- Joined
- Feb 11, 2017
Follow along with the video below to see how to install our site as a web app on your home screen.
Note: This feature may not be available in some browsers.
Questionable context in these screenshots. Especially this one, the smut comment was directed towards someone that said they can't try the new model because they are in public, insinuating that whatever they would generate wouldn't be fit to be seen in a public setting. I know it's hard for current year people to understand humor after multiple layers of irony poisoning and undiagnosed autism, but believe it or not, this is a "joke". (But it's really hard to tell that if you just snip a response without the comment it's responding to.)More news from the coomersphere, the latest edition of Novel AI's image gen is complete ass at the one thing everyone wants it to do (copy the artstyle of existing artists and mixing artstyles) and it is actively devolving into a gigantic meltdown as PR and community managers fail to spin it. And that's a paid service, NAI is not cheap.
It's so bad that it's compared to SD 1.5, with jannies reacting to any questioning of the new model with mutes and bans.
View attachment 7041200View attachment 7041201View attachment 7041274
>you don't NEED to prompt porn anyway, what do you pay $25 per month for?
The meme reaction wave is "reimugate", where they challenge the new model to draw Reimu with extremely questionable results.
View attachment 7041199View attachment 7041270View attachment 7041272View attachment 7041273
As part of their newly announced open source and transparency initiative, Deepseek reveals that their profit margins for their API R1 service ($0.5/MTok output) is... 545%. For comparison, Openrouter providers offer unmolested R1 at ~$7/MTok output. Imagine what the margins on expensive models like OAI's are.
View attachment 7041204
With their current usage at least, they're trying to shift peak demand to off-peak hours with cheaper prices, which at current capacity would raise the margin to ~750%.
View attachment 7041279
"anyone stupid enough to not just use a modified deepseek on a home server must be retarded enough to pay out the ass for our product. they'll literally pay it so they can ask if Barack Obama had sex with a money in june of 1989.What the fuck is this price? Opus is $15/75 with being the best model for creative writing ever released and it's still not worth it, this is more than double.
$75/Mtok input, really?
>a hedge fund destroys a non-profit by being honest about its pricing, leading everyone to question how "non-profit" the "non-profit is"As part of their newly announced open source and transparency initiative, Deepseek reveals that their profit margins for their API R1 service ($0.5/MTok output) is... 545%. For comparison, Openrouter providers offer unmolested R1 at ~$7/MTok output. Imagine what the margins on expensive models like OAI's are
My gay Discord cult is better than your gay Discord cult etc.It's funny how butthurt 4chan gooners are about Drago.
Most users and the overwhelming majority of revenue for these companies is corporate, and I'm not talking real evil gov-corporate contracts like Palantir, just run of the mill companies wanting the latest buzzword tool for their employees. It's far easier for a company to expense a subscription and SaaS licenses for their employees than justifying hardware to run models locally, even if it's dedicated plug-and-play stuff like Digits.to not just use a modified deepseek on a home server
while true, what are they using AI for, every website has that "how can i help you" bot which i assume is AI but much like tech support it doesn't seem to labor intensive and the biggest cost must be being able to handle hundreds of assholes using the site at once and interacting with it.Plus, local models are just not good enough compared to API, there's no contest about that.
This is just false, they're used for images just as frequently.Those reverse proxies are rarely used for images, it's overwhelmingly used for text,
I was talking specifically about those ran by the people named in the complaint, they were mainly LLM proxiesThis is just false, they're used for images just as frequently.
I love how they phrase all this shit like they're some kind of superheros protecting society instead of just a pack of thieving racketeers angry that someone else leeched the shit they originally stole from their customers. Choke on cocks M$.Microsoft claim to have had insiders that ratted out on each other, but I don't really believe it, it was probably spitefags. They must have been monitoring 4chan for a while, since certain links were getting autofiltered/banned for a month now before they went public with it.
You underestimate how lazy and stupid people are, not only there are idiots paying for AI generated content but last time I was in civitai there were users talking how scammers were taking their generations and selling them, when the prompt its right there. Do you see? even the scammers are stupid now, they were taking time to manually mirror the images and cropping the watermarks one by one instead of just using the prompt to generate their own in a second.Who the fuck uses a proxy for DALL-E anyway? SD is free
Because she's a billionaire and the most famous popstar right now, the bad press alone is not worth it.Hilarious how what bothered MS isn't the rampant honest to God fraud done against their third party customers which is an actual crime, or distasteful uses like the metric shitton of loli ERP generated with their system, but goddamn Taylor Swift's ugly mug
That's eerily specific...so every pinoy teenager that googles how to run R1 on their Samsung A20 sees their website first
With how LLMs are trained on prewritten responses, there's a high chance a 4chan user was part of that dataset preparation teamWe're getting closer to the matrix.
View attachment 7049144
GPT-3 had the best greentext generation, I don't know how they did it.We're getting closer to the matrix.
View attachment 7049144
.We're getting closer to the matrix.
it is sort of disturbing how easy it seems to be for those chatbotsGPT-3 had the best greentext generation, I don't know how they did it.
It's all downhill from there.
We're getting closer to the matrix.
i love how Tolken was so pissed about that ending that multiple parts of LOTR are about how he felt the prophecy should have been fulfilled. hence "i am no man" and the walking trees and that third thing.witches tell me "no man of woman born" can kill me
>feel invincible.gif
>enemy army disguises as forest
>should have seen that coming
>learn opponent wasn't born naturally
>realize prophecies were technically true but actually lies
>get head chopped off
>mfw I trusted witches
Also how the fuck could an army disguise itself as a forest?