Michael Frank
@mfrankDude
Followers
749
Following
348
Media
275
Statuses
4,173
VFX Compositor | Technical Director
Joined November 2017
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
Kendrick
• 368180 Tweets
Wayne
• 130230 Tweets
Browns
• 105375 Tweets
Watson
• 84265 Tweets
Lions
• 83180 Tweets
Manuel
• 68480 Tweets
Raiders
• 59425 Tweets
重陽の節句
• 59053 Tweets
ON CH3 iQIYI JACK JOKER
• 45470 Tweets
Baker
• 43272 Tweets
#ZEEgotoNYFWwithCOS
• 36381 Tweets
Rams
• 33716 Tweets
もちづきさん
• 19628 Tweets
シーザリオ
• 19540 Tweets
Stafford
• 17870 Tweets
ZEE x COS
• 17574 Tweets
Angélica
• 16282 Tweets
#OnePride
• 16177 Tweets
Yuhui
• 15227 Tweets
Goff
• 12570 Tweets
#GOAT_70Mviews
• 12011 Tweets
JD 3rd ANNIVERSARY
• 11266 Tweets
サッポロポテト
• 11098 Tweets
ふっかさん
• 11022 Tweets
Pinned Tweet
If you want to see what kind of images are being scraped from the open web (and where they're scraped from), check out the browser I'm working on for LAION aesthetics.
1
8
51
This isn't a Napster moment. This isn't you and your friends using peer to peer file sharing to get free music.
This is a handful of well funded companies and venture capitalists trying to vacuum up ALL THE DATA and use it for million dollar training runs.
14
1K
3K
Wasn't this guy upset that someone else used their prompt without credit?
👉 Let's debunk myths.
It is not true that AI is going to take jobs away from artists.
AI will take work away from mediocre "artists" in the same way that a real artist takes work away from a mediocre artist.
Looking at these crybaby trolls 🧌😭, I don't need to tell you which
587
8
181
30
130
2K
"A dataset, that's like, a 'set' of 'data' right?"
"Right"
"So when you train images on LAION, you're taking people's data"
"No, LAION is just urls"
"Oh you train on just urls?"
"No you have to d/l the data"
"Got it! What do you call the downloaded data?"
"A dataset."
12
160
1K
So effed up.
Platforms are pivoting from "there's nothing we can do to prevent scraping" and "it's your copyright to enforce, not ours"....
TO
"I'ma sell all this stuff you put on our servers and you can't stop me."
NEW from
@samleecole
: Tumblr and Wordpress are preparing to sell user data to Midjourney and OpenAI, according to internal documentation we've reviewed
143
3K
5K
5
314
1K
I've seen this artwork countless times in my life, so I decided to draw it from memory in under 1 minute.
These are my results versus Midjourney's.
@DrTBehrens
It’s never violated. When an artist goes into a gallery and watches artwork he gets inspired. He is storing a compressed version of the image in his memory. That is the equivalent of training an NN. Does the artist violate copyright? No!
99
1
13
27
118
976
I'm always bummed to see others in the art community express defeat over AI plagiarism.
If you don't want to fight, that's cool. You do you. Telling others to give up is just free marketing and normalization.
13
149
930
Losing jobs to tech is one thing.
Losing jobs to tech built on your own work is another thing entirely.
If we don't object to that now, when will we?
22
125
853
That's a shame. 93 million urls are made available to Stability's training data by way of Amazon.
Any commercial product featuring art is exposed. Art prints, book covers, graphic novels, TCGs..
Artist licensed their work for products, not ML training.
We are proud to team with
@AWS
to bring
#StableDiffusion
and the upcoming Stable suite of generative AI models to Bedrock! Soon Stability and AWS customers will use Amazon’s
#Bedrock
to integrate foundation models into their apps securely and efficiently.
20
58
279
5
137
659
I love that AI art generation started with hackers openly talking about moving so fast they'll break copyright, to now having legal scholars bending over backwards to assert the legalities of it.
7
65
601
This stuff is tiring after awhile.
One of the first things that got to me when using Midjourney was the things it *couldn't* do. If you push to explore the space, you start to perceive the boundaries.
Helped me realize how much the system is limited by the input data.
I am delighted to share my artwork with you.
I dedicate much time creating it and as a courtesy I ask that you first ask me before using my art.
I would also appreciate you crediting my work. This way other people can fall in love with my art
1K
775
6K
21
42
581
You've been openly telling artists and creatives to hand over their work for free.
So far it's not much of a dialog, while you only have to entertain it as debate fodder..
companies like yours are actively training on everything while copyright questions are unresolved.
AI is not some sort of natural phenomenon that we have no control over.
AI is being built by us, humans.
Hence, if you're scared of AI, what you are actually scared of are your fellow humans.
You probably have doubts about the ability of institutions to do the right things with
721
236
2K
7
90
573
People who:
"can't draw",
"can't sing",
"can't write",
"can't play music",
"cant dance",
"can't act",
You are gatekeeping yourself
12
69
565
Stock photography company shuts down, citing digital photography and generative AI.
Meanwhile, they've had millions of images scraped for training.
@CanStockPhoto
Idk if anyone is still watching your account, looks like you closed up shop.
FYI your CDNs are still serving up images for anyone to grab via the LAION dataset.
0
2
15
11
184
511
This idea that "you view an image you get inspired" DOESNT SCALE TO MACHINE LEARNING.
For a human being to view the 2 billion+ images used to train Stable Diffusion, at a rate of 1 image per second, would take 77 years.
77 years!
W/out sleeping or attending to anything else!
11
80
457
If your idea of a companion is someone who's needs (truly) don't exist and they're always 100% responsive to you, you don't really want a companion.
Excited to announce v(1.0) of Digi, the future of AI Romantic Companionship, for IOS and Android 🤖
Site:
Twitter:
@digiaiapp
A quick thread on features, and where we go from here (1/13)
3K
670
5K
28
42
436
There is this very incorrect assumption that all data found on the open web is uploaded by an authorized user.
It's not.
Any user of Pinterest knows this.. we just right click -> Pin It = re-upload to a new server.
How is the original rights holder supposed to opt out of that?
3
0
0
9
91
409
This is what
@TheCartelDel
meant when saying using AI art makes your brand appear "feeble" and "impoverished".
You are a mostly empty warehouse with some yard ornaments.
Police were called to an 'immersive' Willy Wonka Experience after families showed up to an 'empty warehouse'
The event reportedly charged $40 for entry, advertised with AI art, and said it would be a 'journey filled with wondrous creations and enchanting surprises at every
2K
6K
95K
3
70
381
I am not a concept artist, but from what I gather.. using SD will likely handicap your abilities in the end.
You need to stand with those who can readily sketch, study form language, adjust the art based on notes..
Concept is more about ideation than full rendering.
6
41
365
I wish everyone who had this opinion had to train "how to draw from memory" for the next 6 months, just to understand how ridiculous their position is.
@DrTBehrens
It’s never violated. When an artist goes into a gallery and watches artwork he gets inspired. He is storing a compressed version of the image in his memory. That is the equivalent of training an NN. Does the artist violate copyright? No!
99
1
13
14
42
343
These folks really don't care where they get their stuff from. We're all just fodder to them.
4
92
333
There must be no financial value to training on copyrighted images, nope. None at all.
Not even for marketing or anything..
Otherwise this would all be infringement right?
OpenAI’s latest image generation model, DALL-E 3, makes it SO easy to create comic books!
Here are 4 panels for a fan-made Batman comic made in under 5 minutes.
Prompts included in the ALT.
Enjoy!🔥
1K
491
4K
13
34
324
Imagine you spend years of your life w/ someone.
You meet, you kiss, you dance.
You start a family, watch kids grow.
You age, get sick, get well.
Then one day, you suffer loss.
As you find words to voice your experience- Sam says, "ur just like, remixing other things, man."
everything 'creative' is a remix of things that happened in the past, plus epsilon and times the quality of the feedback loop and the number of iterations.
people think they should maximize epsilon but the trick is to maximize the other two.
678
822
6K
8
46
305
We keep hearing it's users at fault for copyright infringement,
yet the first thing these people run off to make are the most popular IPs and likenesses.
And there must be no commercial value to that.. or something.. I guess..
Real-time Latent Consistency is next-level fun! I can't get enough!
I forked & modified the main GitHub repo & added Desktop/App Capture as well as batch scripts for easy installation and launching.
Info in thread. 👇
364
117
1K
9
38
290
ATTN: Artists
is now targeting the text of <img> captions before using CLIP retrieval.
If images are captioned - "artist name" should have more results.
If you've ever said, "I tried looking for my work on HIBT, I don't find anything." PLEASE TRY AGAIN.
New feature: just got a *lot* more powerful with caption search 🔎
If you search your name, many more results will appear than before!
If you are present in popular AI image datasets, this search will find you, and you can opt out in a few clicks :)
3
48
95
13
132
285
ML teams aren't approaching companies to ask, "is it okay if we feed the best media in the world into a machine that can emulate it"?
One would have to go back in time to pre-emptively restrict research access to copyright data.
The panic is the result of AI commercialization.
The music industry is slamming the panic button on AI but they had a lot of time to be proactive here. AI music generators have been in the works for years.
Foresight is so important for every industry. You have to lean into the future instead of waiting to be disrupted by it
10
14
109
8
40
280
It's almost like... no one wants to be scraped .. and they already said so.
2
93
291
What do all of these SDXL image generations have to do with copyright? Nothing, I guess.
I strongly side with
@OpenAI
here.
Any attorney who argues that generative
#AI
output infringes on
#copyright
as a general rule does not understand how LLM training data works.
16
1
13
10
54
274
After much d/ling, I have LAION5B sitting on my hard drive and have compiled a total URL count by Domain of each subset 2B-en, 2B-multi, 1B-nolang.
Here they are in Google Sheet form for all to check out:
10
74
271
Fast forward years from now when the court cases are decided: what if all this art appropriation is infringing?
What will software groups have to pay out for being so wreckless?
It's wild that everyone is so full steam ahead KNOWING they're using unlicensed data.
11
37
273
Maybe if AI developers weren't intentionally focused on disrupting the general public, there might be a better public opinion.
Instead we see a large scale smash grab - followed by thumbs up, pats on the back, insistence of legality, and economic prisoner's dilemmas. Thanks AI
Here is Nick Bostrom admitting that AI panic feels out of control right now, "like a wrecking ball" that could "destroy the future"
33
69
363
9
47
264
There's this idea that diffusion models "learn" uncopyrightable facts: paint strokes, color patterns, etc.
This is a misrepresentation.
What they train to do is remove noise gradually, accounting for a keyword bias.
(Remove Noise at Time Step) * Subject Matter
8
49
250
This is the thing that's so hard to convey, yet simple to intuit for artists.
ML as applied to art is just recreating patterns.
It's "intelligence" is a shallow interpolation (and averaging) across the training data. It's not a true depth of understanding, it *needs* plagiarism
If these systems learnt instead, fundamentals such as surfaces, light, occlusion, etc they wouldn't struggle with a relational type of prompt like render the cube ontop of a the cube to the right of something.
1
0
34
2
47
247
Hi
@WaltDisneyCo
@Pixar
What about stuff like this from Stable Diffusion, the already-released-to-the-wild-so-there's-nothing-you-can-do-about-it-"opensource" model?
REPORT: Disney asked Microsoft to prevent AI users from infringing its trademarks. The key issue: Microsoft’s AI image-creation tools were generating a Disney-Pixar logo. | Click image for the story 👇
41
262
805
7
45
252
It's wild that creative originators have to make a case that their continued existence is a bigger net benefit to society than re-generative AI trained on their work.
F---in' wild.
3
44
252
Hollywood is more in danger of harming itself with this tech than being displaced over night.
People think the ability to "make movies" will compel people to watch them.
It's been pretty easy to make movies for awhile now. The amount of unwatched stuff out there is huge.
Not a great take. What happens if generative AI is found in massive breach of copyright and privacy? What happens then?
#ai
#HollywoodStrike
33
27
202
14
32
248
People: AI will help solve climate change, poverty, and end war.
AI: Sure, let me help you with that.
AI: Stop burning fossil fuels, share your wealth freely, end corporate citizenship, and don't fight each other.
People: Nah. That's too much. What else you got?
7
35
240
"Nope not a good picture... click...
not that one either.. click..
nah i dont like that either .. click
If I could only capture it the way *I* see it"
Introducing – Paragraphica! 📡📷
A camera that takes photos using location data. It describes the place you are at and then converts it into an AI-generated "photo".
See more here:
or try to take your own photo here:
1K
5K
23K
13
24
230
Reminder that most major platforms that host training data have anti-scraping Terms of Service.
Hilarious that OpenAI and Microsoft get to specify that people can’t use their products to create competing products, given that they’ve been building their products to compete against millions of people off of stolen work.
15
525
2K
3
63
240
Today on the deadline of
@CopyrightOffice
's response period,
I've written + recorded a walkthrough of whatsinlaion
Determining the scale of copyrighted material use is too massive to demonstrate in a comment submission.
We need tools such as these to truly visualize it.
10
74
226
One of my favorite parts of this is the section of Emad quotes.
RE:compression
1/ Mega pleased to give you all a hell of a class action update!
Our MASSIVE 96 page amendment is out! Look at the AMAZING roster of artists joining
@GeraldBrom
, Greg Manchess,
@zemotion
,
@GrzegorzRutko14
,
@upandoutcomic
,
@adamtotscomix
&
@HawkeSouthworth
!
12
837
3K
8
71
231
It doesn't matter how "massive" the human element.
It's capitalizing on someone's name, likeness, and voice without their permission. The attempt is more emulation than parody.
The reality is - if the creator's of this were anybody special, they'd use their own voice and image.
I don't think you understand how this special is made. It wasn't done just using AI. There's a massive human element that went into this. One that I'd argue is probably more prevalent than what was created using artificial intelligence
But you're just a whiner who farms for
34
0
6
10
33
220
The subtext of the loudest AI marketing efforts:
"You are behind"
"You are deficient"
"You are not capable"
"You are not enough"
12
31
227
After ignoring the Terms of Service on hundreds, thousands, maybe millions of websites.. AI developers expect everyone to listen to *their* terms, by prohibiting training on AI outputs.
I don't agree at all with copyright expansion. The music industry is the worst copyright regime on the planet + copying it for everything is a terrible idea. But I do agree it's hypocritical for AI companies to prohibit using their output for training.
10
15
85
5
39
213
When I hear sentiments like this from artists, I can't help but wonder what they've "gained" from the using generative AI.
When your work is necessarily a derivative of your peers' work, how much is yours to claim?
Have they even looked into LAION and ?
11
27
217
"..you know what we have now.. is 4-5,000 artists sign up for Spawning. half of them have opted out, half of them have opted IN"
It's a real joy to announce our
@huggingface
partnership🙂
From today, eligible datasets hosted on Hugging Face will show a data report powered by our API, informing model trainers what data has been opted out and how to remove it.
Another big step towards consenting training!
39
106
353
10
33
187
"scraping, spidering, or crawling" are already violations of your Terms of Use.
Why not default to opting out your customers? (Or better still, enforce the Terms of Use?)
@CindyLouWho_2
We understand the concern. To clarify, AI models currently train on all public web data. This feature is not actively opting you in, but instead offers you the option to opt-out and prevent your site from being crawled by Artificial Intelligence crawlers.
150
2
29
4
30
192
What is being proposed here would be the film industry cannibalizing itself.
Many works in LAION (and likely MJ's data) are film/tv stills, concept art, matte paintings, and storyboards.
Billions of dollars of value just packed away for $10/$30/$60 a month.
This is Fair Use?
I wrote this on August 29, 2022
***
I talked today to a person who works in the film industry in Hollywood, and he shared some interesting insights that, with his permission, I’d like to share with you.
Filmmakers are talking about Midjourney as a tool used for film production
28
36
362
8
32
185
Prominent AI developers lead the philosophical charge against protecting IP from machine learning.
Hand waving the fact that model weights are copyrighted and patentable properties in their own right -- derived from a training corpus of protected properties and used for profit.
@SpriteAttack
@survivor_343
When a human artist gets inspiration from the style, the themes, or the individual pieces from other human artists to produce a new piece, is that stealing?
114
4
53
3
31
181
Today I was msg'd by a photographer who saw URL counts for LAION.
Couldn't find their work on or via .
I ended up searching on my PC. After an hour: 540 hits on LAION2B. If you need help searching, hmu. I'll try to make time.
12
37
183
"The New York times is not telling the full story"
What's in your datasets OpenAI?
4
23
183
If you make it sound like a hype thread, then you can do whatever you want.. I guess.
Wow... MidJourney V6 can replicate almost any animation style.
And it's extremely easy to do.
10 wild examples with prompts:
416
311
3K
13
23
179
It's the difference of releasing the source code vs. releasing the trained model.
Users have no control over what's in the models, all of that copying was performed already by the developers.
Before:
AI learns just like a human.
After:
AI software is just like audio or video recording devices, photocopiers, or cameras, all of which can be used to infringe on copyrights. Manufacturers of those products aren’t held to account when that happens, so why should AI.
28
216
829
3
34
182
Ahem.
Try millions(billions) of uncredited data point taken to author for-profit software.
If the bar for training is this low, there is no bar. Copyright death.
9
15
176
How is that platforms can both have Section 230 immunity AND be able to sell data that's not theirs?
If the data uploaded isn't yours, how tf are you selling it??
This does not compute.
6
40
221
I called this line of thinking.
As if voice actor/imitation is synonymous with machine learning deconstructing and reconstructing someone's performance.
We need to enforce our laws. Failing that, draft new ones.
@DeadProxy
@AbelowRob
License for what? The tamber of a human voice? Lmao. The only thing that’s questionable is the use of an artists likeness in tiktok promo when they aren’t involved with a song.
4
0
8
3
19
171
How is anyone to know they *have* to opt-out?
What we have are arcane sources of datasets, from vaguely described locations being used to train models (w/ bs research disclaimers).
New tech and an "info should be free" ethos doesn't give carte blanche to do as one will. /1
3
23
176
Imagine using AI to generate a cover song IMMEDIATELY after the song's release. Then going on to praise the virtues of AI.
It's just annoying derivative coat tail riding behavior. You do you, but c'mon. This is simply rude and disrespectful.
@DeanSpencerArt
Not discouraging but encouraging that you consider using
#aiart
as a tool. Like I took your
#art
and asked AI to re-create a portrayal (it even left in the lighting), here is what it created
#dndart
517
1
32
1
24
173
Truth. They should sue.
😐
hoping AI companies remove your work from their datasets (after already including it without consent) is not going to be enough
17
209
894
4
20
169
The thing about ballpoint pens is *one* could use them to forge someone else's handwriting after TONS of human practice and effort.
Or, with 15 minutes of a recorded voice + a next-gen iPhone; *anyone* can prank call Grandma to ask for her pension money.
Democratization, baby.
4
22
160
Imagine having to build ML on licensed data.
@survivor_343
Imagine you were a portraitist when photography was invented.
Imagine you were a performing musician when recorded music was invented.
Imagine you were a scribe or calligrapher when the printing press was invented.
168
2
52
2
18
152
Artists - this is a big deal for visibility into LAION.
Brute force searches of "artist name" on 5 billion lines of text would take me 40 min+. They are now available to us in a few seconds.
Opt-out or not, awareness is half the battle. What you do with that info is up to you
Thank you,
@mfrankDude
for pointing out that the new caption searches on use the search terms in any order!
Today we've added an ordered search option using quotes. So quoting "artist name" will match the terms in order and return even better results!
1
3
12
2
36
148
It's amazing that ML trainers pre-emptively claim Fair Use as a DEFENSE. Essentially saying,
"We know we are infringing copyright. We believe and hope this is legal. The only way to prove us wrong is to TAKE US TO COURT. Feel free to SPECULATE in the meantime 🖕🖖"
4
32
148
Not mad at this. If you wonder why I oppose genAI using artist works, it's because (to me) it feels like an appropriation.
It takes from the creators and doesn't give back, much like weasely lawyers arguing over creator's rights.
We *do* need more artist focused IP rights.
2
20
153
If anything, we need more clarity that training ML models is a direct derivative of the artists themselves.
Impersonation on outputs alone doesn't cover commercial gains won by Midjourney and Stability AI.
They do more than allow artist name prompts, they train ON their work.
Today, Adobe announced our support for the
@WhiteHouse
Voluntary AI Commitments to promote safe, secure, and trustworthy AI. In addition, we’re committing to develop AI to empower creators, including by proposing a new federal anti-impersonation right.
53
19
50
6
30
149
AI enables the worst kind of entitlement, as if someone *deserves* a seat at the table for writing knock off jokes and using Carlin's voice.
No, dude. We have every right to be critical. Go study up on whatever "RIP profession" and learn some f'n etiquette.
4
26
144
We've all seen it.
"something something rhetorical question"
Something something.. "so I asked ChatGPT to blah blah blah <bluecheck mark> for me"
---------------
Just .. don't ? If we cared what ChatGPT 'thinks' we KNOW WHERE TO GO ALREADY.
6
17
146
I feel like many who want to normalize using unconsenting artwork for AI training, haven't done any real digging.
Have they looked on ?
Have they seen the URL totals?
I wanna believe it matters, that if one could see all the content, they'd think again.
4
24
145
Agree, there is a missed opportunity here to develop such technologies as a cooperation, not appropriation.
It's concerning to see tech development set apart from the arts. By taking without asking, tech frames itself in opposition, competition, yearning to "win" or "disrupt".
But while these models can be used to imitate an individual style, reducing their function to mere copying (or even stealing) misses the point: they represent not so much an appropriation of individual works as an appropriation of collective creativity by private companies
11/14
3
27
224
3
20
147
Hello
@amazon
@ajassy
@JeffBezos
!
Your CDNs (images-na.ssl-images-amazon.comm, , +more ) hold 93+ mil URLs in LAION5B dataset used to train Stable Diffusion.
Per your ToS, copyright is owned by you or your suppliers. /1
3
29
139
Hi
@Hurricanehugo13
-- this is a follow up thread to address this question. Are copyrighted works used to train AI image generators?
(This is a separate question from do AI image generators recreate copyrighted content.)
The answer is unequivocally, YES. 🧵
@mfrankDude
and if there was definitive proof that these systems were improperly scraping and using bona fide copyrighted material that is a lawyer's dream of a class action lawsuit, once again no major action taken in regards to what
1
0
0
3
22
140
Highlights from
@kortizart
@StevenZapata
@SaveriLawFirm
's talk "Art and Law: Empowering Artists against AI"
@LightBoxExpo
"The cat is not out of the bag. Many people we talk to are in a position to make an impact. They hear us and say 'that doesn't sound right, we can fix it"🧵
2
30
139
"
There’s no point in acting surprised about it. All the planning charts and demolition orders have been on display at your local planning department .., so you’ve had plenty of time to lodge any formal complaint and it’s far too late to start making a fuss about it now.
"
It's now possible to block ChatGPT's crawler on any website you control.
Added the block: it's a no-brainer.
Why? ChatGPT cites no sources. It's a one-way relationship where OpenAI takes what is published on the internet, and then doesn't give a single reference or link back.
119
2K
5K
4
16
141
We can only hope. In discovery, Midjourney would need to disclose the original generated image so we can see the alterations he's made.
We're all guessing it's been hardly altered.
This looks like it's going all the way to the Supreme Court.
10
8
34
2
16
133
Young artists are getting their first glimpse of how artwork gets appropriated by others for profit.
A tradition with a long history - art contests, art tests, straight up copyright violations.. now ML training.
If we don't object to this now, when will we?
4
23
133
Shopify users now have a method to find and remove their work from training. Amazing!
Spread the word!
The service is up! Try it out and see if your Shopify has been scraped!!
0
7
26
3
32
123
If anyone wants to start a "post-studio" Hollywood -- sign me up. I'll be your VFX Artist.
We can let them generate derivative nonsense forever while we all actually respectfully collaborate on real, human artist driven projects.
The greed is almost a legendary comic trope - one fun fact - when I was a youngin- I did a film (with a boom box ) and somehow I got points - net not gross.
Never expected to see any money - but the film became quite famous - so about 10 years ago - I looked again at
1K
8K
58K
4
14
119
GenAI case study. An open eyed Prisoner's Dilemma - two creative companies come to diff conclusions.
Ethics, infringement - "Don't you know that's derived from every other board game/RPG creator's artworks?"
Profit, speed, efficiency - "I can't *NOT* use AI because I must"
Seeing a major game/publisher openly use AI art for a very well funded game is equal parts infuriating and disappointing.
I can't say it any clearer: There is no ethical use of AI art for profit currently. There is no model, no work around, that isn't intellectual theft.
31
132
679
2
19
123
There's a kind of irony to Creative Commons pro-AI articles that claim all publicly viewable art as Fair Use for AI:
MFs you are the creators of CC0!
Keep building a culture of well-defined sharing and stop supporting grey area bullshit appropriation
5
15
120
AI music authors relying on Fair Use arguments are grossly underestimating the litigious nature of the music industry.
See "Blurred Lines" case.
A ruling that seems unjust for human authors but 100% correct for ML derivative works.
4
15
111
Arts and entertainment are 4.2% of the US GDP. That's 966,000,000,000! Way too much money ppl be spending!!
But no worries, we're gonna take all the art for FREE and pass savings on to you! 0% GDP dawg!
(along with every other industry we can figure out how to plagiarise ✌️)
1
17
114
Generative AI demonstrates a lack of this kind of thoughtfulness.
Where is the sketch? The heart of the piece? The *idea* that is borne from the artist and held true to the end.
A traditional artist could stop at the 10-15% mark and you'd still KNOW the most important ideas.
2
7
106
What I find exceptional about the logic behind
#2
--
If software is not to be thought of as software, but instead something worthy of personhood, why are you inventing digital slaves?
Why do people fear AI? I hear three reasons:
1. Cynicism — the belief that it is rational not to cooperate
2. Humanism/racism — systematic bias against machines, denial of their potential moral worth and personhood
3. Conservatism — fear of change, fear of the other tribe
None
83
27
131
6
13
108
The idea that GAI weights are not infringing is bullshit.
Shifting RGB values to statistical spreadsheet values doesn't magically absolve you from re-distributing a *new, licensable, property* built on unlicensed data.
2
18
104
I love that all these ToS invite you to upload whatever you want, ask you to authorize derivatives, then handwave liability in the case of copyright infringement.
Anything more strict than that would "break the internet" guys. You don't wanna break the internet, do you?
2
14
106
There needs to be a third way option where open source =/= wrecking the economies of unlicensed data creators, infringing on trademarks, and playing fast + loose with rights of publicity.
Keeping AI as closed source is unethical. It could delay crucial inventions like curing diseases and solving climate change. By sharing AI as open source, we can solve these problems faster and better.
122
75
300
5
18
103
Even the fact that no image is kept on screen for more than a second.. it's flashy, but I'm over it just as fast.
The sad thing is it can only be helped by feeding it even more data, or targeting a real human's work with narrow fine-tuning.
1
5
97
Hi
@Reddit
, do you also have plans to protect image content hosted on your platform?
LAION5B contains ~345,000 urls from *.redd.it CDNs
4,357 of those results ranked above 6+ in Aesthetic score.
2
20
98
In light of discovery of CSAM material found in the LAION5B dataset, whatsinlaion is down.
There is no reliable filter method to offer broad access to LAION urls and insure this content is not transmitted.
I'll be considering the scope / utility of this project moving forward.
4
26
97
If we're gonna try to refactor IP laws just for AI.. can we first acknowledge that Arts and Entertainment rakes in 271 billion to the US GDP?
@Rahll
I get what you’re saying, but if I wanted to copy marvel’s frames, couldn’t I just screenshot the movie? 🤔 and if I’m a photorealistic artist, at what point does my art become copyright infringement?
All I’m saying is that there’s a debate to be had about how the concept of IP
27
0
5
4
26
95
How is the OS community THIS dense though?
80-90%+ of these open source ML projects are built on infringing datasets.
If you want people to respect your licenses RESPECT EVERYONE ELSE'S.
The open-source community may soon have to transition to using BitTorrent to distribute checkpoints for large machine learning models, rather than relying on centralized model hubs.
49
95
860
5
15
93
There are some who wish to abolish copyright, or interpret it a manner that suits their needs (which we all do).
There is an economics to the current model that we could be lose in this process.
Licensing, residuals, IP support markets rewarding value post-creation, not before.
1
15
91
@rom1504
It's interesting that you flip the ethics of choice on its head.
That it is more ethical to *not ask permission* than it is to make a simple request.
It might be tricky to scale, or costly, but patience would ultimately be on your side. IP protection isn't forever.
1
1
92
The
@FTC
is concerned with unfair competition, not sure why they'd need to hear from those unaffected?
There is no requirement that Generative AI consumes unlicensed data points to compete with the original creatives.
@BadMuthaHubbard
There are lots of artists who welcome generative AI and use it in their creative practice
37
0
10
1
15
90
For anyone using random, unlicensed datasets and making open source projects that allow commercial use --- YOU are the problem.
If you're doing the same and charging money for software, F*** YOU on top.
0
21
89
The aesthetic doesn't really belong to anyone, it's just crunched numbers and text prompts.
The most interesting thing is that it's possible. Completely overshadows any artist name or statement.
2
3
84
Authors: I'm 100% confident -
The missing piece of your self-published best seller is that AI generated cover that demonstrates how highly you value originality.
Don't be left behind.
2
9
85
Anyone who thinks taking humans out of the loop will work under all conditions has never owned a Roomba.
4
20
87
I'm saying... $10k/mo Midjourney users??
For anyone using random, unlicensed datasets and making open source projects that allow commercial use --- YOU are the problem.
If you're doing the same and charging money for software, F*** YOU on top.
0
21
89
6
6
83