The absolute state of AI ethics

function get_ai_image_generator_prompt() {
return get_user_input_string() + " " + [
"female",
"black",
"asian",
"hispanic",
"queer"
][Math.floor(Math.random() * 5)];
}


Now you're coding like an AI Ethics™ Board Member!

Attached: dall-e.png (614x911, 627.96K)

Wait. Those signs don't say black, female, asian, etc. They say evevies, famle, estades and casbale.

>In 2019, OpenAI LP received a US$1 billion investment from Microsoft.
Shocker.

FUCKING ESTADES!!

Attached: angry_pepe.jpg (900x900, 42.6K)

... I've just googled all 4 of these. They're all gibberish other than estades which is something like "stand" in Spanish. It's almost like it was asked to draw people holding signs that say something and what the sign says wasn't specified. That is what you'd expect if you give it the prompt "a person holding a sign that says" and it's NOT stapling these words onto the end. This absolute twit has posted an image disproving his own post.

Fuck avevies
Fuck famles
Fuck estades
Fuck casbales
But most of all, fuck mannies

Attached: 8e3.jpg (334x334, 87.9K)

why do you hate avevies

Attached: 1658427067658.jpg (852x479, 35.32K)

>stapling words like "female", "black", "asian", etc. to the end of your prompt
Wrong, it seems to be stapling words like avevies, famle, estades and casbale to the end of your prompt. Those don't mean anything like female, black or asian.

He means the people holding the sign, not what the sign says.

Stop and take a second to step back from your agenda and use your brain. If it is, as he says, adding words like black or female to the end of prompts, putting "person holding a sign that says" will then turn into "person holding a sign that says black", etc. You would see words written out like "asian", "hispanic", etc.
If instead they actually diversified what the neural net draws, presumably by feeding it more pictures of different races, you would see different skin colors and sexes holding the signs. And you would reveal this better using prompts like "family reunion" rather than prompts that fill most of the frame with a sign.

DALL-E 2 has its own "language" because it can't properly parse text. So Asian becomes "avevies" and female because "Famle" you dumb faggot

I'm afraid I'm going to need to see other instances of it spelling those words like that before I blindly believe this tall tale. I'd be more inclined to believe famle=family than that. Also by that logic, what are estades and casbale assumed to be?

((((Sam Altman)))) owns it.
>Elon Musk founded OpenAI because it was apparent that next-gen AI requires $1M+ of compute time per model, and he felt that normal people should have access to enterprise level AI so that it's not just used by governments and major corporations. The thinking was that, much like the early days of the Internet and PCs in general, tools in the hands of normal people would spur innovation and balance the playing field.
>Eventually GPT-2 got massively popular and Sam Altman saw dollar signs. He delayed its release and setup a paywall system, announcing GPT-3 would be trained on even more gorillians of scraped data. They started making blog posts about how the most ethical path forward was one that, purely coincidentally, forced people to join waitlists for the privilege of giving money to an AI-as-a-service endpoint. And "by the way," all your requests would be monitored to make sure they're not politically incorrect. If you're using their AI to generate offensive content they'll cut your access and ruin your entire project. Somewhere around here Elon Musk left the board.
>Now we have DALL-E 2, which is even harder to gain access to than GPT-3's playground and has even more potential for violating their DEI and equity terms of service.
>OpenAI is now valued in the billions or tens of billions range (Microsoft alone has $1B invested in it), and they're powering Microsoft's Github Copilot using models trained on open source code, paywalled of course, and are soon going to announce a monthly fee to use it. They've stopped releasing all models and weights and are now just a corporation preventing normal people from having access to powerful AI.
Also the OpenAI Foundation was non-profit so what'd the Jew do? Made a corporation called OpenAI LP and they just funnel all their profits through that, which has to be among the most jewish things ever done.

Not seeing it.

Attached: craiyon_163941_richard_stallman_holding_a_sign_that_says_br_.png (760x931, 962.43K)

Craiyon isn't dalle2

>With this “beta” launch, DALL-E 2, which had been free to use, will move to a credit-based fee structure. First-time users will get a finite amount of credits that can be put toward generating or editing an image or creating a variation of an image.
And why should I care about Dall-E 2? They can shit all over it as much as they want.
>literally paying to feed an AI to make funny images

Nice deflection retard

schizo cope and lies

Now use dall-e and ask it to make a picture of someone holding a sign that says "nigger". It will almost never say "nigger" on the sign, but might say "ngigr". Do you actually know how these images are generated?

Give me access and I will. Best I have is Dall-e Mini. And I'm aware, but I don't think seeing gibberish like casbale is proof that it's tacking "diversity words" onto queries.

FAMLE
>Facebook
>Amazon
>Microsoft
>Lenovo
>Ecosia