@timnitGebru The #LAION origin story is quite mind boggling. It's like the #XKCD "Dependency" cartoon, but for evil - https://www.bloomberg.com/news/features/2023-04-24/a-high-school-teacher-s-free-image-database-powers-ai-unicorns

A High School Teacher’s Free Image Database Powers AI Unicorns https://archive.vn/UIu69 #AI #images #LAION (interesting)

#Verpasstodon

ChatGPT-Alternative OpenAssistant: Eine Konversations-KI für alle

Die LAION-Community hat mit Andreas Köpf und Yannic Kilcher eine Open-Source-Alternative zu ChatGPT gebaut: OpenAssistant ist samt Datensätzen frei verfügbar.

https://www.heise.de/news/OpenAssistant-Konversations-KI-fuer-alle-ChatGPT-Alternative-frei-verfuegbar-8963873.html?wt_mc=sm.red.ho.mastodon.mastodon.md_beitraege.md_beitraege

#ChatGPT #Chatbots #KünstlicheIntelligenz #LAION #LinuxundOpenSource #MachineLearning

This petition from #laion is likely our best bet for equitable, human-friendly AI development in this decade.

In the name of scientific progress we collectively built the Large Hadron Collider. Let’s do it again for #ai

Sign this petition to democratize AI research by establishing an international, publicly funded supercomputing facility equipped with 100,000 state-of-the-art AI accelerators to train open source foundation models.

https://laion.ai/blog/petition/

#fediverse assemble!

After 3 days of adding countermeasures to the AI Horde ratings, the amount of ratings we record per day, has dropped significantly, but we're already seeing a improvement in quality.

It's just a pity I had to again waste my time from improving things, to prevent people from ruining everything.

#aiart #laion #StableDiffusion #aihorde

Adobe's #aiart app debuted today, and they're talking big about using licensed training data and compensating creators.

AFAIK it's the first general-purpose AI tool from a corp with a content library big enough to compete with scraping. They've used that library for years, powering things like content-aware fill in #photoshop

It's a big move, potentially a declaration of war on #LAION startups and #Midjourney.

*But...*

(1/🧵 )

#Adobe #Photoshop #Firefly #AI

#OpenChatKit provides a powerful, open-source base to create both specialized and general purpose #chatbots for various applications. We collaborated with #LAION and #Ontocord to create the training dataset. We are releasing OpenChatKit 0.15 under an #Apache 2.0 license, with full access to source code, model weights and training datasets. This is a community driven project, and we are excited to see how it develops and grows. https://www.together.xyz/blog/openchatkit

Stable Horde in cooperation with hlky just provided 250K ratings, including 50K images with 3 ratings each to @laion_ai.

These are based on the diffusiondb dataset. Our amazing community is churning out about new 100K ratings a week!

#stablediffusion #laion #stablehorde

I ask anyone choosing to use AI art apps, to consider & be curious how they work. It's not just harmless fun. This video by artist Steven Zapata, gives a breakdown of exactly how AI art apps work https://youtu.be/tjSxFAGP9Ss #midjourney #stablediffusion #laion #lensa #prismalabs

Sarah Andersen, cartoonist and the illustrator of the comic, "Sarah’s Scribbles" at https://sarahcandersen.com/, has written an excellent and deeply personal article about AI, alt-right, art, and appropriation.

https://www.nytimes.com/2022/12/31/opinion/sarah-andersen-how-algorithim-took-my-work.html

#SarahAndersen #Art #AI #AIArt #ArtificialIntelligence #AltRight #StableDiffusion #Midjourney #DALLE #LAION

Exploring 12 Million of the 2.3 Billion Images Used to Train Stable Diffusion’s Image Generator - some interesting info/insights on the #LAION dataset. #stableDiffusion #ai #ML

https://waxy.org/2022/08/exploring-12-million-of-the-images-used-to-train-stable-diffusions-image-generator/?unapproved=45149&moderation-hash=1fb24385014533d79611aaaee1c5686a#comment-45149

“Data sets such as LAION-5B must be deleted and rebuilt to consist only of voluntarily submitted work. AIs trained on copyrighted art must also be pulled. Above all, the work of real people should be valued, not exploited to enrich a few tech plutocrats. We are, after all, on “team human.””

— Molly Crabapple

#AI #LAION-5B

Now, "Common Crawl" just scrapes all data from the web indiscriminately. They only respect websites that explicitly ban THEM specifically in the "robots.txt" file.

They offload the responsibility of actually sorting through the meta tags, robots.txt etc down the chain as part of their Terms of Use. So technically, anyone who uses CC data needs to respect robots.txt and meta tags.

#LAION doesn't even check the HTML or robots.txt. They just download everything.

Nobody is accountable.
#Ai

RT @camenduru@twitter.com

automatic1111/stable-diffusion-webui space is trending on 🤗 Hugging Face 🎉 running with A10G GPU thanks to @_akhaliq@twitter.com and @multimodalart@twitter.com ❤ and @gradio@twitter.com team #stablediffusion team ❤ #laion team ❤ and all stable-diffusion-webui devs ❤
https://huggingface.co/spaces/camenduru/webui

🐦🔗: https://twitter.com/camenduru/status/1604495019955494912

This is crazy, #StabilityAI is asking artists to opt-out manually their artwork from the #LAION dataset when the question is why are these images in the dataset in the first place
https://arstechnica.com/information-technology/2022/12/stability-ai-plans-to-let-artists-opt-out-of-stable-diffusion-3-image-training/

Opt out is now available for the #LAION dataset! This is the dataset underpinning many generative models but especially #stableDiffusion go to http://haveibeentrained.com to opt out or opt in now!

Consent matters and this is the way to revoke consent to your works being trained!

Stability AI plans to let artists opt out of Stable Diffusion 3 image training - Enlarge / An AI-generated image of a person leaving a building, thus op... - https://arstechnica.com/?p=1904587 #machinelearning #stablediffusion #emadmostaque #stabilityai #spawning #biz⁢ #optout #aiart #laion #ai

Opt out is now available for the #LAION dataset! This is the dataset underpinning many generative models but especially #stableDiffusion go to https://haveibeentrained.com to opt out or opt in now!

Consent matters and this is the way to revoke consent to your works being trained!
Explainer vid here:
https://youtu.be/SMMjqKWLsbc
Boosts welcome!
#mastodonArt #aiArt #art

New dataset for training AI image generators. Contains images with text description.

#laion #laion5b

https://laion.ai/blog/laion-5b/

Artist finds private medical record photos in popular AI training data set - Enlarge / Censored medical images found in the LAION-5B data set used t... - https://arstechnica.com/?p=1882591 #machinelearning #stablediffusion #googleimagen #consent #privacy #biz&it #laion #ai