/lit/ - Literature

>>19394975
Haven't seen a single AI passage yet that isn't schizophrenic raving.

>>19394975
>Would you still write knowing that an AI can generate something better than you're capable of in less than a minute, even if you dedicated your whole life to writing?

Writing is a retarded waste of time even without AI because literature is dead and culturally irrelevant. This is only a problem for art forms that actually matter.

>>19394975
>Writing is a retarded waste of time even without AI because literature is dead and culturally irrelevant. This is only a problem for art forms that actually matter.
Der ewige Zoomer

>>19394975
Or maybe it's so shitty they don't dare to show it and embarrass themselves.

>>19395142
>art forms that actually matter
so currently none except maybe cinema

>>19394975
Writing is meaningful because of intent. It responds to a prior statement in a conversation spanning history. AI is fundamentally incapable of intent, because it’s a kind of predictive modeling — it spits out the most likely word, based on the preceding 5,000. This is the opposite of conversation; it’s blind agreement with the input.

>>19394975
>Have we reached the point where AI has defeated humanity in term of art production?
Depends on how good they are at building their models. They're programmers - so it might require more time.
>What would the future of human artists be?
I actually shudder thinking about this.
>Would you still write knowing that an AI can generate something better than you're capable of in less than a minute, even if you dedicated your whole life to writing?
Probably - but I'd do what I do now, which is write it and then just never try to publish. I think the really interesting bit isn't art per se, but will AI surpass us in philosophy?

>>19395165
No.
https://openai.com/blog/dall-e/
When this goes public, graphic design as a field is obsolete.
GPT-3 was already good enough to mass produce genre fiction, with humans only needed to guide and edit. It is impossible to know for sure what GPT-4 can do as of now, but given the massive improvements GPT has shown between each increment, it may very well end up replacing fiction writers entirely. If not, GPT-5 certainly will.

>>19396381
Somebody already tried.
https://philosopherai.com/
It's paid now, but its answers always were along the lines of a teenager skimming Wikipedia. Always wrong, but it might fool somebody who didn't know anything about the subject.

>>19394975
>not making the model public
it's a marketing play and it's working

>>19396411
I mean rate of improvement is 2^N man. This is the first 5 years of large scale ML.

>>19394975
>Would you still write knowing that an AI can generate something better than you're capable of in less than a minute, even if you dedicated your whole life to writing?
I still play chess even though a computer will beat me every time. Some things are worth doing for their own sake.

>>19396530
The problem is that the resources needed for the next AI get far more difficult with each generation. GPT-2's resource usage was nothing compared to GPT-3, and GPT-4's power consumption is said to be staggering. A workable form of GPT-3 took just a year after its precursor, GPT-4 is predicted to take several years.
Another problem is that these AI comapnies have full-time employees who work around the clock to prevent the AIs from saying offensive things. It's possible the next generations of AIs will be incapable of saying uncomfortable truths.

The AI can think but........CAN IT FEEL

People still buy handmade vases and the like even though a factory could mass produce vases that are better by empirical standards. I guess future human art's interest would be derived from the mythos around its creator. We live in a world where every piece of fiction has thousands of pieces of fanfiction (some might even surpass the original writing in quality), but people still seek out pieces stamped CANON by somebody.

>>19395085
fair

the same guy who's a big investor in openai just blogged about investing in cold fusion or some other fanciful boondoggle and he dropped out of his undergrad so he's an easy mark. i'll believe openai is real when it's "open" to the public.

>>19396785
damn that actually kind of decent

>>19396594
Can YOU feel?

>>19394975
>OpenAI refusing to allow public use of the full DALL-E (automatic picture generator) and GPT-4 (automatic text generation)
So what? Have you heard about ruDALL-E or GPT-J or other similar projects?
>>19396575
OpenAI said GPT-4 won't be much bigger. Better architectures and training methods were developed since GPT-3.

Sing, OpenAI, the anger of Peleus’ son Achilleus

Honestly though that avocado chair is amazing. To think I could just imagine something, say a few words, and that thing could spew out a dozen prototype designs for me.

Or imagine a netflix of the future where you sit down in front of your viddy wall and say, Netflix, tell me a story about a man and his wife's boyfriend, and Netflix would deliver a whole series on the theme, employing the faces of actors and people of the past like puppets. Or, a future porn site, PornAI, show me my neighbour, Katie, getting railed by the binman with the big belly and the four cats from Mrs Hobbs' home, and then I see it all at once, Katie, and the binman with the big belly, and the four cats, orgying just for me.

>>19395142
>This is only a problem for art forms that actually matter.
porn is the only culturally relevant art form of our age, and AI will improve it immensely. I unironically look forward to this.