According to the editor of award-winning sci-fi/fantasy magazine Clarkesworld, and some personal contacts, the short fiction world has encountered one of the most predictable consequences of this new era of generative AI: being flooded with computer-created dreck. While many have focused (and rightly so) on the possibilities for using this technology to manipulate public opinion, or the consequences for labor, its most immediate impact has been to the world of art.
Venerable publisher Tor has accidentally been using AI-generated book covers by hiring unscrupulous stock-image artists. In this case, it was actually a bad enough image that they had to use an in-house artist to fix it (the human figure was legless). Game studios have begun using generators as part of their workflow for creating textures, though the samples I’ve seen for this do look pretty good. A heavily-AI-assisted image won a photomanipulation contest last fall, though it did have a (highly questionable) disclosure attached. In ways great and small, at every level, the art world is becoming saturated with generative content, to the point where you can accidentally consume and even reward it.
On the written side of things the effects have been less immediate, at least as far as we can tell.
In non-art news: I’m sure many of us have seen the articles about students using AI to cheat on essays and the like. Microsoft will be adding some generative features to the Office suite. Google has had autocomplete for entire emails for a while now. AI-assisted coding has been around for years and continues to improve. (It’s still worse than an intern, and requires just as much oversight, but it has its uses.)
In art news: it looks like fiction markets will have to change how they function. Clarkesworld editor Neal Clarke wrote a recent post about how his submission pile is now unmanageable. For the first time in a long time (or ever?) he has had to close submissions because the number of ‘writers’ he has to perma-ban looks like this now:
Towards the end of 2022, there was another spike in plagiarism and then “AI” chatbots started gaining some attention, putting a new tool in their arsenal and encouraging more to give this “side hustle” a try. It quickly got out of hand… the number of spam submissions resulting in bans has hit 38% this month. While rejecting and banning these submissions has been simple, it’s growing at a rate that will necessitate changes. To make matters worse, the technology is only going to get better, so detection will become more challenging. (I have no doubt that several rejected stories have already evaded detection or were cases where we simply erred on the side of caution.)
This is probably just how things are going to look for a while in the art world. Even if we imagine that these stories aren’t scammy, and are just by aspiring ‘writers’, this is something everybody will have to live with. Eventually, I imagine, auto-rejection tools will improve, just like they did for spam emails, but just this second, things are untenable. There’s not an easy answer. It will just take more time to filter out the bullshit, impoverishing everybody in the process.
The backlash has been swift, if incoherent. A splashy class-action suit alleging copyright infringement by some of the companies producing these image models has been filed–but it is riddled with factual errors and cannot, by my understanding, be reconciled with existing interpretations of copyright law. Many writing contracts now make you state that you did not use an AI at any step of the process, which to me misses the forest for the trees. Chatbots can be a helpful brainstorming tool–should that get me automatically rejected, or force me to lie? What if GPT-3 spits out a sentence I really like when I give it a few pages of a story and ask it to write the next paragraph? A friend and I are working on a game right now using AI-generated images for placeholders while we code–does that count? Where exactly do we draw the line between drawing a hill with Adobe’s AI brushes, which illustrators have been using for years, and drawing a hill with Stable Diffusion?
These tools are part of the world now, and in my opinion a smart artist would do well to learn them.
On the other side of the discourse, things are… worse. There is a sizable contingent that views the entire art world as illegitimate gatekeepers and would like to dance on its grave. Opinion-haver Jon Stokes (who proudly had a “web3” .eth crypto domain last year, and whose opinions on all things should therefore be disregarded) had this to say:
Seriously, the link up top where the guy is flipping out about AI-generated content in his short fiction journal? …Who cares? If the story is good then I want to read it. If it’s bad then I don’t want to read it. Maybe just decide if it’s good or bad, and if the problem is that submission rates are too high, then charge people for submissions…
Creators are worried about competition from untalented normies who didn’t earn the ability to produce quality work and therefore aren’t that invested in the work itself or in any sort of guild, discipline, community, or other social institution that’s connected to creative work.
No, dingus, it’s because you’re in league with a bunch of scammers, and aggrieved young men who don’t understand the difference between “passable, discardable works” and “top-tier artisan works that deserve to be featured in Clarkesworld“. If an AI-generated story were capable of that–which is just laughable to consider at this point–this might be a discussion worth having.
Instead we get to have discussions like this one I had yesterday:
What’s illegitimate about being upset that you’re drowning in shit submissions all of a sudden?
— ☕️ hoopy frood 🍵 (@TynanPants) February 21, 2023
Luddite nonsense. Develop better systems for screening quality. Responding to labor-saving technological innovation by calling to ban it. He’s not concerned with the quality, just the means of producing it. If he wants to hire staff writers and only publish them, he can do that
— Росія іди нахуй 💸🇺🇸🌐🏗 ⚛️ (@rnoyfb) February 21, 2023
Instead he elicits submissions from the public and wants to externalize the cost of screening them to his completely medieval standard.
— Росія іди нахуй 💸🇺🇸🌐🏗 ⚛️ (@rnoyfb) February 21, 2023
He also accused me of wanting to artificially restrict the supply of writing. I replied “Wow lol” and he blocked me, the precious snowflake.
Open thread, I suppose.
Baud
It’s not art until an AI cuts off one of its ears.
Qrop Non Sequitur
It’ll just respawn.
Baud
I’ve been meaning to play with AI more, but I’m not creative enough to figure out how.
Major Major Major Major
@Baud: If that’s not sarcasm I’m happy to walk you through it.
Old School
But are blog posts real or was this composed by AI?
Kent
HS teacher here.
Although I’m a science teacher I’m not particularly concerned about AI. There are plenty of tools to deal with it.
For tests and in-class assignments you can always go back to the old blue exam books and have students do hand-written essays. Or if you don’t want to read their increasingly crappy handwriting there are plenty of tools you can use to lock down their Chromebooks to keep them from googling or using AI web sites to complete in-class essays.
If I was an English teacher or History teacher or some such, I would break essay assignments down into a lot of different manageable chunks with graded check-points so that they couldn’t just present a finished essay with no documentation of how they got there. So, for example:
Step 1: Research and compile your sources and summarize the main points of each one and put them into MLA reference format (there is already software for that)
Step 2: Outline your arguments and organize them.
Step 3: Write your introduction
Step 4: Complete your first draft
Step 5: Circulate your rough drafts for peer-review and identify changes
Step 6: Complete your final edited product and summarize the changes that you made from the rough draft.
No matter how good AI is, it isn’t going to do all of that for you. Teachers already do all of those things to avoid the problem of paid essay writing mills which have been around for years. There are whole towns in India that are supported by writing term papers for lazy American college students and HS students.
Baud
@Major Major Major Major:
No, I’m serious. I’m really only familiar with ChatGPT. I’ve done a couple of searches and Q&A just to see how it works.
Baud
@Old School:
Protip: Look for typos.
NotMax
AI manuscripts = Power of babble.
Major Major Major Major
@Baud: Interested in text or images? Most things cost money (or have a pretty low free tier) unless you have a good graphics card… ChatGPT is the best free-tier open-access text thing available right now, to be honest.
The Moar You Know
I like the idea of the teacher I read about who is requiring the class to use ChatGPT for all their submissions, with an attached explanation of the inputs they gave to the bot and why they picked those.
Which you could also generate with the bot, but that really starts involving a lot of work.
zhena gogolia
@Kent: If they want to piss away their chance to learn how to think and write, who am I to stop them?
Baud
@Major Major Major Major:
Oh, I’m out then cuz I’m cheap and not that interested. I was curious about trying out images though, since ChatGPT doesn’t do that.
Baud
@zhena gogolia:
Nominated!
Kent
If you are a teacher it is what you are paid to do.
If I left it up to my students to decide for themselves whether or not they wanted to piss away their education I’d probably have 50% of them fail.
Kent
How much randomness is there? If 5 students submit the same exact inputs do they get 5 identical essays? Or do they all come out differently? That would be interesting to analyze.
Baud
Didn’t education go through something similar when it came to math and sophisticated calculator apps?
Qrop Non Sequitur
@Baud: To avoid this, I’ve had calculus teachers straight up ban calculators from the classroom. Others have had mixed, calculator and non-calculator sections, but show your work.
Major Major Major Major
@Baud: I think some macs can run Stable Diffusion natively if you don’t have a gaming computer… there are some playgrounds too
https://huggingface.co/spaces/stabilityai/stable-diffusion
free tier at the official one: https://beta.dreamstudio.ai/faq
don’t know what this is but looks free too https://stablediffusionweb.com/
Ken
No, just the supply of artificial writing.
I am seeing an odd anti-parallel between thse AI “authors” that are flooding the system so humans can’t be noticed, and the AI “evaluators” in HR departments that reject humans for not having the right keywords in their letters.
NotMax
@Baud
Images? Perhaps not the best of the best but for a free tool it ain’t bad: Picsart.
Baud
@Major Major Major Major:
Awesome. Thanks!
Qrop Non Sequitur
@Ken: Isn’t all writing artificial? Doesn’t the act of writing produce what may be termed an artifact?
bbleh
“Tech will make up for both my lack of talent and my unwillingness to learn or work!”
Math Guy
@Qrop Non Sequitur: I always asked my students to show their work: if I can’t do the problem in my head, I doubt that they can either. Word problems and geometric questions are also a good way to test a students understanding of the material.
Ken
There’s some evidence that the AIs, especially those generating spam, use typos to convince the humans (and spam filters) that the letter’s not from an AI.
(Though “convince” isn’t the right word, since it implies some goal-directed volition on the AI’s part. It’s just trying variants, and re-using the ones that improve the response metrics.)
Major Major Major Major
@Qrop Non Sequitur: “show your work” is what matters for that, yeah. Some people have suggested including ‘track changes’ with your text submissions but I don’t think people will want to share that, it’s a bit invasive, idk. It’s not an intractable problem but it’s going to be very annoying until some standards are developed around this.
Qrop Non Sequitur
@Math Guy: Good thinking.
Qrop Non Sequitur
@Major Major Major Major: I’m familiar with the track changes function but I only have a very weak guess how it may help with detecting AI submissions.
Ken
@Qrop Non Sequitur: I was using “artificial” to refer to the source (human, or AI), not to the product.
Qrop Non Sequitur
@Ken: Yeah, that makes sense. My bad.
That was about what I was thinking. Edits can done manually afterward, though, just to make it look right.
@Major Major Major Major: M4, are you able to unlock a nym? I’m looking for a change.
The Moar You Know
@Kent: That’s a good question. I would assume “reasonably close but not word-for-word identical”, but I have not used it and do not know.
My wife (who is a teacher) and I recently reviewed a few papers done on it. I’d say the writing quality is about that of a poor ninth grader. Very few tells – one of them being that, if the paper is about someone, the bot will use the full name every time it comes up, which few humans would do.
Major Major Major Major
@bbleh:
Now you get it!
@Qrop Non Sequitur: AI writing emerges fully formed from the head of the machine, so the track changes would just look like a big one-time paste job with little-to-no editing.
different-church-lady
Garbage in, garbage out.
Major Major Major Major
@The Moar You Know: ChatGPT has a fairly low “temperature” (randomness setting), but the model itself can change in between queries.
@Qrop Non Sequitur: I’m not totally hip to how the process works any more, but if you drop a comment under your new nym (might need a new email too? idk?) I am happy to approve it, which will whitelist you.
different-church-lady
Hey you know what? I do want to restrict the supply of writing. I want to restrict it to the good stuff.
The Kropenhagen Interpretation
Ah, that new nym smell..
And thank you.
different-church-lady
@Major Major Major Major: How would it respond to the question, “What did I just ask you?”
Ken
Some books I’ve read show little or no sign of editing, so this test isn’t conclusive.
Major Major Major Major
@Ken: unless you’re reading anime-inspired Kindle Direct books, they probably did go through like twelve drafts. Not everyone is a dazzling writer!
@different-church-lady: you should sign up and give it a spin! “It’s part of your education” as my dad would say.
Baud
@The Kropenhagen Interpretation:
You trying to get Schrodinger’s Cat to like you?
The Kropenhagen Interpretation
@Baud: If this were a legitimate first step, I don’t know what the second would be.
The name parallels didn’t escape me, though. But I had to use it.
NotMax
Is AI the text equivalent of Milli Vanilli?
//
Baud
@The Kropenhagen Interpretation:
I’m sure there’s some world out there where she does.
Bill Arnold
@Ken:
Or just generate text, and apply insert-random-typos filter using a list of common typos and only changing a fraction of instances. E.g. word misspellings:
https://en.wikipedia.org/wiki/Commonly_misspelled_English_words
The Kropenhagen Interpretation
@Baud: 🤣🤣🤣 I’m dead.
Citizen_X
Bob Dole does not like this rule.
Ken
Only in some universes.
zhena gogolia
Looks as if Biden gave a great speech in Warsaw. I have to go out to dinner now, but I hope it gets discussed on BJ.
The Kropenhagen Interpretation
@Ken: Oh for sure, very literally in quite a few.
Major Major Major Major
@The Kropenhagen Interpretation: not nearly as many as the number of ones in which you were never born, though.
The Kropenhagen Interpretation
@zhena gogolia: Biden speeches give the feels.
The Kropenhagen Interpretation
@Major Major Major Major: I suppose there are infinite of each. And infinite with each unique death.
Then again, not all infinities are created equal.
Baud
@zhena gogolia:
AL or Adam will be on it.
Major Major Major Major
@The Kropenhagen Interpretation: surely the infinite set of universes where humans evolved is smaller than the infinite set where we didn’t!
The Kropenhagen Interpretation
@Major Major Major Major: I stumbled upon multiversal thinking my senior year of high school just thinking about simple geometry.
These questions have plagued me ever since.
mwing
Fortunately I make physical stuff, small stained glass panels. But I wonder what will happen when people start using AI as an aid to their cartoon designs? (The stained glass design on paper is called a cartoon).
I can easily see real stained glass artists, or people who are more hobbyists, like me, using AI to generate “leads”- ideas for panels that are basically patchworks of other people’s work.
Then redrawing and redesigning from at least parts of the output…but you know, we kind of do that a bit already by hand.
The thing is the AI doesn’t know what stained glass is, and what it’s technique and limitations are, so I’d expect to see a lot of output with uncuttable/undoable design, but maybe…not?
Maybe not if the data set is based on real produced work, like the AI would, unknowingly, eventually, , “get” the rules?
this whole thing is so weird I don’t even know how to think about it.
Searcher
The thing is, if/when an AI could write unlimited good literature effortlessly, given a few prompts… no one would pay writers again. You’re never going to really win using AI spambots to submit stories for publication. The success of the strategy immediately destroys the profitability.
Sister Machine Gun of Quiet Harmony
Tech Bro snowflakes are extra icy. I can remember more than one heated argument about 3 years ago on whether crypto was the future. I maintained that while the technology can be legitimately useful for other purposes, crypto itself was only useful for crooks and speculators. I made a couple of them very angry. 😂
The Kropenhagen Interpretation
@Sister Machine Gun of Quiet Harmony: crypto is great if you want to burn the world faster.
Sister Machine Gun of Quiet Harmony
@The Kropenhagen Interpretation: Yes. Those people are a subset of crooks.
NotMax
@Searcher
God, I Hate Shakespeare (Reprise).
;)
Major Major Major Major
@mwing: so these AIs don’t really know what anything is, as you or I conceive of knowledge. They can be bad at anatomy because they don’t “know” what a human is. They do know what pixels “a picture of Joe Biden waving” is made of, more or less, but not what’s under his suit if that makes sense. ChatGPT is really bad at knowing what letter a word starts with because it doesn’t know what a letter is—it operates at the level of a word. Lots of delicious philosophy of mind discussion to be had here about what humans “know” but that’s for another time.
You can fine-tune a model by showing it a lot of stained glass, though, and it will become good at making stained glass designs. Any “knowledge” of physical limitations of the form will only be inferred from the pixels though. If that makes sense.
Jeffg166
For future artists AI will be a tool to use to produce or coproduction work.
Math Guy
@Baud: The cat both like and dislikes you till you try to pet it.
Cathie from Canada
On a side note, one of the commenters to a recent AI piece I read suggested this: whenever you are chatting with a suspected bot, just ask “But what about the squirrels?” and a bot will actually try to reply, while a human will just say “Huh?” So it becomes a sure-fire test. From now on, my question to the brave new world of AI Chatbots will be “But what about the squirrels?”
The Kropenhagen Interpretation
@Math Guy: That assumes a real possibility the cat likes you.
Major Major Major Major
@Jeffg166:
If the combination of luddite detractors and really-bad-at-PR boosters don’t end up getting the damn things banned first.
I believe the UK has already proactively said that image generators are not copyright violators. And over here I think the big houses will be in favor of the tech. So I have some hope that our courts and legislators can be correctly bribed and cajoled.
Mr. Bemused Senior
@Cathie from Canada:
What about Naomi?
Hoppie
Almost fifty years ago when I couldn’t afford it, I resolved to only by original artwork. I now have a remarkably wonderful SF – and other – original, mostly oils and acrylics, collection. They really can’t yet be faked. I love living in an art gallery. It may end up at UC Riverside, since the spawn does not seem interested. ‘Nuff said.
RSA
Wow. I wonder whether anyone has thought about setting up a Web site that just churns out Disney- or Marvel-themed images? That might be an interesting test.
Anoniminous
Gary Marcus is running a timely criticism of ChatGPT & etc.The latest screed is: What did they know, and when did they know it? The Microsoft Bing edition. For those who don’t know who Gary Marcus is …. Gary Marcus
Feathers
@Kent: My ADHD ass would just fall apart trying to do all that nonsensical make work. My “rough” draft is always pretty close to my final draft. I can’t tell you how many classes I just ended up dropping because a professor demanded that my paper be graded based on how much I improved it from the rough draft, rather than how good it actually was. That was all before I was diagnosed and before I found out I have a tiny working memory. I come up with the best way to write it and my brain isn’t going to give me another way. It’s a polish or start from scratch on something else. I thought my problem was writers block and ended up dropping out of several college programs.
Teachers don’t understand that demanding a certain workflow will make the task impossible for some students. And there’s no possible documentation that will get you an accommodation for this sorts of issues.
Another Scott
There was an Ars Technica story recently about a guy who beat a Go computer by playing unconventionally. We’re not quite doomed yet. ;-)
I had a fancy slide rule in high school in the late ’70s, just before affordable calculators were becoming common. I’m not sure that we understood the math any better than folks using calculators a few years later – it was all to easy to read the wrong scale in the middle of a calculation! Maybe they’re easier to use now, but back in my day one had to understand quite a bit of math to use those things effectively.
I still vaguely remember a lesson on how to calculate square-root long-hand. It’s kinda like long division, but there are steps involving squares and suchlike. Kids these days don’t know what they’re missing!
Anyway, software assistants will get smarter and more ubiquitous. Some of it is good. Some of it is bad. Some of it is just different. The main thing is, in my opinion, is getting kids to think and exercise different parts of their brains. Being a master with a slide rule, or being able to calculate square roots long-hand 45-50 years ago doesn’t mean a whole lot now… Kids need to know how to look stuff up, how to figure out whether the stuff they find is accurate or true or relevant.
I like the idea of incremental assignments. Maybe another way to approach the ChatGPT cheating problem is for students to evaluate and critique an essay that’s already provided. “This ChatGPT essay says that Albert Einstein was a watchmaker in Zurich and played the glockenspiel when he was on break. On weekends, Einstein liked to drive Nuvolari’s Auto Union racer around the Nürburgring. …
1) How many errors of fact can you find in the first 10 sentences of the essay?
2) What evidence supports your count in #1?
3) …”
Cheating is bad, because it corrupts. And cheaters who don’t feel bad about cheating or never face consequences for doing so, and thus don’t stop doing it, become monsters later in life…
Cheers,
Scott.
different-church-lady
@Sister Machine Gun of Quiet Harmony:
Crypto was the future. It was just a future filled with failure.
Major Major Major Major
@RSA: Disney loves this stuff, I suspect. Talk about a productivity multiplier! As for a website full of knock-off/fan-art, we already have those, they’re called DeviantArt and Pixiv and ArtStation and, and…
There’s no copyright violation you can do with these tools that you can’t do with a paper and pencil, other than the scale of the thing.
different-church-lady
@Cathie from Canada: That was me! “What about the squirrels?” was something I came up with years ago as a parody of the kinds of letters to the editor a small town newspaper would print.
CliosFanBoy
Meh, still better than anything L Ron Hubbard ever wrote.
different-church-lady
@CliosFanBoy: Well, he was a bot too.
CliosFanBoy
Who dafuc is that bozo on Twitter with the Russian (?) name? The one slinging around the “luddite” smear??
Major Major Major Major
@CliosFanBoy: Name translates to ‘russians fuck off’ or somesuch, as I recall. His politics don’t seem too appalling, really, looking over his feed. But at the end of the day–who cares about him, not me! What a loser.
Another Scott
@Major Major Major Major: No “suspect” about it. They have a history…
Disney Reused Animation Scenes (3:21).
Cheers,
Scott.
RSA
@Major Major Major Major: Interesting! I don’t have a good understanding of the tech industry.
Unrelated (but within the scope of the post): OpenAI released an AI-vs-human-written-text classifier late last month. It’s not very good…
But as with cybersecurity it’s something like an arms race, so maybe it can be improved.
CliosFanBoy
@different-church-lady: I thought he was a Thetan? or am I getting my cults mixed up? ;)
CliosFanBoy
@Major Major Major Major: Ah, OK. I wondered. He had a Russian name but the NATO star as his avatar.
different-church-lady
I’m not sure how this applies, but it reminds me of how 40 years ago everyone thought sophisticated drum machines were going to make drummers obsolete.
Brachiator
@different-church-lady:
Auto-Tune has almost made actual singing irrelevant.
Kent
At the HS level it is almost mandatory unless you are talking about talented AP kids. Otherwise you end up with a ton of kids putting it off until the deadline and then turning in crap.
rekoob
The unofficial results in the Virginia 4th Congressional District race (as of roughly 8p Eastern):
Jennifer McClellan (D) 50,110 (69.52%)
Leon Benjamin Sr. (R) 21,718 (30.13%)
The remainder are write-in votes.
VA 4-CD Special 21 Feb 2023
different-church-lady
@Brachiator: It’s most certainly made actual singing impossible to hear.
Elizabelle
@rekoob: Thank you, rekoob.
Just coming here to celebrate. McClellan was heavily favored, but every victory is a victory
May Wisconsin’s vote also be a set of victories.
Brachiator
@Searcher:
There are people trying to get AI to get good at scams. Unlimited fraud (letters from Nigerian princes, attempts to get people to reveal personal information, hacking, phishing, etc) creates great dreams of wealth.
I also expect to see AI used in political campaigns, fund raising, cold calls for donations, etc).
kalakal
There’s an awful lot of music software thet claims to be AI. hmmm, nope.
Some are billed as smart compressors, eqs etc notably Izotopes Ozone. It actually does a pretty good job of mastering but it’s really just a glorified way of taking a mix and applying a chosen formula to it
More in line with creativity is stuff that claims to write entire songs. It’s pretty terrible. There are chord sequence, rhythm, riff, and melody generators which are great for giving ideas but don’t pretend to write a song. But the ‘AI’ song writers are dire
CliosFanBoy
@rekoob: Good, the republican candidate is one of those African_American nuts that the republicans seem to dredge up to show they’re really NOT bigots.
different-church-lady
@kalakal: Problem is, as far as I can tell, real life song writers are also pretty dire nowadays.
Another Scott
@rekoob: 🎉
Excellent.
Thanks.
Cheers,
Scott.
kalakal
@different-church-lady: lol!
One fun thing is with midi. Virtual instruments these days are amazing*, and if you’re a bad keyboard player they can help you out by snapping your plunking away to a grid so your timing etc is fixed via a command called quantize. As this makes it very robotic there is also an option to introduce an element of sloppiness called humanize. You also do this if you’ve straight up programmed it
* Some are not so hot eg saxophones
Major Major Major Major
@kalakal:
This was the state of AI images a year ago so just you wait!
kalakal
@Major Major Major Major:
I’m sure it will get better, I can certainly see it as an idea generator
One product that has impressed me is in the field of photo ‘improvement’/ processing/ restoration is Topaz Photo AI, it’s a mix of their denoize, sharpen, and gigapixel ai packages with some extra. I have a ton of old slides etc from my dad a lot of which need help due to age etc. It’s pretty time consuming so I fed a few into this beastie.
It doesn’t always work but it has a pretty high hit rate and when it does it’s sodding amazing. It’s reall impressive & has saved me a lot of time
zhena gogolia
@Feathers: THAT’S HOW I WRITE. Maybe that’s why I can’t stand asking for drafts. Figure it out and THEN write it.
Major Major Major Major
@kalakal: AI upscaling is really impressive, yeah. A lot of graphics cards have an option where you can run the game at a lower resolution, but it runs everything through an upscaler, and you can get a pretty great resolution:frame-rate ratio that way a lot of the time on a consumer-grade card.
different-church-lady
@kalakal: Oh yes, I’ve played with a lot of it. But I don’t do much music, so I’m far from an expert.I did recently have a spot of trouble on a live-to-tape music video with a vocalist who… well, let’s just say his ambition outstripped his abilities. We were going for the feel of a live acoustic performance and auto-tuning in post just made it sound crappy and fake. In the end I just comped a few takes together and everyone was happy. But later I went into Nuendo and started playing with the built-in pitch correction tool, which can give you quite a bit of manual surgical precision. And I discovered that what I needed to do to get a natural sound was correct the *middle* of the notes while leaving the slides and dips at the beginnings and ends intact. It’s something the plug-ins just don’t do by themselves.
kalakal
@different-church-lady:
I hate autotune, it’s so fake
There is some remarkable pitch correction software around. Melodyne seems to be the favourite amongst the cool kids, repitch is another. Melodyne can work on polyphonic sources which is rather amazing. They all need supervision as they’re basically spectral* editors with fancy UIs. I use Nuendos cousin, Cubase a bit, i think they both use the same pitch correction software. If you really want to get in the weeds Steinberg produce Spectralayers but that takes a fair bit of learning to put it mildly. Life is usually too short unless that is your job
You got that dead right, nobody is consistent throughout a note.
* I was so disappointed when I learnt they had nothing to do with ghosts
PJ
@different-church-lady: Well, programmed drums, if not “drum machines” in the ’80s sense, are everywhere in pop music. In fact, most pop music these days uses very few actual musicians – most of what I hear is just programmed samples, with the lead vocals from an actual human, but heavily treated. A lot of soundtrack work, whether it’s for TV or movies or video games, is also just programmed samples.
Tehanu
Bret Devereaux latest post was about people using ChatGPT to write college essays and why they shouldn’t. Really thorough. Link: https://acoup.blog/2023/02/17/collections-on-chatgpt/?utm_source=feedly&utm_medium=rss&utm_campaign=collections-on-chatgpt
BellyCat
Taught a masters level architecture studio last semester. Stumbled blind into this topic with a student who decided to outsource idea generation to AI. Interesting discussions. Tragic results.
The frontier I’m curious about is if/when AI can generate digital 3D models (rather than static 2D images) which can be built (by full scale robotic printers n’atch!).
Paul in KY
@Major Major Major Major: I would assume so.