To anyone saying AI is uselss
>>108767066How many times are you going to ask how to do the same thing over and over?
>>108767143what you gonna do about it, white boy?
it was already dying
>>108767066This is need and all but this doesn't change the fact the bubble will likely pop. Saas will not disappear into Oblivion the minute it happens like dumb fuck tictokers say, but it will get a fuck ton. More expensive. Big players like anthropic and open AI are already hinting at turning to token based billing as opposed to the subscription model, which is heavily subsidized as of right now. If you can run decent models locally then you're mostly fine. It's the API pics that need to be worried
>>108767066Talking to an AI is more pleasant and productive than people. Especially most of the people on stack overflow.>Ask question>Lucky enough it didn't get closed/deleted etc>Get useless answer from jeet>Get person answering a question they wish I had asked and not what I actually asked>Asked to justify use case for my questionFuck people. Useless bastards.
>>108767066>atwood's face after he sold it in june, 2021 for $1.8 billion to some retard investment group
>>108768151>>108767066this is fucking hilarious, that was literally the absolute top. i mean part of it is also the new management's incompetence but it'd have dunked under OG founders too after gpt3
It is not useless. It's useful for making a Google search since it is effectively a search indexer.It's useless for doing an actual task - useless as an "agent". I have watched the different AI plays Pokemon streams on Twitch since they launched and they have barely improved.
>>108768054the big boys are already tightening. i have a student github copilot sub i get for free, and i used to get a limited amount of premium model tokens, and if they ran out i still had access to older models for free. then they limited the premium models so i got less token.. now they have limited the older models too, so if i go over a weekly allowance i get no access to any model at all. yeah i'm thinking it's aiover
Same thing is happening to Wikipedia. People have donation fatigue and most of the admins retired. There's only about 50 active admins still doing 99% of the moderation now.
Claude Opus 4.7 still plays Pokemon like a retard. I bet it still won't beat it. Which hints that investors are wrong about this leading to mass replacement of white collar labor.
>>108768189who cares about pokemon
>>108768171Even the Gemini app is starting to show signs of that. I have the normal planned for Gemini and I used to be able to use Pro all the time but now I'm getting those " usage of this model is restricted until x:xx PM/AM" messages (The same one anthropic models are infamous for showing users). If even Google has to start restricting usage of certain models you know they're fucked.
>>108768187yeah that's great and all, but have you donated yet?
>>1087681988 year olds, of which can beat Pokemon.
>>108768198>>108768208It's a good benchmark for an agent. The game is designed so that you can only make progress even if you play like utter dogshit.
>>108768187Wikipedia has enough money, just because they do regular donation drives doesn't mean they're cash-strapped. All the AI companies will eventually have to fund Wikipedia and sites like Reddit and 4chan because otherwise the models will go out of date if don't have information to scrape.
>>108768198I think an AI that can't beat Pokemon efficiently cannot possibly become a super-scientist, or do anything that generates enough profit to overcome the billions sunk into this.
>>108768151they made more than 1.8 billion selling their data to AI retards
>>108768274source? all the big models just scraped it before people even started protecting. and chink etc models still just scrape it because it's virtually impossible to prove it was scraped FROM their website
>>108768237>>108768222May I ask you guys why you expect these tools to be good at anything and everything out of the Sun? That's not going to ever happen for obvious reasons unless we figure out how to replicate consciousness. No one even knows how the fuck that works. I'm not trying to make excuses for trash models but in every single thread related to AI people keep bitching and moaning as if they expect these things to be "le heckin super intelligence AGI"
>>108767066>>108767181
>>108767066this graph doesn't prove its usefulness
>>108768335>being advertised as a "cognitive enhancer" that can program code and shit>Windows being turned into an "agentic OS">AI bots replacing customer support>expect it to work as advertised>WOW let's not get ahead of ourselves here
What little use Stackoverflow has is gone. It loads very slowly now. Many of its pages are also deindexed. Most of the answers are from 10 years ago and outdated.
>my name is Ozymandias, king of kings...
>>108768391>Dimwit fell for obvious grifting
yeah bro they are totally asking ZERO questions on SO
>>108767066If the average stackoverflow questions are being replaced by retard LLMs, then stackoverflow was full of retards.LLMs aren't going to replace actual research and understanding. It's going to be used for grinding STEM related stuff and replacing code monkies..
>>108768712Not zero, but there was only ~3100 questions in the past month.
>>108768237>I think an AI that can't beat Pokemon efficiently cannot possibly become a super-scientistIt's actually significantly easier for an AI to become a prolific scientist than it is for an AI to beat Pokemon.There are mountains and mountains of written words explaining how to become a prolific scientist including all the intellectual tools and common goals they are reaching towards and the artifacts of their work are all written word. Pokemon is a real time simulation that has no concrete goals and comparatively little written about it compared to the entire body of scientific work. There is no artifact produced whatsoever from beating Pokemon.I don't think LLMs will ever come up with ground breaking innovation, but 99.99999% of the scientific discipline is incremental paperwork validating the things that 1 in a trillion geniuses come up with.
>>108768335>May I ask you guys why you expect these tools to be good at anything and everything out of the Sun?nigga because that's how these tools are being fucking sold
>>108768780Grow up kid, you knew fuck. Well. These were not going to live up to the high. They are good at specialized tasks. Regular ass PEOPLE are not good at being amazing at everything, so why the fuck did you expect a static, somewhat deterministic neural network to be the same? Let me guess you got fleeced by the nft craze too?
>>108768774>incremental paperwork validating the things that 1 in a trillion geniuses come up withIf only. AI can't validate anything empirical. It might be able to validate some trivial mathematical theorems. However most "science" is actually reshuffling existing work to generate new slop papers while making sure you add all the important citations and buzzwords. That's something AI can sort of do.
>>108768789>so why the fuck did you expect a static, somewhat deterministic neural network to be the same?Because that's how you retarded shills keep selling them. Maybe sell your product with this vapid honesty and see how much investors are willing to pour into "specialized tasks" as opposed to fully automating the knowledge workforce.>Let me guess you got fleeced by the nft craze too?You aren't very good at reading comprehension are you?
Website full of annoying cunts telling you not to do what you wanted to do
>>108768821> Obvious shill said xyz therefore, I need to take everything they say at face value.
>>108768904>it doesn't matter if the sales pitch for the product is complete bullshit! just slop slop slop it up goy!
>>108768162>this is fucking hilarious, that was literally the absolute top.You're putting the cart before the horse. The usership dropped because they bought it out, not after they bought it out; most M&A's destroy value since the investment needs to be returned on, which usually means aggressive cost-cutting and short-term gains.
>>108768961Just use the right tool and you'll be fine. What the fuck exactly is the issue you're having? Are you getting filtered?
>>108767066well yeah 99% of the questions are homework, and now they just ask chatgpt so they can look like a retard in private.
>>108768774Actually, no, I don't think the knowledge you need to create scientific breakthroughs is encoded in scientific papers. At minimum, you need to interact with reality directly for that.
>>108767151who said he was white?
>>108768222its not a good benchmark at allit plays pokemon by trying to look at gamefreaks shitty graphicsno duh it's gonna suck at it
>>108768198/v/ diptwits
