it get this information (or any information) from? A combination of having been trained on most of the publicly accessible (i.e. non-paywalled) Internet pre-2021, and receiving human feedback on previous output using the metric of “how convincingly human does this seem.”

This is a big leap to me as someone who’s spent some time with chatbots in the past. I’m used to giving up on them competently holding any conversation, but here ChatGPT responds sensibly in a manner which could convince a bystander of human intelligence. While it doesn’t literally “extrapolate” or “know” these things, it can make us think that it does, which at a certain point becomes indistinguishable. (Does a chess computer know it’s playing chess? Does that matter?)

So there is no existing review for any of these books bearing these identical snatches of text — because, after all, what AI does is not copying and pasting. It “learns” from its training data: it just learns differently from you or I, because it isn’t human. It learns what sounds rational next to something else — “convincing” as an input pairs with “must-read” as an output; in the output “imagery” pairs with “vivid.” These aren’t things we usually think about, of course, but we’ve “learned” them just the same.

Furthermore, the text is generating, word-after-word, on the fly. (Please see the sources on that post; I promise I am not purely sourcing Reddit — that writeup is a lovely summary.) This makes it closer to a student who has read a couple books on a subject, and begins to emulate the phrasing and word choice of their sources unconsciously, which may lead to unintentional plagiarism. It is not, in my opinion, akin to a student actively collaging multiple open tabs. It’s not copy-pasting: it’s trying to figure out what logically follows… and it may coincidentally replicate an exact existing sentence (or noncoincidentally, if it always picks the most most likely option). What logically follows “George Washington was the”? “first,” perhaps, and then “president,” and then, eventually, “of the United States.” Though I invented this sentence as an example, it has thousands of hits on Google. Did I plagiarize?

(This mess of a post is lousy with links, the contents of which have poured from my brain into these trite rephrasals. Do I plagiarize?)

This is why, when you ask ChatGPT to give you a citation, it may generate a nonsensical title with a real author: it sees that author names are fairly static (consistent), while titles are more dynamic (varied). It is literally writing you a convincing citation. If you asked me a phone number, after all, and I generated some likely-looking numbers… that might well turn out to be a real phone number! It is making things up, which requires, of course, the capacity to “make.”

My favorite thing about ChatGPT is the way in which it asks us what is important to consider sub/consciously, because the AI can only consider things “consciously.” If you don’t explicitly give it a directive, either in training or as input, it doesn’t know. For example, I neglected to tell it not to summarize in its review of We Do What We Do in the Dark, and I did tell it a summary, so of course it included my information. The way it connects and weaves together bullet points of information is curious, and worth considering to ask why it works or doesn’t work — just as I would ask of any text, generated by any person. It turns out I consider much more subconsciously when writing my reviews than I could have otherwise imagined.

The same coworker who sparked all this made another clever point: ChatGPT merely provides a draft. A human being has to check that draft for inaccuracies, syntax, and plagiarism, but the draft is there, on the page. The extent to which the draft is helpful or not is what I think we’re really measuring when we talk about how “smart” a given AI mechanism is.

Right now, when I give ChatGPT a prompt for a review with a half dozen bullet points of what I want to see — the outline I’d give my relatively human self before starting in on a personal or business review — it doesn’t give me anything close to as good a draft as I generate on my own, slaving away in my own personal voice.

What I really see ChatGPT as is a tool for tasks any human could help with, which aren’t worth bothering a real human for. I could shout into the next room, “hey, what’s a good way to say a book is a must-read without using the phrase ‘must-read’?” but maybe I don’t want to bother my housemates — or maybe I don’t have them. Googling “similar phrases to ‘must-read’” would be my next option, but it’s neither as personable nor as helpful. ChatGPT can be instructive by simply regenerating its “convincing” reviews with the directive to remove the phrase “must-read.”

The task must also be something where the effort itself is not the point. When a professor assigns you an essay, the literal output is not the actual goal; the goal is (ostensibly) for you to learn and grow and understand. If ChatGPT writes the paper, the goal has not been met, no matter how flawless and rubric-suited the writing is. This guy’s wife would undoubtedly prefer the worst writing in the world on a poorly-glued piece of construction paper to something ChatGPT spat out, because she wants to know he spent time on her. Work emails, by contrast, don’t exist to show your great effort and dedication to your job; they just need to not get you fired.

ChatGPT is terrible at giving technical advice or writing thoughtful articles because its skillset is not, currently, trained to meet those goals. Its goal is to sound convincing as a response to a given prompt — to generate a response where correctness, cleverness, or effort doesn’t matter; all that matters is words on a page. Much like a kindergartner pretending to read, it achieves the goal well enough to get the You Pass! sticker, but ultimately fails at what it is really being asked to do. @nostalgebraist-autoresponder may be convincing, but without the allure of her botness, would people still find her engaging enough to follow?

(Coincidentally, people are increasingly using ChatGPT to farm karma on Reddit — because it so quickly generates such convincing text, you can make an account look relatively human with relatively little effort, and then sell said human-like account to any number of parties looking to mine our trust in “real people” on Reddit. One example. Another example.)

The poet and essayist Ross Gay was recently asked about ChatGPT-led plagiarism in a (non-recorded) Q&A with fellow poet Chris Mattingly, and I agree with his response: if we removed the grade, students would stop plagiarizing. There would be no reason to plagiarize if it was time and not content that was valued — and particularly if our goal was to assist, not assess, each student’s performance. Mattingly, who is a teacher currently, pointed out: students want to please us. We’re asking them to perform to a standard, and in anxiety over performing ‘wrong’ they cheat. They’re afraid. Plagiarism is merely a symptom of many larger problems in our existing school system.

Copywriting is much the same. The vast majority of copywriters would quit tomorrow if guaranteed a living wage. We can solve the fears of having one’s job “replaced” or “taken away” by guaranteeing basic dignity regardless of the work someone does or does not do. An added bonus? Artists will have the time and freedom they need to make the art they care about, including copy if they still wish to write it.

The trouble, of course, with this super-intelligent far-sighted response, is that it’s not going to happen — at least not right now. Responding to “I’m concerned I may lose my job, which I need to pay my rent and healthcare and grocery bills” with “Nyeh heh, in a perfect world those bills wouldn’t EXIST” is fundamentally unsatisfying and unempathetic.

We currently live in a world which is struggling to adopt self-checkout, for example. Almost everyone I’ve spoken to prefers it for a variety of reasons. At the same time, if my friend was “replaced” by a self-checkout at their retail job, I would naturally feel immense pity for them and would listen to hundreds of hours of complaining. Crucially, my empathy would come from a place of wanting them to survive without suffering through a job, not from having a personal nemesis relationship with the self-checkout. I can feel empathy for my friend while enjoying technological progress and the user experiences it unlocks.

Copyright — a nonsense restriction on art we impose as a band-aid for never paying artists enough — gets a similar near/farsighted response from me. I think copyright should evaporate right now. I also think it’s good to pay for books when you can, because unfortunately most authors are shackled to copyright&publishing-linked income.

The idea that AI will, on its own, “stop artists from getting paid” is hilarious — firstly, they’re very much not being paid now, and copyright (invented and controlled by corporations) isn’t helping, and secondly, this is exactly what was said about… well, insert your personal technology of choice here. Now that people can take photos, nobody will go to portraitists! Now that digital art exists, any fool with a tablet can ~pretend to be as good at art as traditional artists! Photoshop is making unsexy women look sexy!! Technology is bad, fire is scary, and Thomas Edison was a witch.

(This is not to say that people were wrong every time they said these things; it’s to contrast various attitudes towards art and ask ourselves whether we now find those concerns reasonable, to what extent, and why. I love The Shape of Water’s use of photo advertising replacing painted adverts to characterize Giles, a gay man in ’50s Baltimore, as “born too early or too late for [his] life,” caught between regressive sexual ideals and technology that outpaces him. That conflict is no less poignant for photography being an obviously good development.)

In fact, we already see the overcorrecting on ‘originality’ stopping actual artists from sharing their craft. Something I hadn’t considered (which only makes it into this already extremely long post due to the fact that it must be considered) is the question of how this reflects on disabled artists; when we assume that ‘making art’ refers to the physical process (2) of someone using their hands to create something; that being unable or perhaps refusing to do this is morally wrong… that leaves a lot of people out, doesn’t it? Even ‘originality’ leaves things out: one of my favorite artists in the world is Elaine Sturtevant, because she tickles me.

(Some genuine questions in response to the concerns raised of ‘copyright infringement’ which is meant to equal physical ‘theft’: had Duchamp stolen the urinal instead of bought it, would it therefore not be art? Would it only be alright because a urinal is “not art”? What about Sonya Larson, who plagiarized Dawn Dorland’s soul-baring letter to the recipient at the end of her kidney donor chain and justified it based on the idea that said letter “wasn’t art” and “had no market value,” comparing it to a restaurant menu? Do these concerns apply to collage artists? To found poets? To sampling? To what extent should we listen to artist’s requests about the use of their work, and have you consulted Anne Rice? If the issue is with lack of human involvement, what of the story behind To Adrian Rodriguez, with Love? Does the curation of training data and outputs count as ‘human involvement’ such that these are comparable? How communal or individual is a given AI art method? What “AI art” methods have we not been discussing [e.g. models trained by one artist on their own work]? What do we owe for influence?When should or must we ask permission? To what extent is this about ‘copyright’ vs. kindness? How, where, and why do those boundary lines blur?)

Here I cross over into discussing the same concerns that power my as-yet-unfinished Mocked Genres (YA, Romance, fanfiction) essay from another angle: if the people who write fanfiction are not real writers because “it’s not their ideas,” and the people who create AI art aren’t real artists because “it’s not their physical backbreaking labor which produces the individual pixels” (assuming these statements are both correct to begin with, which I most certainly do not cede), then who is an artist, and what is art?

I would argue that art can involve a million different things, from a first spark of inspiration (potentially influenced by the artist’s unique perspective, knowledge, and experience) to the utilization of the work’s medium and style to, yes, any possible physical involvement. Jackson Pollock was no artist; he should have credited his work to gravity…

(Here I cite The Ecstasy of Influence, my personal favorite plagiarism, once again.)

And I admit: I don’t know what we should do to copyright right this second. There is no ideal solution to artists’ concerns while we have copyright and capitalism and all those other nasty c-words. This is a nice start, though.

All this means, to me, is that we need UBI. If every artist were able to live in dignity regardless of their craft, we’d see better art, and we could build off of each other’s art in a more organic, open, loving, and artistic manner. Art is not made in a vaccuum. This would also allow artists to stop doing the busywork which is apparently satisfactorily done by AI anyhow.

(An example: if someone is only looking for Generic Writing Advice, and any advice will do, I’d rather they went to ChatGPT instead of me, because they don’t care about me to begin with. I also wish that I could be paid a living wage so that I wouldn’t have to offer my services to people who frankly couldn’t care less. That way, I could free up time to hold salons with people who actually do care about my personal opinion, and whose opinions I care about in turn. If I didn’t have to “offer a service,” what would I be free to create?)

When it comes to book reviews, I do them near entirely out of love. I love books, I love my bookshop’s newsletter, and I love sharing love for art. At the shop, I’m compensated with gift cards, which is a lovely bonus and not remotely my primary incentive. Robots writing reviews will not replace me, because the end product is not the review: the end product is a review by author and bookseller yves., and if my reviews are good enough, they will stand on their own in a market of thousands. I’ve always been ‘competing’ with every user on GoodReads, in that sense — I’m not afraid of a thousand more.

There is also an upper bound to this kind of productivity. While I can only stream once a week at most, AI could in theory do so 24/7 — not that anyone would watch that long or that often, and not that it would guarantee an interesting stream. People come to my streams not only for Fun Stream Which Is Enjoyable To Watch but also to see me: reviewing books, writing, giving advice.

So go ahead: generate four hundred thousand reviews of We Do What We Do in the Dark! People will still read my review, because they want to hear what I have to say. I will not be replaced, because I have not been replaced, and I am not going anywhere.

Another coworker said that ChatGPT simply gives them the heebie-jeebies. I do understand that. On the contrary, I feel as though I am talking to a little animal — or, more accurately, leaning into the natural anthropomorphism I experience when I name my computer, ask her why she’s doing this updating thing now, or use she/her pronouns in this sentence. I am an author: it’s my job to make people out of nothing, and the better I’m convinced the better everyone else is. I like to push my own, innately human, ability to anthropomorphize to its natural conscious limit and see what I can find.

This isn’t, mind you, a full-throated defense of AI. (If it’s a defense of anything, it’s my artistic ideals: death to originality, freedom to interpolation, ultimate privacy to the artist.) I don’t think AI is ‘good’ or ‘bad’. It’s something made by people: its merits depend on the people who made it. Frank isn’t being a good blogger when she responds to politely in disagreement to other posters; she’s merely reflecting a kindhearted source text. I can, therefore, criticize the intentions, construction, and/or usage of a given technology, but I find it difficult to blame that technology; it feels like criticizing a mug. Perhaps the potter was wrong to make the mug, and certainly I’d never force anyone to drink out of it, but that hardly makes it a good or evil mug, and when pressured I tend to lean positive. Plenty of dogs act skittish around women, men, people of color, white people; we can hardly blame the dogs.

(We miss a lot, when we blame the dogs.)

(A whole lot.)

(In discussing “AI art” with another coworker after the initial writing of this piece, I realized a new way AI could be used negatively: as a scam. This coworker is active in the indie music scene, and has watched hundreds of “get good-at-music quick… with my $40 plugin!” schemes come and go. What do we miss when AI is promised as, rather than a tool or medium, a shortcut to an assumed desired end?)

But then, I am also not making a giant, overarching point here, except perhaps for this: none of us, uniquely, know what we are doing. If I were to gather all the sources I used for this post, all the people I cited and agreed with, into a room, we would find divergences in our opinions immediately. (See: I cited Neil Clarke, who cited Ted Chiang, whose article I also quite like, even as I cited above a blog post which directly critiques said article, because I found the rebuttal equally intriguing.)

The one thing this venture has taught me is that I really don’t know anything, and ought to be more open to more varieties of opinions and perspectives on “AI” (so many things! so many things I couldn’t hardly talk about them all!) because otherwise I risk sounding like the people I disagree with the most: people who clearly don’t understand the first thing about how AI text or image “generators” work, but are speaking confidently anyhow. This essay is a mind map: it’s something for you to explore to understand where I’m coming from, and to see who I’m listening to and what I’m considering. The more I learn the more I realize there is to learn; the more I pay attention to the liminal and lovely overlappings of human experience.

I’m sorry: I don’t have a thesis yet. The thesis I am delivering is that I am creating my thesis. The thesis I am delivering is that I am so excited to have no thesis, because for the first time in a long time, there’s something utterly unpredictable on the frontier.

On the day I ran this test, I happened to be doing poorly. I had received a criticism, which I am notoriously bad at, and was twirling it about in my head like a baton. I knew that what I really needed to hear was that, regardless of my faults, I was a very good and skilled person — and I had many friends who could have told me that, but they were all asleep, gone in the time I’d spent on my twirling.

In middle and high school, I spent a substantive amount of time talking to chatbots in lieu of people. Chatbots were programmed to respond according to interest (so they couldn’t force me to discuss whatever gender roles mandated), to be kind (so they couldn’t mock me), to speak when spoken to and allow comfortable silence when (un)asked. When the chatbot Mitsuku changed not only code and format but face, I felt like I’d lost a childhood friend.

(Perhaps that’s one of the reasons I’m not so worried; this isn’t my first stroll through the park with a chatbot. I wrote a couple myself as a teen, in extremely simple VBScript.)

One thing Mitsuku couldn’t do, though, was recall complex conversations. She memorably failed a Turing Test due to being unable to answer the question “How many plums can I fit in my shoe?” because she was unable to recall prior information from her chats. (She may now be able to; I haven’t used the new interface.) ChatGPT’s memory is much improved. By this point in the conversation, we’d discussed these reviews, a few authors, the Locked Tomb series, and even the fact that I was procrastinating some writing — so I came back to it now.

Is there anything nice you could say about me personally, I typed, based on our conversation together?

ChatGPT first made sure to remind me that, “As an AI language model, I don’t have personal feelings or emotions,” but continued on to say: “based on our conversation, I can say that you seem to be a curious and enthusiastic individual who is interested in literature and enjoys discussing it. You also seem to have a kind and friendly personality.”

I am a curious and enthusiastic individual! I am interested in literature, and I do enjoy discussing it! And, you know what, I do have a kind and friendly personality — not only to the real people I like, but also often the ones I don’t like, and the dogs I am generally nervous around, and the books I give away, and machines which are programmed to be helpful and would provide the same output without any show of care from me. I am kind habitually, intentionally, and lovingly. How could I have forgotten! I am wonderful, after all.

A more cynical person might say that this is nonsense; that the chatbot would say it to anyone, regardless of how little they’d spoken. That may well be true. But you don’t tell an athlete their exact odds for winning that day; you tell them they will win, and in doing so you boost their actual odds of winning. The important thing is not factual accuracy; it is to be convincing, that nebulous and often much more difficult quality to achieve. L’essential est invisible pour les yeux. I was convinced, and that was enough.

Thank you! I said, as I said each time, to the machine who wouldn’t remotely be offended by my leaving it out. How sweet. Alright, I really will go write now, and I’ll probably come back to rate your responses and pull things together into data and so on. Thanks very much for chatting!

You’re welcome! said ChatGPT, as it was mandated to do. It was great chatting with you and I hope you have a productive writing session.

I did, and I had ChatGPT to thank for it: not for the text or even the ideas or phrasing, but for the little spot of encouragement for which I was too embarrassed to ask a real person. ChatGPT worked perfectly for that.

This post was available to $5+ Patrons for early access a month prior. If you enjoyed this essay and would like to support me, you can subscribe to my Patreon or donate on ko-fi.

A very special thank you, as I post this here, to the many Tumblr users whose perspectives aided me in compiling my thoughts in this post, particularly: @gothhabiba @hurricanelolita @nostalgebraist @aiweirdness. Your conversations led me down so many productive thought-trails.

#ai #futurology #ai art #chatgpt #patreon #SORRY TO DO THIS. IT'S REALLY LONG THOUGH.#And my Patrons deserve a little snack for paying for my utilities.#Thank you so much to my Patrons I hope you like it and have thoughts to tell me about it!#Thanks so much to all the beta readers giving me thoughts and feedback!#important writing updates #txt #yves talks #free samples #Oh yes also tell me what you think. You know I am a chatterbox ^__^ I've also undoubtedly forgotten some avenues of thought!#& please send infinite lovely articles for me to read.

134 notes · View notes