121
Pack it up. Nothing can be funnier than this. 
Or perhaps, nothing can be funny anymore. 
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
this post was submitted on 17 Aug 2024
121 points (99.2% liked)
the_dunk_tank
15912 readers
533 users here now
It's the dunk tank.
This is where you come to post big-brained hot takes by chuds, libs, or even fellow leftists, and tear them to itty-bitty pieces with precision dunkstrikes.
Rule 1: All posts must include links to the subject matter, and no identifying information should be redacted.
Rule 2: If your source is a reactionary website, please use archive.is instead of linking directly.
Rule 3: No sectarianism.
Rule 4: TERF/SWERFs Not Welcome
Rule 5: No ableism of any kind (that includes stuff like libt*rd)
Rule 6: Do not post fellow hexbears.
Rule 7: Do not individually target other instances' admins or moderators.
Rule 8: The subject of a post cannot be low hanging fruit, that is comments/posts made by a private person that have low amount of upvotes/likes/views. Comments/Posts made on other instances that are accessible from hexbear are an exception to this. Posts that do not meet this requirement can be posted to !shitreactionariessay@lemmygrad.ml
Rule 9: if you post ironic rage bait im going to make a personal visit to your house to make sure you never make this mistake again
founded 4 years ago
MODERATORS
He doesn't own shit in this case, he just had one of his lackeys make whatever LLM he ripped off for his chatbot prompt a Flux server instance. Flux itself is open source, has nothing to do with him or anything he's touched, and runs on midrange consumer hardware. It's also as horrifying as it is fascinating, because despite only a modest increase in system requirements over Stable Diffusion it's starting to lack the really obvious flaws that earlier models have.
I don't know what Grok is under the hood, because it doesn't make sense for it to be its own independent model over just a modified version of some other presumably open source LLM that had permissive enough licensing for a derivative work to not mention it (or his lackeys just ripped one off and didn't credit it at all), but the image generator that it's prompting is just a Flux instance. So basically one of his lackeys set up some servers running something like comfyui (also open source) servers set to its remote API mode and got his chatbot to send API calls to them on request, and those servers are just running some basic workflow with the default Flux checkpoint.
I just want to emphasize that here he's trying to leach off open source research tech that he doesn't own and isn't involved with in any way.
Exactly. Flux isn't his and has nothing to do with him or his shitty companies and bumbling lackeys, he's just a middleman trying to grift off open source tech.
Extremely horrifying open source tech, but open source tech nonetheless.
Yep. With a relatively modern midrange computer and the most basic of technical knowledge anyone can set up and run at least Stable Diffusion (and if they have an NVidia GPU "relatively modern" extends back to like the better 10 series cards from over a decade ago) and do basically anything with it, limited primarily by their VRAM and RAM vs the image size.
The one saving grace is that despite how trivially accessibly extremely powerful tools are, most of the AI enthusiast community is comprised of dipshit chuds who struggle to operate a simple prompt input box on something like A1111 and cry about how hard and confusing comfyui - which is literally just a node based flowchart that holds your hand through the whole process - is to use.
At this point there's no putting the brakes on it, no. All the tools someone could ever need already exist and either have a broad use (basic computer hardware) or are free programs and relatively small model weight files - wiping out image generating AI at this point would be like trying to wipe out media piracy.
And it's only going to get worse as the relatively crude and inefficient algorithms are improved further. Flux itself required a 3090 or 4090 when it was released because of its extremely high VRAM requirements, and now after just a few weeks people have managed to squeeze it down to run on old 10 series cards with 8GB of VRAM. That's a terrifying photorealistic generator running on ancient hardware, even if it doesn't run well.
Even though the AI bubble is going to burst because actual tech companies are struggling to monetize shitty proprietary image generators and garbage LLMs with no use value and their investors are getting impatient and annoyed at their losses, the tech itself isn't going anywhere and even if the major tech funding dries up there's a ton of open source independent work being done by enthusiasts.