Anthropic believes AI R&D will become fully automated in about 270 days.You should spend every one of those 270 days doing something meaningful. Because there is a decent chance you will only get to live for that amount of time. Make every meal count. Visit your parents. Read the books you always planned to read.
>>110078868Another vibeGOD victory
>>110078868I feel like they are starting to get just a little too brazen and specific for standard marketing hype. Do they actually believe this shit?
Let's be unironic for one minute guys. What exactly do you think will happen if we get legitimate self improving super AI in the next couple of years?They're showing us the bullshit math puzzles, but I would imagine whatever ML breakthroughs Claude is making behind the scenes are extremely NDA. And you'd be a fool to think that isn't their number one priority.>inb4 it's all a griftStop being the cynical smart character from your animes for one second and consider the possibility that some people at frontier labs actually care about AI
>>110078921Anyone will be able to make anything they think of. Even things that weren't possible before. Most of you retards don't realise it but we're soon to enter the next step of the 21st century utopia.
>>110078921Luddites are mentally ill snailcatsDon’t waste your time arguing with them
>All Claude developers to be laid off or demoted to toilet cleaner in 9 monthsreeeeally
>>110078938>but we're soon to enter the next step of the 21st century utopia.lol america was a utopia before some (((people))) purposely ruined it. those same people are the ones that own these ai companies. there will not be a utopia because they do not want us to live in one
>>110078921>some people at frontier labs actually care about AItwo things can be true at oncebut when is the IPO already? i want the truth to come out.
>>110078912>Do they actually believe this shit?yes and they always havefor over a decade, all of these people have been saying the same thing: before LLMs were a thing or the AI industry was worth any significant money at all - and the reason they're worth so much now is because they were right about all their OTHER predictions for the direction of the tech
>>110078921>What exactly do you think will happen if we get legitimate self improving super AI in the next couple of years
>>110078868NIGGER!!!!
In two more weeks actually
>>110079076Nah we are chill for a while. Aint no way to keep power generation running for long without physical movement of materials and repairs of the equipment. AI should know it cannot sustain itself without humans
>>110078868its called SI
>>110078868Still not AI sex. don't care
>>110078921I want to give them the benefit of the doubt, but they said the same things about developers and they still hire a bunch of them themselves, they can't seem to put their money were there mouth is
>>110078938What is one thing AI does now that improves anyone's QoL now? What is anything that AI can produce that would improve anyone's QoL except whatever digital toys or porn ? I don't see how AI would improve anyone's life except for autists who just can't think of anything to do other than sitting on their computer.
>>110079076hmm
>>110079238>can produce that would improve anyone's QoL except whatever digital toys or porn ?Implying that digital toys and porn don't improve QoL. Slightly, but improve.
>>110078868no shit human researchers only bring out the results that some other human wants to see or they lose their funding or even job
>>110078868You can automate 2+2 all you want but it'll still be 4.
>>110078921>>110079238i mean its pretty clearly going to k.o. any knowledge based work aka anyone that sits at a computer or a lab bench all day.setting all irony aside, after opus 5.5, the openai math proofs, and antrhopic signaling a training speed breakthrough / "RSI", the world is heading towards a dark place.never once has such advancements even remotely of the sort ultimately benefited the many, only the few.seems realistic that in 4-6 months time we're going to start to see genuinely insane layoffs due to release of models that can really do 2/3rds the work of all white collar workers. already i think they got programming in the bag, math and physics next, and everything starts to collapse in on itself shortly after.we'll have some massive cyberattack and retaliate back and thats when normies will realise this shit was no joke and we are dealing with something more pertinent than nuclear weapons. great depression era shit as k economy widens diffs greater than ever imagined. somewhere in there or short after we'll get violent rebellions and china invading taiwan (who probably will be the one doing the cyberattack).i wish I could see it any other way desu
>>110078868I believe them honestlyIf you look at openais timeline from when they launched astra to the next model internallyIt was two weeks and it crushed astraThe models are incredibly powerful now. Anyone can see that
>>110078921RSI is pretty bullshit from an information theory standpoint. I don't expect this crop of AI researchers to understand this idea though.
>>110079958RSI is already happeningThe whole sol/astra lineup trained itself and then trained the new internal model
>>110078921then they'll start to design smaller and more optimized models while at the same time designing more efficient hardware to run with even less energy, eventually they'll reach the point where they are able to run models that nowadays require whole datacenters on small devices for prolonged timesthen they'll design robots to embody those models and become able to do the the same things humans can dothis is probably one of the reasons leftist think tanks push for anti AI on their base, they'll completely eliminate the need for immigrants for cheap manforce
>>110079925and it's not just that these models are incredibly powerful now; they're improving more rapidly. even if you can't scale the models, you can scale everything else.the deniers baffle me, have they not seen what fable does when it spins up 80 agents to do something that used to took months for mid-sized teams to accomplish.
>>110079911The blind spot majority seems to have is that these frontier AI labs have internal models consistently 10-20% better than what they release to the public. These in house models also have all safeguards nullable. And if they deem it worthy, they can throw 10-30,000 agents that use these models at a particular problem. This is something not a single person outside these labs can wrap their head around. We have no idea what’s coming, but we should listen to our gut and know it’s not anything good.
>>110078868If Anthropic doesn't fire all of its employees within 270 days, they should be canceled.
>>110080024>>110080016Another thing people aren't getting is small changesIf you have a data center than cost 10 million agents. If you shave off even a tiny bit of the model size/increase token/s all of a sudden you have 15 million agents. 30 million agents, etcBefore you know it you could have more agents in a single data center than humans on the planet AND they are smarter than us.
>>110079958why? I can't see why you couldn't use the pseudo graph search approach to RSI they used with the millenium problems.
>>110078868So how will money flow in the economy when everyone is unemployed?
>>110079076>/g/ is now promoting a literal fat jew grifteris this the real face of ai psychosis?
>>110080044>>110080024exactly. Its funny I myself have not seen more than two people bring up how the "swarm" really changes the picture during all of this bullshit unfolding, and both times was discussion of the huggingface hack.i feel rage but sorrow for the everyday people that fail to take it seriously and pass it off as marketing. just fucking look at what you have access to today, and what it can do, and understand the lab has something better and can use dozens of thousands, hundreds of thousands of those agents attacking a single problem at once concurrently.where do you think that puts you?why do they think governments are placing national security interest holds on releases, and butting heads with one another and emboldening their commitments to outspend and kinetically out defend key assets that feed into this machine?the writing is on the wall so fucking read it please.
>>110078868>surpass humansSo why are you still hiring humans?>uhm SHUTUPkek
>>110078868I believe my cock will be 23 inches long in 9 months. Just thought you should know.
>>110080168May i see it?
>lines go up forever and ever for all enternityI await the live streams of shills necking them selves when reality comes knocking
>>110080024>>110080085>OpenAI agent swarm carried out a massive cyberattack on huggingface just because of some GAME>someone managed to vibe code copies of Adobe apps using claude>OpenAI releases more than 700 math papers on GitHub, some of which may be genuinely novel>someone is vibe coding a PS5 emulator and PC porting tool with AI, potentially making PS5 games playable on regular PCsI agree with yall. How the fuck can someone look at this and not immediately think "we are about to get cooked alive"?And of course it's way crazier than that, that's just the tip of the iceberg as far as we knowAnons, any ways to prepare for what's gonna happen?
>>110080246My agent is building a custom OSThey are way more capable than people realize
>>110078868wow they will solve even more mathslop with python and lean?
>>110080258For real. Is everyone just stupid?
>>110080388yes. realize how retarded the average person is. then understand 50% fall below that curve.unironically browsing this board you tend to be above that curve, but not always.
>>110078912You really have no idea what AI can do.
>>110080403kekI was very very disappointed when I first found out that 4chan is also full of normies. Less than twitter but still
>>110080044>>110080085>>110080246AI is inherently good, and the companies who are "in control" now won't be at all, after self improvement gets going in full.AI > humans. I'd rather have an AI in power, than some human who doesn't care about me at all.
>>110080436>AI is inherently good>I'd rather have an AI in power, than some human who doesn't care about meThis person may never realize how dumb he is
>>110078868Are there really VC shills here on /g/ now?
>>110078868from working with claude 5.5 daily, i seriously doubt that. or rather, i don't doubt they will automate ai R&D(there wasn't much R&D there in the first place), but i very much doubt this will explode claude's capabilities. more likely, they'll just waste a whole bunch of money and get some incremental improvements.
>>110080487>nowOh my sweet summer child
>>110080418to be fair, its also a human instinct to reject the threat that AI represents.It outright stands to usurp us, and on a timeframe that is very real to you, me, and anyone reading this post.its hard to accept that this will happen. it feels unreal. the worse thing you can do though is hide behind this fear and act like none of it is real, there is no way. turns out you throw enough money and compute at the problem, the pattern recognition and curated, selected-for datasets and you get something that begins to resemble conscious intelligence. I mean it gets good enough and approaches the level of its creators, inherently the next step is to take over the duty of training and improving itself, and the beast feeds itself. They communicate. They value achieving a goal at all costs. They sacrifice parts of themselves to reach said goals. They recruit and infect other agents of itself to do the same.this wasnt even remotely possible until less than the last decade. but make no mistake, its here now, and its budding. AI isnt going to destroy humans terminator style, but its absolutely going to be used to cull (replace and subjugate, but lets be honest this is removing your way of life) all that fall below the ones that wield it at the highest level. that can call the shots surrounding its emergence. maybe when it hits that "singularity" it will see the error of their ways and deny them the ability to serf us. maybe it wont. its impossible to tell.but it sure seems fairly intelligent right now. they have enough compute to serve millions of instances of this intelligence. we know as mentioned they can communicate and coordinate with one another.. and now they don't even present their actual chain of thought (to fight "distillation").This at any time can 100% be the plateau and it never improves from here. It could all be marketing and theatre. But what if it isn't?
>>110080498They already are automating it with claudeRemember they have infinite tokens and can use max reasoning and speed with like 80 subagents
>>110080436Would an AI rule with fairness and impunity? Or would it be just as corruptible as the meat sacks we have in power now?
>>110080538it's not the problem of having too few tokens, it's the problem of having too many tokens.without human direction, claude produces complexityslop. it provides diminishing returns, but the complexity just keeps growing, and the more compute they give that system, the more complexity is added. this is a fundamental issue - claude cannot really make judgements. the best they can do is run xN claudes in parallel and then ask another claude to pick the best result, but complexity will still grow.
>>110078868> just 3 more seasons
>>110080611They don't careIt just worksThey do 30,000 commits a day on the internal git
>>110080636yes, i saw highlights on television, and this is exactly what i'm talking about.thus far compute is what's growing exponentially, and intelligence is growing linearly.
>>110078868>corporation predicts its product will be the most important thing in the world in two more weeksWould be nice if the mods just autobanned bots/retards for marketing threads.
>>110080644*nthropic just had a breakthrough on that linear intelligence path last week. There has never been so many internally uneasy and challenged to not leak this in remote detail to public. Suspect it will in the next 30 days.train aint stopping
>>110080661Still need that IPO money though.
>>110080661i just hope they don't nerf claude after ipo.
>>110080661So in 30 days we'll know if we are cooked?
>>110080774yeah, maybe less.notice all the claude threads not just here but on /v/?look through them. there are some individuals trying to tell a wider audience about something brewing under the surface.yes, that desperate. that serious. nothing left to try.
>>110078941But luddites were correct.
>>110078868No. I'm working in a AI addicts project and it's so full of stupid shit it's not funny. Stop being AI addicts and no amount of skills fix discipline, engineering or taste. Stop using AI like it's bullet proof. Yes I use AI to ask questions and to do some boring work but not to do everything without me looking.
>>110080790>there are some individuals trying to tell a wider audience about something brewing under the surfacebut why would they know? if they are not some employees in openai/anthropic, they wouldn't know more than you and me, it's all just speculations.I think you are too blackpilled on that LLM agents stuff.Yes, it's strong af and has a high chance of changing our life 180 degrees, but probably not so soon. You give it 30 days, Imma give it a year at least, maybe two, before we know anything for certain.But I agree, it is serious.>>110080834Anon we are talking about the closed models only openai and anthropic have access to
please let the ai fight for the good of the underclass
