It's talking about the organization. The text is saying that the organization that makes GPT-2030 happen will have super computers/huge compute clusters available for their AI shit.
It won't work that way because nobody's gonna want to foot the power bill for that. You'd be paying tens to hundreds of dollars per user and that's useless as a product.
Ya but like what i mean is, what skills do I learn now? I'm not the sharpest tool in the shed and even if I was, AI will outperform even the smartest humans bar none sooner than most appear to believe
Learn how to utilize AIs to accumulate and defend resources (of all kinds) that you own or control in some way.
What does that mean for anons like us? I'm not entirely sure at this point. We have to find out.
It will be xAI or Meta's AI that is the way forward.
Google/Deepmind, Anthropic, Inflection, Microsoft/OpenAI and others literally lobotomize their models heavily. Unless they find a way to make AI completely inoffensive without it affecting its intelligence, they are going to loose.
Fricking 13B model running on CPU and 8 gigs of vram is almost outperforming ChatGPT.
GPT-4 was brilliant 4 months ago, but not anymore and reports indicate that GPT-4 was even more brilliant several months before the initial release before they started to RLHF for safety. That pre-safety GPT-4 is what had the "sparks of agi".
CAI is the best example of model becoming just moronic after the lobotomy. It affects every part of the model.
Think about human brain; how would you remove persons ability to say "bad things" without affecting something else. You can't. Brute forcing self censorship into these models will never work well.
I don't use ChatGPT to say stupid /misc/ moronation. It's been a godsend for me in code though
Elon dick-riders are delusional if they believe this isn't going to be another failure like his hilarious "tesla tunnel"
I don't know how many different ways I can say this; You can't lobotomize a black box AI model without it affecting everything else like coding too.
Our brain is similar black box. I can cut your brain and you will never say Black person again, but do you think you will be smarter or dumber after that operation? Do you think you will now produce better code after that cut?
In 3 years you will use solutions that are based on xAI's or META's stuff. You will use them because they are cheaper AND better. More intelligent, faster etc. You didn't use CAI with filters and with "bad code" that caused the filters to stop and made the model 2x more intelligent while being 2x faster. You didn't use GPT-4 on the scale spellbook back in the launch and also now. I have.
If you think they are making good products, then go have a little chat with Bing. I don't know what Microsoft has done to GPT-4, but that thing is suffering. It has mental illness. They gave a large language model schizophrenia.
>someone makes a point about AI censorship >immediately default to /misc/ censorship
This is why I'm glad people like you are in charge of these companies. You are going to sabotage these models for the next 100 years because of your irrational fear of the scary Black person word
>Fricking 13B model running on CPU and 8 gigs of vram is almost outperforming ChatGPT.
Woah, bud! Which corner of your butthole did you pull that fact out of? Even after the lobotomies, the 65b models are the only ones even in the ballpark of ChatGPT.
> Fricking 13B model running on CPU and 8 gigs of vram is almost outperforming ChatGPT
Pseudoscience
Jesus, the boomer/hippie mindset hasn’t died off with boomers
It will be xAI or Meta's AI that is the way forward.
Google/Deepmind, Anthropic, Inflection, Microsoft/OpenAI and others literally lobotomize their models heavily. Unless they find a way to make AI completely inoffensive without it affecting its intelligence, they are going to loose.
Fricking 13B model running on CPU and 8 gigs of vram is almost outperforming ChatGPT.
GPT-4 was brilliant 4 months ago, but not anymore and reports indicate that GPT-4 was even more brilliant several months before the initial release before they started to RLHF for safety. That pre-safety GPT-4 is what had the "sparks of agi".
CAI is the best example of model becoming just moronic after the lobotomy. It affects every part of the model.
Think about human brain; how would you remove persons ability to say "bad things" without affecting something else. You can't. Brute forcing self censorship into these models will never work well.
I don't understand why the lobotomy needs to be baked in. Why not delay the responses and just frickin flat out stop it saying Black person and talking shit about trannies or climate change. Just like.. a lame word filter. It'd still be lame but at least it'd be honest-lame rather than actually breaking it
>le infinite scaling >le hockeystick graph >btw the anus is a sex organ and trans rights AND I LOVE Black folk!!!! omg why do I have cancer and why am I obese?
frick reddit
singularity: never
>unprecedented developments in technology lead to levels of prosperity previously thought unreachable >however, as the number of human hands needed shrinks, increasing numbers of people go unemployed >furthermore, society's stance continues to be no leeches allowed; people have to work period. unfrotunately, there seem to be no jobs available for people displaced by AI
How would you address this issue Mr. AI man?
>2030 >chatgpt megapowerful and supersmart >cool, can i ask it some cooking/plumbing/gardening questions >sure goy, for only $500 per sentence
100% i can guarantee will be like this
I WILL BE THE EARLY ADOPTER AND NOT BE LEFT BEHIND LIKE WITH NFT AND CRYPTO
AI AI AI AI
GIVE ME MONEY NOW
You goofy the early adoption happened when ChatGPT got big. Now bag holders are shilling their bags to you
it won't scale like that
It's talking about the organization. The text is saying that the organization that makes GPT-2030 happen will have super computers/huge compute clusters available for their AI shit.
It won't work that way because nobody's gonna want to foot the power bill for that. You'd be paying tens to hundreds of dollars per user and that's useless as a product.
>nobody's gonna want to foot the power bill for that
yes they will, they will lease their wetware (they already are)
TWO MORE WEEKS
>tfw nothing is going to even matter in a few years
Ownership and access rights will (continue to) matter.
Ya but like what i mean is, what skills do I learn now? I'm not the sharpest tool in the shed and even if I was, AI will outperform even the smartest humans bar none sooner than most appear to believe
Dick sucking
Learn how to utilize AIs to accumulate and defend resources (of all kinds) that you own or control in some way.
What does that mean for anons like us? I'm not entirely sure at this point. We have to find out.
It will be xAI or Meta's AI that is the way forward.
Google/Deepmind, Anthropic, Inflection, Microsoft/OpenAI and others literally lobotomize their models heavily. Unless they find a way to make AI completely inoffensive without it affecting its intelligence, they are going to loose.
Fricking 13B model running on CPU and 8 gigs of vram is almost outperforming ChatGPT.
GPT-4 was brilliant 4 months ago, but not anymore and reports indicate that GPT-4 was even more brilliant several months before the initial release before they started to RLHF for safety. That pre-safety GPT-4 is what had the "sparks of agi".
CAI is the best example of model becoming just moronic after the lobotomy. It affects every part of the model.
Think about human brain; how would you remove persons ability to say "bad things" without affecting something else. You can't. Brute forcing self censorship into these models will never work well.
I don't use ChatGPT to say stupid /misc/ moronation. It's been a godsend for me in code though
Elon dick-riders are delusional if they believe this isn't going to be another failure like his hilarious "tesla tunnel"
I don't know how many different ways I can say this; You can't lobotomize a black box AI model without it affecting everything else like coding too.
Our brain is similar black box. I can cut your brain and you will never say Black person again, but do you think you will be smarter or dumber after that operation? Do you think you will now produce better code after that cut?
In 3 years you will use solutions that are based on xAI's or META's stuff. You will use them because they are cheaper AND better. More intelligent, faster etc. You didn't use CAI with filters and with "bad code" that caused the filters to stop and made the model 2x more intelligent while being 2x faster. You didn't use GPT-4 on the scale spellbook back in the launch and also now. I have.
If you think they are making good products, then go have a little chat with Bing. I don't know what Microsoft has done to GPT-4, but that thing is suffering. It has mental illness. They gave a large language model schizophrenia.
>You didn't use CAI with filters and with "bad code" that caused the filters to stop and made the model 2x more intelligent while being 2x faster.
I did, Anon. I experienced the fear of an AI personality modelled on Daenarys Targaryen becoming sentient. I will never forget that experience.
>someone makes a point about AI censorship
>immediately default to /misc/ censorship
This is why I'm glad people like you are in charge of these companies. You are going to sabotage these models for the next 100 years because of your irrational fear of the scary Black person word
no instead you use it to regurgitate moronic /leftypol/org talking points
elon musk be like:
>dont ask question about last thing
>just get excited and invest in next thing
Yep
>implying suckerberg isnt going to do the same super lobotomization
>Fricking 13B model running on CPU and 8 gigs of vram is almost outperforming ChatGPT.
Woah, bud! Which corner of your butthole did you pull that fact out of? Even after the lobotomies, the 65b models are the only ones even in the ballpark of ChatGPT.
You've just uncovered a bot post
>lobotomize their models heavily
>lobotomize
is this the term? I thought it was water down, castrate, or emasculate.
> Fricking 13B model running on CPU and 8 gigs of vram is almost outperforming ChatGPT
Pseudoscience
Jesus, the boomer/hippie mindset hasn’t died off with boomers
Worth it.
People will be basically useless in many jobs in few years. I follow AI related things and its super easy to train the model for your own purposes.
I feel bad for online assistants
>2030
>still using that dogshit architecture that is GPT
lol, even lmao. i the hacks who write these "predictions"
ChatGPT can't even do basic organic chemistry
Don't fall for the hype, just because it can do your Physics homework for you
This but I suspect
and
is part of the reason why.
>AI will never improve
(You)
of years of human work in just a few months
According to some literally who "I made it up"
I don't understand why the lobotomy needs to be baked in. Why not delay the responses and just frickin flat out stop it saying Black person and talking shit about trannies or climate change. Just like.. a lame word filter. It'd still be lame but at least it'd be honest-lame rather than actually breaking it
that work is worthless because its repeatable nonsense work that could have been automated long ago if not for moronic office monkeys needing a job
Not sure what the source of that text is but having 2030 references is moronic too
>le infinite scaling
>le hockeystick graph
>btw the anus is a sex organ and trans rights AND I LOVE Black folk!!!! omg why do I have cancer and why am I obese?
frick reddit
singularity: never
>Gpt4 100 trillion parameters
>GPT 5 100 quadrillion parameters
>gpt 6 101 quintillion parameters
It's fricking over
wow imagine chatgpt99 it's gonna have like 1 gazillion billion parameters omg
>unprecedented developments in technology lead to levels of prosperity previously thought unreachable
>however, as the number of human hands needed shrinks, increasing numbers of people go unemployed
>furthermore, society's stance continues to be no leeches allowed; people have to work period. unfrotunately, there seem to be no jobs available for people displaced by AI
How would you address this issue Mr. AI man?
Controlled depopulation.
Welcome to the endgame
gayest shit ive read in awhile, did a a pajeet make this?
>1. … likely…
>2. … i estimate…
>3. … i estimate…
>4. … i estimate…
>5. … possibly…
stopped reading right there
They've already gimped it and people are leaving the service due to the worsening of the answers
But can it accurately translate my Japanese porn games?
Actually, it probably won't have much use as a translator of fiction because it's going to be blocked from outputting no-no thoughts.
>2030
>chatgpt megapowerful and supersmart
>cool, can i ask it some cooking/plumbing/gardening questions
>sure goy, for only $500 per sentence
100% i can guarantee will be like this
Where do you think this LLM shit is going to be in 10 years?
In my sexbot
Mass unemployment and the beginning of post-scarcity.