|
How far are we from the situation where AI is able to create and label its own training data, creating an infinite learning feedback loop?
BoldFace fucked around with this message at 14:26 on Sep 16, 2022 |
#
¿
Sep 16, 2022 14:24
|
|
|
|
| # ¿ May 14, 2024 06:14 |
|
|
I wonder if tracing becomes even more common with the advent of AI image generation. To me, it seems more ethical than straight up copying another artist's work.
|
|
#
¿
Sep 18, 2022 18:47
|
|
|
I have always been skeptical about all the VR stuff, but if we can train an AI to build huge 3D models of places like these to an arbitrary degree of detail, I wouldn't mind exploring them.
|
|
#
¿
Sep 21, 2022 00:02
|
|
|
https://twitter.com/minimaxir/status/1572272628504883200
|
|
#
¿
Sep 21, 2022 02:52
|
|
|
https://twitter.com/MetaAI/status/1575495462894723072
|
|
#
¿
Sep 29, 2022 19:19
|
|
|
For those with weak or even average GPUs, using Google Colab can be a viable alternative to start experimenting. It doesn't even cost anything unless Google decides that enough is enough and closes the tap to free GPU resources.
|
|
#
¿
Oct 4, 2022 21:08
|
|
|
Not creepy at all. https://youtu.be/aoQ5EUjN_LM 
|
|
#
¿
Oct 11, 2022 18:50
|
|
|
Speaking of blowback, a new chatbot startup character.ai from former Google and Meta AI engineers had to close their site for days and completely purge and lock down their discord server. The reason? They announced that moving forward, they will not support users having sex with their bots. DDoS followed.
|
|
#
¿
Oct 12, 2022 16:29
|
|
|
mcbexx posted:I hope someone can find a way around the problem with extra anatomy on images with certain aspect ratios or exceeding the resolution threshold by too much. Can't you generate the initial image at 512x768 and when you get a good result (or at least one with the right number of limbs), use it as a starting point in img2img with higher resolution output?
|
|
#
¿
Oct 12, 2022 17:19
|
|
|
Sometimes I wonder if these cat pictures would look uncanny to other cats, despite looking normal to humans.
|
|
#
¿
Oct 12, 2022 19:23
|
|
|
It's pretty mindblowing to think how many concepts + instructions of how to draw them can be fitted in just 4GB.
|
|
#
¿
Oct 12, 2022 20:11
|
|
|
It felt more like a "robin hood" leak to me than someone trying to build a new cryptocurrency mining botnet. I have no doubt that even the open source anime models will reach the same level of fidelity soon enough.
|
|
#
¿
Oct 14, 2022 04:15
|
|
|
On the topic of text generation, I have been thoroughly impressed with https://character.ai/ . Things like Github Copilot for code generation and OpenAI's text completion with GPT-3 are impressive, but at the end of the day, they still feel like tools. The character.ai bots are not like the chatbots from ten years ago. These things feel like they have a personality and they can maintain it through long conversations. It's not just verbal interaction either. You can start a fictional boxing match with them and they will describe physical actions they take and how they grow more tired as the match progresses. You can give character a coin and after some time ask them to give it back to you. There is very little handholding needed, and the bots seem to pick up even subtle verbal clues. It feels like we're on the cusp of a major paradigm shift. Computers are finally starting to understand natural language on a more sophisticated level than Apple's Siri or Google's search engine. The impressive thing about Dall-E and Stable Diffusion is not how pretty the images are, but how they can transform written language into something else without losing too much of the original meaning. If humans understand language and computers understand language, a completely new channel of communication is formed. It will fundamentally change how we interact with technology in the future.
|
|
#
¿
Oct 17, 2022 05:01
|
|
|
I've heard somewhere that images in batches use sequential seeds. If the seed number is set to 340129172, fourth image in the batch would use the seed 340129175.
|
|
#
¿
Oct 17, 2022 14:15
|
|
|
Elotana posted:Is the NovelAI model good for anything other than single-subject waifu portraits? I saw that it has better coherency than StableDiffusion at non-square aspects, but that's a pretty low bar to clear and MJ + D2 (with outpainting) can handle those fine too. Is there any improvement to the practical things like action shots? Hand-object interactions? Multiple characters? I think it's a problem with the training data. The training images have human-generated tags like "1girl, 1boy, red hair, blue hair, apple", but there is no tag for "a red-headed girl giving an apple to a boy with blue hair". If there is only one subject in the image, almost all of the tags refer to that subject and the AI is able to draw it pretty well. It is possible to use something like inpainting to select different parts of the image to draw different characters in, but it's difficult to have them interact with each other.
|
|
#
¿
Oct 17, 2022 17:04
|
|
|
https://twitter.com/StabilityAI/status/1582053531686645760
|
|
#
¿
Oct 17, 2022 18:12
|
|
|
https://twitter.com/_akhaliq/status/1582825597059104769
|
|
#
¿
Oct 20, 2022 02:56
|
|
|
What exactly is the dynamic here? Why was this model not published through the same org as previous versions?
|
|
#
¿
Oct 20, 2022 18:19
|
|
|
It's a big mess. Seems RunwayML also released a 1.5 model specifically finetuned for inpainting. HuggingFace already incorporated support for this inpainting model to their diffusers-library and deprecated the way it used to work.  
|
|
#
¿
Oct 20, 2022 20:30
|
|
|
|
|
#
¿
Oct 20, 2022 20:39
|
|
|
Ripper Swarm posted:open-source legality aside, what does 1.5 have over 1.4? Maybe better inpainting performance with that finetuned model? I'm still waiting to see comparison images.
|
|
#
¿
Oct 20, 2022 22:09
|
|
|
Stability.ai's CIO made a blog post and answered some questions on Reddit. They maintain that 1.5 was leaked by RunwayML, which they call "a bad faith actor" who reneged on promises made privately. Someone from Stability.ai did request HuggingFace to take down the model, but Emad on discord clarified that it was not a "legal takedown request". Later they dropped the takedown request altogether and it seems that the leaked model can stay up. https://danieljeffries.substack.com/p/why-the-future-of-open-source-ai 
|
|
#
¿
Oct 21, 2022 04:39
|
|
|
Emad has said that they will not release NSFW models going forward. 
|
|
#
¿
Oct 21, 2022 18:40
|
|
|
Lexica.art seems to have added free image generation to their site. I remember them receiving hefty amount of funding a while back. I wonder what their eventual business model will look like. 
|
|
#
¿
Oct 21, 2022 23:29
|
|
|
They did ban all other AI art, though. https://twitter.com/DiffusionPics/status/1584957146256506883
|
|
#
¿
Oct 25, 2022 19:44
|
|
|
Similar vibes. https://twitter.com/_akhaliq/status/1585608077042147328
|
|
#
¿
Oct 27, 2022 17:00
|
|
|
Character.ai added image generation (probably stable diffusion) option for bots.
|
|
#
¿
Nov 9, 2022 15:48
|
|
|
I haven't tried it myself, but someone seems to have finetuned SD1.5 with Midjourney v4 images, which sounds pretty interesting. https://twitter.com/rameerez/status/1590883316022284288
|
|
#
¿
Nov 11, 2022 06:55
|
|
|
https://twitter.com/EMostaque/status/1598182752523722752 Also, seems like the bot functionality to generate images for free on the official SD discord is back. Uses SD2.0 and produces 768x768 images.
|
|
#
¿
Dec 1, 2022 22:20
|
|
|
I wonder if it would be easier to start with something like tracker music. 
|
|
#
¿
Dec 3, 2022 19:47
|
|
|
The judicial system will become much more efficient after a judge AI has been trained by feeding it all written court records.
|
|
#
¿
Dec 3, 2022 20:00
|
|
|
frumpykvetchbot posted:It feels like magic. I'm having a fruitful conversation with a machine. 
|
|
#
¿
Dec 9, 2022 22:48
|
|
|
https://twitter.com/sharifshameem/status/1602416016360759296
|
|
#
¿
Dec 13, 2022 02:53
|
|
|
TheWorldsaStage posted:There's an open revolt happening on Artstation If I'm being completely honest, I had never even heard of Artstation before Stable Diffusion. If anything, AI has been good for their business.
|
|
#
¿
Dec 14, 2022 17:45
|
|
|
https://twitter.com/weirddalle/status/1603023590303338498
|
|
#
¿
Dec 14, 2022 18:38
|
|
|
KakerMix posted:https://www.forbes.com/sites/paultassi/2022/12/14/high-on-lifes-walls-are-covered-in-midjourney-ai-art/?sh=610427474142 I'd say it's an improvement over how things used to be (and still are). 
|
|
#
¿
Dec 14, 2022 20:44
|
|
|
https://twitter.com/MrBeezul/status/1603120528935428097
|
|
#
¿
Dec 14, 2022 21:22
|
|
|
This site seems like a pretty good combination of a ChatGPT-like model and a search engine. If Google doesn't start doing something similar, it risks being left behind. https://www.perplexity.ai/ 
|
|
#
¿
Dec 16, 2022 03:13
|
|
|
I guess that checks out. 
|
|
#
¿
Dec 16, 2022 23:04
|
|
|
|
| # ¿ May 14, 2024 06:14 |
|
|
This will be much more interesting than that silly class action, but I still don't think they have a real case.
|
|
#
¿
Jan 18, 2023 00:24
|
|