Typical garden gnome pilpul practice where they use terms like "open", "free", and "private" to try and condition people into trusting them, before flipping the table on people and saying something like "oy vey I never said that, I just falsely advertised in a legal manner"
But they have a lot of source code though?
I'm looking here: https://github.com/orgs/openai/repositories?type=all
I haven't kept up with them but it looks like they release source code for stuff after a while that the research itself has been published. So DALL-E 2 will probably get published in a while. What people really are bitching at here is that they aren't able to have free use over their servers. And of course not, that shit costs money and providing it for free use would conversely allow no one to use it, because shit tons of people would be trying to connect to it to "get a chance".
Also to be fair I haven't read the licenses so they could be shit.
>DALL-E 2 will probably get published in a while
gullible, you should look at the current status of GPT-3. any half-decent model is never going to be open sourced by OpenAI.
>OpenAI is an artificial intelligence research laboratory consisting of the for-profit corporation OpenAI LP and its parent company, the non-profit OpenAI Inc.
Google published a more realistic version of Dalle-2. They won’t release it either. Think about the worldwide consequences of virtually zero-cost content creation at scale. It’s coming though…
this would literally solve the cp problem >cp is bad because it hurts real people
but if we generate with ai then there will be no victims
it's a win-win-win situation
Consider piracy of XYZ - you can go to jail because you are causing financial losses that might bring corporations gay out of business. Sounds reasonable, yes? You accept it? Now consider pirating pizza instead - did you cause financial losses to the creators thus protecting children from abuse, or did you support them with that act?
Reminder anything prefixed with the word "open" is nonfree software. It is nothing more than mimicry meant to deceive users into surrendering their freedoms under the false assumption that just because parts or all of the source code is accessible. Please use only free software that protects your freedoms and do not fall for corporate marketing.
Meaning the only reason they aren't releasing shit like this is because free content creation will eat into their profit? It really isn't because of social justice or something?
You know that shit's already illegal right? Even e-boi is illegal in most of the world and unprosecuted in the rest. japan is the exception but the UK, US and UN are always trying to get them to fold
the protect act of 2003 is off the books and if the question was brought before the scotus then it would immediately be ruled as unconstitutional. cope eurogay.
it's exactly about that, don't pay heed to delusional schizos here. cp is illegal because it encourages a black market that rich westerners and poor third worlders will engage in. the same way you will get prosecuted in the US for fucking a child overseas as if you had done it in the US itself, the government protects children wherever they are, or at least attempts to >lalalala they are corrupt aaaaa
wow, insightful. it's no secret that bad actors can easily sneak in harmful and extremely gnomish legislation under the guise of "protecting children" but that doesn't mean trying to stop ivan from abducting and raping children is a gnomish plot.
No, I want to hang you from a tree because you're a sick fuck that is incompatible with society and has nothing to offer it other than moral degradation
If you killed yourself right now you would not only improve the quality of this board but also society as a whole
but where does the training data come from?
we have not yet created NNs that can generate output unrelated to what it has been trained on, if you feed it adults in porn it can made, at best, barely legal/barely illegal stuff, but it wont make porn with 9 y/os or whatever you want to jerk it to
>we have not yet created NNs that can generate output unrelated to what it has been trained on
Not even remotely true, however GANs are notorious for copy-pasting e.g. the "outline" of objects in the train set and then "painting" a texture on it. Other models don't have these problems, it's a consequence of the mode-chasing problem from the dynamics of the two submodels. >but it wont make porn with 9 y/os or whatever you want to jerk it to
That's a completely unrelated problem to which there is already a ready solution: instead of generating from sampling the latents, you provide a reference input that gets encoded, then take a sample from the latents and decode the result. This is similar to how style transfer works.
I think you're underestimating things. This picture is generated by Imagen. I highly doubt they trained it on images of cups made out of fruit floating in liquid and holding various types of seeds. I would be more surprised if you couldn't generate dubious imagery with legal pornography and legal images of minors.
Fuck me, you could build a shitty social network and fill it with bots posting from a big ass IPv6 range, best case scenario it is good enough to attract real people and you can sell it off afterwards.
Literally yes. They let "vetted people" have "access" to their model and "post examples" aka they paid people to post particularly enticing examples while avoiding the bad results, possibly with retouching via photoshop.
>/aids/ is already working on their own image generator. Your petty concepts of morality will not stop us from cooming.
I've tried disco diffusion but I can't make any coomable images.
I'm starting to think A.I. just isn't there yet.
10 months ago
Anonymous
Try centipede diffusion and comment out "load_safety_model"
Reminder anything prefixed with the word "open" is nonfree software. It is nothing more than mimicry meant to deceive users into surrendering their freedoms under the false assumption that just because parts or all of the source code is accessible. Please use only free software that protects your freedoms and do not fall for corporate marketing.
Per the open ai safety requirements, all outputs from applications using GPT-3 have to go through a safety filter first before being sent back to the user.
The looping thing I've personally seen even when using GPT-3 directly with no filter, but you can tune GPT temperature so that it gives more random responses. It's pretty fascinating to use
No, best option right now is NovelAI which uses the open-source GPT alternatives.
It's not as powerful as GPT-3 but it still works great for generating smut.
if you want to talk to GPT-3 you need to apply for access from open AI. if you dont want to do that, you can try out GPT-NeoX-20B which is available for free and scores somewhat comparably to GPT-3 on AI benchmarks.
Because it would be like giving a nuke to a medieval peasant. While most peasant wouldn’t even know what it is, let alone how to launch it, it only takes one clever and determined shithead to cause a disaster.
Fucking retard. Having easy access to knowledge could also be a disaster. It didn't. It wasn't. Having easy access to fire could also lead to absolute disaster. It did and we overcome it. There is literally no reason to not give the public access to AI.
The peasant class doesn't know how to utilize power. The internet and devices connected to it are such tremendously powerful things and the only thing normies know how to use it for is social media posturing, porn, and watching degenerate music videos on youtube. Its truly a parable about what slaves would do if they were suddenly given immense power, they would just simply sink even deeper into their own slavery.
This is why aristocrats are still aristocrats even when they're going through hard times financially, because their substance presents itself through specific behaviors implanted and embedded deeply by good breeding and good stock. What the proliferation of knowledge has done is dilute the tool that only higher ups know how to use and give it to slaves that would only misuse it. This was of course all done by a rouge group of faux aristocrats (banker garden gnomes) with a very pronounced aim of bringing destabilization and exacting vengeance. This all used to be common sense before the French Revolution.
I bet you love being the dogs of aristocracy. Having nothing to contribute to the world besides just following orders. Kys.
10 months ago
Anonymous
You are unironically just a dumb and disposable animal. Aristocrats today are extra corrupt but they are only really bad because they have so many animals like you at their beck and call. Make sure to keep watching porn, smoking weed, getting vaxxed, and accepting your shitty lot in life like they tell you to, dog.
10 months ago
Anonymous
>say how I hate aristocracy >somehow it means I am at their beck and call
No wonder you like the aristocrats so much. You can't even think for yourself. I don't worship rich people. I don't defend corporations. I don't make excuses for billionaires. You are the one that does that. Pathetic.
10 months ago
Anonymous
There you go. Good job proving my point by not being able to actually read what I say. Literacy is wasted on apes like you.
yeah the elites would NEVER abuse this power i mean they are so trustworthy the peasants are just blood thirsty child rapists, the elites are complete angels and only want the best for everyone
The nuke in this case is someone generating a line of text that isn't fully compliant with how OpenAI believes the word latinx should be used.
You'd think giving people access to something like this and then everything (supposedly) being flooded with fake information would further their agenda because they will have created a problem that they can solve in a way that places them as the absolute arbiters of truth, because there's no way their information would be a lie, unlike everything else on the net.
the network structures are publicated you dont need to reverse engineer shit, just train on your own data as the data they used is often not available for public use
There should be a law masking the use of keywords like open, unless they comply to those word's restrictions and guarantees. Any compary that has Open in its name should be forced to be fully open source in the minimun.
Yeah the "open" in their name means absolutely nothing. There are some people trying to replicate DALL-E 2 and Imagen though.
https://github.com/lucidrains/DALLE2-pytorch
https://github.com/lucidrains/imagen-pytorch
There's some site that pulls the images other people have generated with DALL-E 2 as well. https://dalle2.app/
AI is a bad thing for human society and should be restricted and regulated as much as possible. You just want to use it to play around with and make funny images.
Governments want to use it to influence the way you think.
>should be restricted and regulated as much as possible [by the government] >Governments want to use it to influence the way you think
This is retarded thinking. Your fear is governments using it for bad, but you also want governments to control it. The only option is for it to be free (as in freedom).
Because they're a scam company that does nothing new ever. They literally just copy-paste someone else's work, put an eye-catching headline that is as false as could be, and then of course hide their 'tech' because they're lying about it as evidence even by the shitty demos they make available.
The only value in their stuff is the training time and the data used to train. Thus, they can't possibly release what they have. They also can't release a lesser version for PR reasons: on the one hand it could make it look like their product sucks if the free stuff sucks, on the other it allows them to add more toward their PR narrative.
Because they're a scam company that does nothing new ever. They literally just copy-paste someone else's work, put an eye-catching headline that is as false as could be, and then of course hide their 'tech' because they're lying about it as evidence even by the shitty demos they make available.
This is rampant in the AI community. They run some shit on a supercomputer to make a paper with no code that nobody else can reproduce and then get a bunch of money from investors. Rinse and repeat.
Can confirm. Which is hilarious because the academic counterpart had a hard break from most of CS and was extremely adamant on code-for-everything ignore unreleased shit and even refuse to acknowledge or publish otherwise. They even signed a letter saying they're staying with conference formats and refusing to publish in journals for similar reasons. Then a few years ago the last pillars of that side of the community sold out. All the conferences became industry shilling shitshows and nobody cared anymore about reproducibility, instead focusing on muh duhvurrsity
>Google says it would release its photorealistic DALL-E 2 rival – but this AI is too prejudiced for you to use >Aside from technical concerns, and more importantly, Imagen's creators found that it's a bit racist and sexist even though they tried to prevent such biases. Imagen showed "an overall bias towards generating images of people with lighter skin tones and … portraying different professions to align with Western gender stereotypes," the team wrote. Eliminating humans didn't help much, either: "Imagen encodes a range of social and cultural biases when generating images of activities, events and objects."
Why are these types so scared of reality? Wouldn't it be better for their "cause" if they showed you how evil they think the world the AI faithfully copied is instead of keeping it secret and pretending it doesn't exist while fighting against it on Twitter incessantly.
Half their work unironically goes towards dealing with "ethical concerns" such as text models using terms like "latinx" in a sufficiently positive context.
The people that work at google are actually batshit but luckily all they've done is just brute forcing the existing methods of making bighuge models by throwing enough money at it and not actually anything new.
what about eleutherAI? Pretty sure they wanted to be the opensource version and have made some impressive progress. Of course they don't receive as much funding so its kinda niche.
https://www.eleuther.ai/
Because you will use it if they do
And that's a problem why?
because you might use it for non approved chud things
/thread
>OpenAI
>is actually closed source
pretty scummy tbh
Typical garden gnome pilpul practice where they use terms like "open", "free", and "private" to try and condition people into trusting them, before flipping the table on people and saying something like "oy vey I never said that, I just falsely advertised in a legal manner"
But they have a lot of source code though?
I'm looking here: https://github.com/orgs/openai/repositories?type=all
I haven't kept up with them but it looks like they release source code for stuff after a while that the research itself has been published. So DALL-E 2 will probably get published in a while. What people really are bitching at here is that they aren't able to have free use over their servers. And of course not, that shit costs money and providing it for free use would conversely allow no one to use it, because shit tons of people would be trying to connect to it to "get a chance".
Also to be fair I haven't read the licenses so they could be shit.
>DALL-E 2 will probably get published in a while
gullible, you should look at the current status of GPT-3. any half-decent model is never going to be open sourced by OpenAI.
What's wrong with GPT-3? It seems to be in the list I linked.
>OpenAI is an artificial intelligence research laboratory consisting of the for-profit corporation OpenAI LP and its parent company, the non-profit OpenAI Inc.
This sounds very gnomish.
>Funnel profits from for-profit business into non-profit to avoid taxes, fees and liabilities
Very gnomish indeed.
Because no one is using GPT-J to make cool shit like OpenAI does.
The person making the bot threads is.
Google published a more realistic version of Dalle-2. They won’t release it either. Think about the worldwide consequences of virtually zero-cost content creation at scale. It’s coming though…
this would literally solve the cp problem
>cp is bad because it hurts real people
but if we generate with ai then there will be no victims
it's a win-win-win situation
I could also create infinite variations of your mug, provided I have a single photo of you. I hope you’re thick skinned
why the sudden agression anon?
We already solved that fucking problem with drawings anon, lots of counties fucking refuse to allow it still
*countries
That's not how it works, goy.
Consider piracy of XYZ - you can go to jail because you are causing financial losses that might bring corporations gay out of business. Sounds reasonable, yes? You accept it? Now consider pirating pizza instead - did you cause financial losses to the creators thus protecting children from abuse, or did you support them with that act?
Rules made by the rulemakers exist to fuck you.
why can't life be simple
Meaning the only reason they aren't releasing shit like this is because free content creation will eat into their profit? It really isn't because of social justice or something?
Oh they'll give you social justice alright. Very open one at that.
You know that shit's already illegal right? Even e-boi is illegal in most of the world and unprosecuted in the rest. japan is the exception but the UK, US and UN are always trying to get them to fold
the protect act of 2003 is off the books and if the question was brought before the scotus then it would immediately be ruled as unconstitutional. cope eurogay.
you need to get off this site
>he thinks criminalizing CP is about victims
What is it about then? Genuinely curious.
it's exactly about that, don't pay heed to delusional schizos here. cp is illegal because it encourages a black market that rich westerners and poor third worlders will engage in. the same way you will get prosecuted in the US for fucking a child overseas as if you had done it in the US itself, the government protects children wherever they are, or at least attempts to
>lalalala they are corrupt aaaaa
wow, insightful. it's no secret that bad actors can easily sneak in harmful and extremely gnomish legislation under the guise of "protecting children" but that doesn't mean trying to stop ivan from abducting and raping children is a gnomish plot.
But then the FBI would be out of a job
No, I want to hang you from a tree because you're a sick fuck that is incompatible with society and has nothing to offer it other than moral degradation
If you killed yourself right now you would not only improve the quality of this board but also society as a whole
but where does the training data come from?
we have not yet created NNs that can generate output unrelated to what it has been trained on, if you feed it adults in porn it can made, at best, barely legal/barely illegal stuff, but it wont make porn with 9 y/os or whatever you want to jerk it to
>we have not yet created NNs that can generate output unrelated to what it has been trained on
Not even remotely true, however GANs are notorious for copy-pasting e.g. the "outline" of objects in the train set and then "painting" a texture on it. Other models don't have these problems, it's a consequence of the mode-chasing problem from the dynamics of the two submodels.
>but it wont make porn with 9 y/os or whatever you want to jerk it to
That's a completely unrelated problem to which there is already a ready solution: instead of generating from sampling the latents, you provide a reference input that gets encoded, then take a sample from the latents and decode the result. This is similar to how style transfer works.
>provide a reference input
Where's the problem aside from retardation on your part?
I think you're underestimating things. This picture is generated by Imagen. I highly doubt they trained it on images of cups made out of fruit floating in liquid and holding various types of seeds. I would be more surprised if you couldn't generate dubious imagery with legal pornography and legal images of minors.
>Not sesame seeds
>not dark chocolate sea
It's over.
Your lack of understanding of how sexual addiction works is abysmal.
Fuck me, you could build a shitty social network and fill it with bots posting from a big ass IPv6 range, best case scenario it is good enough to attract real people and you can sell it off afterwards.
>zero cost
If Dalle-2 isn't open to the public then how does anyone know how good it is. Just taking their word for it?
Literally yes. They let "vetted people" have "access" to their model and "post examples" aka they paid people to post particularly enticing examples while avoiding the bad results, possibly with retouching via photoshop.
Good. The public should not have access to this technology
/aids/ is already working on their own image generator. Your petty concepts of morality will not stop us from cooming.
Which board is /aids/ on?
Thanks.
>/aids/ is already working on their own image generator. Your petty concepts of morality will not stop us from cooming.
I've tried disco diffusion but I can't make any coomable images.
I'm starting to think A.I. just isn't there yet.
Try centipede diffusion and comment out "load_safety_model"
>OpenAI
>not open
why did they decide to use this name?
Like bank of america it is a gnomish trick
Reminder anything prefixed with the word "open" is nonfree software. It is nothing more than mimicry meant to deceive users into surrendering their freedoms under the false assumption that just because parts or all of the source code is accessible. Please use only free software that protects your freedoms and do not fall for corporate marketing.
is OpenBSD nonfree?
Basado
>Reminder anything prefixed with the word "open" is nonfree software
name 12
Because it's already being used and you don't even know it
Anyone have a GPT3 API key?
me
im also waiting to get access to DALL-E 2 but i think it'll be a while before i can get it
Do you know of any way I can talk to GPT-3? AI Dungeon, Replika, Emerson bots are too narrow in scope and tend to get caught in loops.
Per the open ai safety requirements, all outputs from applications using GPT-3 have to go through a safety filter first before being sent back to the user.
The looping thing I've personally seen even when using GPT-3 directly with no filter, but you can tune GPT temperature so that it gives more random responses. It's pretty fascinating to use
No, best option right now is NovelAI which uses the open-source GPT alternatives.
It's not as powerful as GPT-3 but it still works great for generating smut.
if you want to talk to GPT-3 you need to apply for access from open AI. if you dont want to do that, you can try out GPT-NeoX-20B which is available for free and scores somewhat comparably to GPT-3 on AI benchmarks.
https://blog.eleuther.ai/announcing-20b/
Because it would be like giving a nuke to a medieval peasant. While most peasant wouldn’t even know what it is, let alone how to launch it, it only takes one clever and determined shithead to cause a disaster.
*couldn’t even figure out
the same can be said of the internet
or a twitter account
>imagine if everyone in the world can instantly see what facts against the elites you have
imagine being this retarded
do you get all your information about AI from media headlines and movies?
Fucking retard. Having easy access to knowledge could also be a disaster. It didn't. It wasn't. Having easy access to fire could also lead to absolute disaster. It did and we overcome it. There is literally no reason to not give the public access to AI.
>Having easy access to knowledge could also be a disaster
It is. It was. That is why it is getting neutered.
How is it a disaster? The peasant class is finally getting more power for themselves.
For one it crashed the birth rates so bad it's literally erasing the societies that committed that mistake. If that is not a disaster I don't know is.
The peasant class doesn't know how to utilize power. The internet and devices connected to it are such tremendously powerful things and the only thing normies know how to use it for is social media posturing, porn, and watching degenerate music videos on youtube. Its truly a parable about what slaves would do if they were suddenly given immense power, they would just simply sink even deeper into their own slavery.
This is why aristocrats are still aristocrats even when they're going through hard times financially, because their substance presents itself through specific behaviors implanted and embedded deeply by good breeding and good stock. What the proliferation of knowledge has done is dilute the tool that only higher ups know how to use and give it to slaves that would only misuse it. This was of course all done by a rouge group of faux aristocrats (banker garden gnomes) with a very pronounced aim of bringing destabilization and exacting vengeance. This all used to be common sense before the French Revolution.
I bet you love being the dogs of aristocracy. Having nothing to contribute to the world besides just following orders. Kys.
You are unironically just a dumb and disposable animal. Aristocrats today are extra corrupt but they are only really bad because they have so many animals like you at their beck and call. Make sure to keep watching porn, smoking weed, getting vaxxed, and accepting your shitty lot in life like they tell you to, dog.
>say how I hate aristocracy
>somehow it means I am at their beck and call
No wonder you like the aristocrats so much. You can't even think for yourself. I don't worship rich people. I don't defend corporations. I don't make excuses for billionaires. You are the one that does that. Pathetic.
There you go. Good job proving my point by not being able to actually read what I say. Literacy is wasted on apes like you.
>Having easy access to fire could also lead to absolute disaster
which is why the elites don't want you to own a lighter.
>Having easy access to knowledge could also be a disaster.
It absolutely was. Mass literacy was a colossal mistake.
Deepfakes already exist, the fuck will OpenAI do that deepfakes haven't already?
yeah the elites would NEVER abuse this power i mean they are so trustworthy the peasants are just blood thirsty child rapists, the elites are complete angels and only want the best for everyone
The nuke in this case is someone generating a line of text that isn't fully compliant with how OpenAI believes the word latinx should be used.
You'd think giving people access to something like this and then everything (supposedly) being flooded with fake information would further their agenda because they will have created a problem that they can solve in a way that places them as the absolute arbiters of truth, because there's no way their information would be a lie, unlike everything else on the net.
because they can't stop it from being casually "racist" when you ask it about blacks.
Just reverse engineer it and make your own version dummy
>make your own version
2 ReLU layers won't cut it anon
the network structures are publicated you dont need to reverse engineer shit, just train on your own data as the data they used is often not available for public use
ok google tutorial how to train an AI on my fapfolder
>just train on your own data
yes I do have few billion pictures and spare A100 GPUs
didnt say its feasible, theres a reason basically none does it, still there are gpt-3 models available on different datasets so people do it anyway
best I can do is about 30,000 images
Cryptography prides itself on being open, why can't AI be the same
There should be a law masking the use of keywords like open, unless they comply to those word's restrictions and guarantees. Any compary that has Open in its name should be forced to be fully open source in the minimun.
Release what?
They've published enough paper to create your own version of what they created, just buy enough super computers.
ClosedAI
e-bois
Could I take this and run it against my furry porn collection to create the ultimate amalgamation of it?
Yeah the "open" in their name means absolutely nothing. There are some people trying to replicate DALL-E 2 and Imagen though.
https://github.com/lucidrains/DALLE2-pytorch
https://github.com/lucidrains/imagen-pytorch
There's some site that pulls the images other people have generated with DALL-E 2 as well. https://dalle2.app/
>https://dalle2.app/
>not a single miku eating hotdogs
soulless
I can think of a good use for this
Can't wait when they manage to replicate DALLE2. I also want to make funny images.
AI is a bad thing for human society and should be restricted and regulated as much as possible. You just want to use it to play around with and make funny images.
Governments want to use it to influence the way you think.
Yep but it's too late to stop it because the nations that take advantage of it will dominate the one that don't so there is no choice
>should be restricted and regulated as much as possible [by the government]
>Governments want to use it to influence the way you think
This is retarded thinking. Your fear is governments using it for bad, but you also want governments to control it. The only option is for it to be free (as in freedom).
I would rather my government control it than China, Russia, or any other government who has a vested interest in destroying my country
They don't need AI to influence the way people think, they already have TV stations that do a much better job at that.
because its all faked to scam more investment and research grant
imagine believing such transparent gnomish trash
>faked
Just say you don't know anything about statistics or mathematical optimization.
>call out obvious garden gnomery
>garden gnome shill shows up with record speed
lmao every time
>Everyone that disagrees with me is just a garden gnome shill
Literal schizo behavior.
>the garden gnome has become flustered, and now resorts to ad hom
lmao
lol
How severe is your lack of self-awareness to not realize that you've been doing the same thing from your first post?
Can you elaborate on the scam? You seem to know a lot about it.
Not him but see
The only value in their stuff is the training time and the data used to train. Thus, they can't possibly release what they have. They also can't release a lesser version for PR reasons: on the one hand it could make it look like their product sucks if the free stuff sucks, on the other it allows them to add more toward their PR narrative.
>openai
>is not open
>openai
>is not ai
>"Open"
>it's closed
Because they're a scam company that does nothing new ever. They literally just copy-paste someone else's work, put an eye-catching headline that is as false as could be, and then of course hide their 'tech' because they're lying about it as evidence even by the shitty demos they make available.
lookup gpt-j
I believe this guy here programs for OpenAI
This is rampant in the AI community. They run some shit on a supercomputer to make a paper with no code that nobody else can reproduce and then get a bunch of money from investors. Rinse and repeat.
Can confirm. Which is hilarious because the academic counterpart had a hard break from most of CS and was extremely adamant on code-for-everything ignore unreleased shit and even refuse to acknowledge or publish otherwise. They even signed a letter saying they're staying with conference formats and refusing to publish in journals for similar reasons. Then a few years ago the last pillars of that side of the community sold out. All the conferences became industry shilling shitshows and nobody cared anymore about reproducibility, instead focusing on muh duhvurrsity
>Google says it would release its photorealistic DALL-E 2 rival – but this AI is too prejudiced for you to use
>Aside from technical concerns, and more importantly, Imagen's creators found that it's a bit racist and sexist even though they tried to prevent such biases. Imagen showed "an overall bias towards generating images of people with lighter skin tones and … portraying different professions to align with Western gender stereotypes," the team wrote. Eliminating humans didn't help much, either: "Imagen encodes a range of social and cultural biases when generating images of activities, events and objects."
Why are these types so scared of reality? Wouldn't it be better for their "cause" if they showed you how evil they think the world the AI faithfully copied is instead of keeping it secret and pretending it doesn't exist while fighting against it on Twitter incessantly.
Because reality is subjective and we have grown out of meme stereotypes created in an era when burning witches was considered a good idea.
>we have grown out of meme stereotypes
Quick, tell me what you think of people who live in the southern USA
Half their work unironically goes towards dealing with "ethical concerns" such as text models using terms like "latinx" in a sufficiently positive context.
The people that work at google are actually batshit but luckily all they've done is just brute forcing the existing methods of making bighuge models by throwing enough money at it and not actually anything new.
They're "open" in the sense that they're openly trying to cause the end times
Just use EleutherAI stuff
what about eleutherAI? Pretty sure they wanted to be the opensource version and have made some impressive progress. Of course they don't receive as much funding so its kinda niche.
https://www.eleuther.ai/
see
HOLY SHIT I MADE MY OWN GF THANK YOU ELON!!!!