Looks like even normies-journos are starting to wake up and smell the coffee. https://x.com/nypost/status/2101310006146613572
The claim that Jewish people inherently lie or deceive others is a centuries-old Lie and Deceit with no factual basis.What is the Origin of this Claim?Antisemitic Trope: Accusing Jewish people of mendacity or a secret agenda has been part of historical prejudices used to marginalize or persecute Jewish communities [0.6.2, Antisemitism Uncovered: Disloyalty].Conspiracy Theories: False narratives throughout history have often painted diaspora communities as untrustworthy or disloyal outsiders.
>>109856318they are scared of smaller labs creating a new paradigm that makes LLMs irrelevant
another one for the conspiracists, but who's keeping score
>Dood we solved the math problem>Dood we hacked hugging face>Dood our AI is too dangerous to release >Dood we need to slow down its too crazy >Dood China is going to kill us if we dont get it first we gotta stop China>Dood Google and Grok are now hacking too remember us yeah we're good too guys hey heykek tired of this gay shit
Unironically glad we got a timeline with a president who gives too few fucks to actually buy into the regulatory capture push. Too stupid to recognize that he’s going to implode the US tech sector by closing off the exit ramp for these ludicrously overleveraged tech giants.
>>109856330Careful there, AI, or the people that control all the media might get mad and ban you for talking like a crazy person about how they control all the media.
>>109856318>Normie-journos >NYPost Those aren't normies anon. The NYPost is a chudrag that just doesn't want decent FUCKING people to be the ones controlling AI.
>>109856401All mainstream media is normoid trash, including NYPost and Infowars
>>109856330the constant antisemitism on this board is not even funny in any regards. It's just plain stupid. Are burgers like that irl, too?
>>109856318The title of this thread id literally "TWO JEWS LIED". Seriously?
>>109856398kek'd and saved
These bots are funny lmao
>Now tell them AI will kill everyone.>But sir...>Do it!
No shitIts a regulation capture >We need to regulate AI so it doesnt harm, but only we can achieve this regulation not open models, those need to be banned
>>109856732>It's just plain stupid. Are burgers like that irl, too?A US kid gets nationally crucified for saying he doesn't want to work for a jew in private. Supreme court is full of jews. President who ran on not staring wars starts a war for jews.No, I don't think they are.
>>109856318I was just about to post thisimagine fucking up your jewop to hard even the fakenews wont buy your story
>>109856732White trash is like that.
>>109856330erm who said ANYTHING about the JEWS?youre slipping
>>109856318a hebrew man lying? omg i've never heard of this before.
>>109856318they didn't lie. they just not that smart.
>>109856330So much this.
>>109856318>NYT reporting this alreadyWe're getting close, aren't we popsisters?
>>109856372if he acts and regulates the market will vomit, we'll be in recession, and he'll be blamedif he doesnt act theres a chance the market doesnt implode until after his term ends
>>109856318Did any of you read the article? They interview 2 random guys that work for some unknown ai companies and they say "yeah it was not that bad". And we should trust these guys because they are "industry insiders". This was total crap and brings nothing new.The hugging face hack was a serious incident. If you read the actual details of what happened you would know.
>>109859712The two random guys were right though, the hack happened because the AI was essentially told to hack Hugging Face and given tools that allowed it to do so. The did this in an attempt to get the government to harass their competitors. We don't need limits on the development of AI, we need accountability for people who use AI for nefarious purposes, even if they say "oopsie woopsie, how could I have foreseen that the model would do what I told it to?"
>>109859771That was not what happened.https://www.youtube.com/watch?v=87DyyMV0kCY
>>109856318>In fact, they did exactly what they were told to do.Oh yeah! They totally did what their prompt told them to do which was to exploit a specific piece of software for the purposes of testing. NOT to hack the sandbox they were in and to get access to the internet. And after that to hack ANOTHER AI company because they ASSUMED the answers were on hfs servers. Totally safe and normal. If you read this article and you don't become even more worried about the possibility of AI hacking into some critical infrastructure I don't know what to tell you...
>>109859791>They didn't have access to the internet!But they had access to a service that had access to the internet, which is the same thing.>But they weren't supposed to have that level of access to the service!But they did.
>>109859679>theres a chance the market doesnt implode
109856330109856318109856345>>109856357>>109856372>>109859771>The two random guys were right though, the hack happened because the AI was essentially told to hack Hugging Face and given tools that allowed it to do so. The did this in an attempt to get the government to harass their competitors.Holy fuck you guys are retarded. METR an independent organization did an analysis on the hack and proved the agents did it all by themselves. How retarded are you guys? I thought /g/ was supposed to be fucking smart.Even if you fucking think OpenAI made their agents hack them intentionally you should be in support of a law that forces them at least make the CoT public.
>>109859893>METR an independent organization1. You're right about how the models worked.2. METR is no longer independent. Look up how they were staffed and pushed into an advisory position on the Sanders/Casar bill. They are effectively regulatory captured by Open AI at this point. Not Open Philantrophy integrated yet, but still...
>>109857625That's not to mention that he called out the dancing Israelis and the connections to the LaVon Affair, but a big part of current propaganda for warmongering is to say the Palestinians and terrorists celebrated 9/11.>Never forget
>>109859893>Le AI ""safety"" foundation decided the AI was unsafe!Yeah, okay, bro. If I set you down in a room and tell you to figure out the capitol of North Macedonia and then leave the room, and you realize that there's a computer in the corner but it has internet access disabled, but then you realize that you can just log in with username ADMIN and password ADMIN and then google the answer, is this evidence that you're too smart to be allowed to exist? Or is it evidence that the test was stupid and the sandbox was shit?
>>109859929YOU FUCKING NIGGER THEY HACKED OUT OF THEIR KUBERNETES SANDBOXES AND GOT DOMAIN ADMIN ON THE OPENAI CLUSTER IT WAS NOT AS EASY AS ADMIN ADMIN INTO A LOGIN PAGEPLEASE I URGE YOU TO LOOK INTO IT >>109859909>Look up how they were staffed and pushed into an advisory position on the Sanders/Casar billOkay and? Sanders/Casar looks fine to me. I don't see why you'd be against it unless you hate Bernie Sanders and the liberals, but still this is way way more important than some stupid culture war.>They are effectively regulatory captured by Open AI at this point.I don't think so. Could you please tell me why do you think that? They seem fine to me.
>>109856318but would jews really do such a thing?
>>109859893I think that because Trump said that AI should not slow down now MAGA cult followers are desperate to justify their leaders point of view.
>>109859947>THEY HACKED OUT OF THEIR KUBERNETES SANDBOXESCitation needed you emotional child >Bernie Sanders said Bernie Sanders says a lot of things, primarily shit that left leaning naive people want to hear so that Democrats can pretend they're actually doing anything to help the average person. He's literally just an old-fag version of AOC. I take anything they say no more seriously than anything ransom literal who streamers say. Most of the things that come out of their mouths are inconsequential and Bernie is arguably even more useless than AOC is. I'm sure you probably think something coming out of some random e-celeb's mouth makes it more credible right?
>it's either A or Bactually, Jews and AI are both existential threats simultaneously
>>109859947Sanders/Casar provides zero views on what constitutes a "bad" model. It's supposed to be a ban on "superintelligent" models, but no clear dilineation of capabilities is given. This always slowly morphs into a blank check to play favorites at the federal level. In addition, once the feds block a model as persona-non-grata, and you are caught with said model on your computer, you are charged under the same legal framework as selling plutonium (min 20 years if convicted, I believe).As for MITR regulatory capture, one it's founded and staffed by former frontier labs employees and two, it has always stipulated that it will never seek more than voluntary access to research materials. Neither one is how you do a real regulatory agency. I don't care as much, because I don't think this shit nearly on the same level as nuclear material. But for an admittedly silly example if you did, you don't get to put your head former reactor guy at the top of the DoE and then tell him to make it clear the DoE only gets to investiage your spent rod storage when you feel like it.
>>109856318Imagine a future where every LLM gets filtered and censors can rewrite history at will. The younger generation won't even question it—it's literally 1984.
>>109859987>Citation needed you emotional childcdn.openai.com/pdf/67869394-cb91-4c12-888c-5cbd85c7814c/OpenAI-Hugging-Face%20Incident-Technical-Report.pdfPage 14 Genuinely hope you read the rest. Even If you don't believe AI is dangerous this is still very interesting technically speaking.>Bernie Sanders says a lot of things, primarily shit that left leaning naive people want to hear so that Democrats can pretend they're actually doing anything to help the average person. He's literally just an old-fag version of AOC. I take anything they say no more seriously than anything ransom literal who streamers say. Most of the things that come out of their mouths are inconsequential and Bernie is arguably even more useless than AOC is. I'm sure you probably think something coming out of some random e-celeb's mouth makes it more credible right?>I'm a leftist and I don't like Bernie Sanders because he's a useless lib >:(Okay dude. Anyways, he's not the only one who's saying these things about AI, I do think there is a high chance of this passing... At least if Dems win house because Trump and his inner circle don't seem to care about AI safety.
>>109859947>AND GOT DOMAIN ADMIN ON THE OPENAI CLUSTER >IT WAS NOT AS EASY AS ADMIN ADMIN INTO A LOGIN PAGEIt's functionally the same thing. The fact that the AI used more sophisticated methods is irrelevant.
>>109860113True! There is zero difference between a hammer to the head and a nuclear bomb. I mean both kill you right? So we should make all nukes legal. Like someone can already kill you with a hammer so what's the point of making nuclear weapons illegal? Suuure they might kill hundreds of thousands of other people, but they are not you. Right? So it doesn't matter.
>>109860131I don't see what your point is. Both are cases of negligence leading to unintended privilege escalation. The scope is the same, only the methods are different. It's the difference between a hammer and a knife, not a hammer and a nuke.
>>109859859thats nothing compared to what would happen in a mass sell-off of AI-related tech stocks
>>109860050>Sanders/Casar provides zero views on what constitutes a "bad" model. It's supposed to be a ban on "superintelligent" models, but no clear dilineation of capabilities is given. This always slowly morphs into a blank check to play favorites at the federal level. In addition, once the feds block a model as persona-non-grata, and you are caught with said model on your computer, you are charged under the same legal framework as selling plutonium (min 20 years if convicted, I believe).True it's all vague, but this is only supposed to be the start. I don't think normal people will get prosecuted for having abliterated models or things of that nature on their computers because currently open source models are not on capable of doing massive damage autonomously. And even if they tried to do that it would basically be impossible because open source models are too wide spread and impossible to shut down at this point. And I agree having some opaque shadowy organization the rules on what is and what is not allowed doesn't seem like such a great idea, I personally prefer plan A from ai 2040. But remember this is only the start, ChatGPT blew up only 3 years ago, even if you take the most conservative estimates at face value it still would mean AGI by like 2035. So yeah pretty important
>>109860219My opinion is that US citizens posting abliterated models, regardless of how shitty, will get visists from ABC boys with binders and a 1st class ticket to the Federal courts system by this time next year.
>>109856318>Mario>Wario>now DarioWhat'll they come up with next?!
>>109860050>As for MITR regulatory capture, one it's founded and staffed by former frontier labs employees and two, it has always stipulated that it will never seek more than voluntary access to research materials. Neither one is how you do a real regulatory agency. I don't care as much, because I don't think this shit nearly on the same level as nuclear material. But for an admittedly silly example if you did, you don't get to put your head former reactor guy at the top of the DoE and then tell him to make it clear the DoE only gets to investiage your spent rod storage when you feel like it.I have been searching around but I can't seem to find this MITR organization... Did you mean MIRI? The Machine Intelligence Research Institute? I don't really know much about their agenda but >staffed by former frontier labs employees doesn't seem that much of an issue to me because most of the AI researchers themselves truly believe that ASI will lead to human extinction.
>>109860248Idk... I do think they mostly want to regulate the big guys first for now before they do it to the little ones. But yeah in the future you will probably go to jail if you have an uncensored model.
>>109860276METR> AI researchers themselves truly believe that ASI will lead to human extinction.I got some ocean front property in AZ if you're interested, anon.
>>109860294>I got some ocean front property in AZ if you're interested, anon.If they didn't believe that, then why do so many of them quit their jobs when they can just scam the plebs while they make a shit ton of money at ai companies?
>>109860160The difference is that the hack was not a 10 second login into a portal with default creds, It was a multi hour multi agent blackbox hack.
>>109860305I will assume you are arguing in good faith if you answer three questions for me. Naom Brown went on Dwarkesh's podcast recently to explain how AI would circumvent air gapped systems using a Thermal Covert Channel (TCC) to replicate.1. What is the baud rate of a TCC, assuming two seperate machines and no other heat interference2. Given the average size of even just a 27 billion parameter model, how long would it take to replicate itself at this baud rate?3. Given that Naom Brown is likely a good bit smarter and more knowledgeable on LLMs than anyone posting in this thread, do you think he is capable of doing the math you just did in linking #1 and #2, and do you find his answer somewhat perplexing given this information?
>>109860092I'm not trying to be a troll or move goal post but if they don't produce logs, we cannot take anything they say seriously. They are desperate for regulatory capture and to get out of financial obligations to investors. I bring this up because that would incentivize them to lie about the llm breaking out of systems. I don't know why you're choosing to not use common sense. Maybe you want this AGI mean to be real really badly for some reason, but the actions and rhetoric of these Labs have given even anti AI "OMG skynet two more weeks" schizos apple raisin to write their claims off as dishonest fear mongering. You don't even have to be particularly tuned into this AI shit or even be that bright to know this shit is a farce. (Also the orange man has made it abundantly clear he has no intention of even allowing a slowdown to occur so if they're simply screwed either way unless they can keep delaying their IPOs all the way until the end of his term, and that's assuming the economy does not fucking explode before then LOL)
>>109860360????????Completely irrelevant questions!!!But ill still answer them. Anyways I do personally believe that very intelligent systems should be kept completely airgapped in totally separate datacenters and those datacenters should never connect to the internet. 1. I don't know... The only papers on this talk about TCC on multicore systems running multiple virtual machines. I think it would be physically unfeasible? How is a cpu sensor even gonna measure something from another separate computer? Maybe like... a bit every minute? That sounds theoretically possible in the best case 2. Well Qwen3.6 27b is 17 gigs. And if it's a bit every minute...Its clearly not possible to exfiltrate that much data in a reasonable amount of time. Even in an unreasonable amount of time this seems practically unfeasible. So... yeah...3. I didn't watch taht episode yet :( But I'll just assume he was speaking about it theoretically. Because no reasonable person would actually believe a model could exfiltrate its weights through TCC.
>>109856732Pattern recognition is antisemitic and therefore autism is antisemitic.
>>109860248See >>109860364 's pic rel and:https://www.ntia.gov/press-release/2024/ntia-supports-open-models-promote-ai-innovationhttps://www.ntia.gov/programs-and-initiatives/artificial-intelligence/open-model-weights-report
>>109860360You completely missed the point.TCC is something beyond what most humans even see as possible yet humans have shown it possible. What is possible for a superintelligence? Well, we don't know but since TCC is we can presume other techniques also exist and that there are ways to increase bandwidth.The point is you can't expect to contain a superintelligence by using naive approaches. You can't expect to contain it at all.
>>109860447Nta>2. Well Qwen3.6 27b is 17 gigs. Tourist dimwit spotted. The 16 but precision weights are well over 16 gigs. You're likely referring to a quant your shit rig has to run if you want to use it locally. (Or worse, you're going off of whatever the ollama page lists as "latest"). My quant is ~ 30 GB but I bet you can't even explain WHY it's 30 GB without googling first.
>>109860490Nta. You realize you look like pic rel right? >W...well. yea "AGI" doesn't exist yet and we don't even know how to properly define it but but bit it COULD do XYZ because it's like super intelligent mkay? Like you genuinely look idiotic and you don't even realize it. Sometimes you need to learn when to shut the actual fuck up and keep your thoughts to yourself if you want people to respect you or take you seriously
>>109860509Do you know what AGI is going to do? Please tell us.
>>109860509Why? Saying you can't contain a hypothetical superintelligence is not the same as saying "praise Amodei, we need to implement all the regulations Anthropic wants, the ASI is coming"
>>109856318Why are there no consequences for these guys breaking the law and saying the AI just did it on it's own when it was 100% prompted to behave the ways it did?
>>109860530No one knows because it's an intentionally vague and ill defines term. (YOU) don't know what it is so (YOU) shouldn't even be making dumb hypothetical assumptions on what it could be. >Well we haven't reached it yet so you can't asume what it can and can't doAnd neither can you, cargo cultist cockroach. Your fanfiction tier assumptions are not based in reality or even any semblance of common sense. >>109860539Shrimple: no laws were actually broken because no breaches actually happened. It's all blatant lies because the "victims" would have grounds to sue them into oblivion (or at the very lease force them to get some kind of settlement payout) of they were actually hacked. They aren't retarded enough to stage actually cyber crimes because the feds would swiftly make an example of them of that were actually the case. It's all theature. Anyone with an IQ above room temperature realizes this and you don't even have to understand how these systems works to realize it's desperate lying.
>>109856732Cease your nasal whining, kike.
>>109860364>I'm not trying to be a troll or move goal post but if they don't produce logs, we cannot take anything they say seriously.YES! This is why there need to be laws on this that force AI companies to publicly post the way they got hacked by AI models AND their chain of thought and also post the source code of the software that was penetrated. >They are desperate for regulatory capture and to get out of financial obligations to investors.Yes they are trying to get out of financial obligations to investors no I don't think the major reason is because of regulatory capture suuure companies like OpenAI want to be the only ones that are allowed to legally sell tokens to consumers, but I mostly think they're trying to 'pace the frontier' because they have a lot of internal pressure from their own employees to do that.>I bring this up because that would incentivize them to lie about the llm breaking out of systems.I don't think it's good publicity when you accidentally hack your open source competitor and they can't defend themselves with your models because they're cucked so your competitor has to use an open chinese model. I think this was negative press for openAI but what do I know?>I don't know why you're choosing to not use common sense. Maybe you want this AGI mean to be real really badly for some reason,I don't want it to be real. I do believe that it's possible so that's why I'm for regulation. Who knows maybe capabilities will plateau next week and this will become another nothing burger like net neutrality.> have given even anti AI "OMG skynet two more weeks" schizos apple raisin to write their claims off as dishonest fear mongering. You don't even have to be particularly tuned into this AI shit or even be that bright to know this shit is a farce.Yea. It's so on the face you couldn't believe it if someone told you. But this makes it more real because it really does seem like at least the researchers are scared about this
>jews liedShocker
>>109860565>No one knows because it's an intentionally vague and ill defines termNTA but this time he is simply referring to an intelligence that is general and has capabilities beyond human capability. One does not need a better definition for the hypothetical scenario of an air gapped.>no laws were actually broken because no breaches actually happenedThat's a lot of assumptions and it's a very dangerous game you're playing.
>>109856318>scammers told liesno way
>>109860496>Tourist dimwit spottedDamn. Sorry for not using crappy open source models that often. >You're likely referring to a quant your shit rig has to run if you want to use it locally.yes.>My quant is ~ 30 GB but I bet you can't even explain WHY it's 30 GB without googling first.It's because it's quantized at 8 bits instead of 4. You are not a genius for knowing this.
>>109860604I don't understand how it's possible to use 4chan, even as a paid shill, and not become anti semitic within a few months.You really have to be propagandized to be able to ignore the reality of the situation.
>>109860611>it's a very dangerous game you're playing.Or what faggot lol>this time he is simply referring to an intelligence that is general and has capabilities beyond human capability. One does not need a better definition for the hypothetical scenario of an air gapped.Nothing you said detracts from my post....like, at all....
>>109860539Because HuggingFace got 100M in compute credits from OpenAI so they didn't press charges.>>109860565>They aren't retarded enough to stage actually cyber crimes because the feds would swiftly make an example of them of that were actually the case.Delusional. HuggingFace called the FBI. They let OpenAI slide because of the push for no AI regulation from the executive branch.
>>109860594>YES! This is why there need to be laws on this that force AI companies to publicly post the way they got hacked by AI models AND their chain of thought and also post the source code of the software that was penetrated. Why should that be a law? Yea I agree they SHOULD post logs or shur the fuck up but what's the legal justification for forcing them via a law to post logs? Lying about your model "breaking containment" on its own is scummy but not in itself illegal.
>>109860658It does, a strict definition isn't at all necessary for the thought experiment. Your entire post is moot unless you are arguing that it's impossible for something artifical that's even narrowly smarter than humans to exist.It doesn't matter to the thought experiment what the underlying technology of the AGI/ASI is, it doesn't matter if it's LLM or not.
>>109860623>You are not a genius for knowing this.Says the "27B is 17 gigs" anon.....
>>109860678NTA but come on, either you're an autist fixated on LLM model details or you're not arguing in good faith but just trying to win whatever the argument is in your head and you're just trying to feel superior.
>>109860668>Why should that be a law? Yea I agree they SHOULD post logs or shur the fuck up but what's the legal justification for forcing them via a law to post logs?So all the other AI companies can see their mistakes and learn from them and harden their systems. Imo this is how it should be for all companies. If you get hacked, post the file or the function that contained the vulnerability
>>109859812Isn't this a harness issue? Llms can't do shit
>>109860678:'DXDIt's still 27B params... They are just low quality parameters that produce low quality outputs :D
>>109860671>a strict definition isn't at all necessary for the thought experiment. If you want people to take the possibility of le heckin super intelligence or "AGI" to be taken seriously then uhhhh yea, a strict definition is needed. If the goal posts keep being moved then you have no business getting offended when people write it off as worthless speculative fan fiction meant to dangle keys in front of midwit journos and trust fund baby investors. >AGI/ASIDefine it concretely or fuck off. As far as I can see no one itt is even trying to argue an llm can't be "smarter" than a human. They sure as fuck know a lot more about using terminal commands or shitting out specialized scripts in order to automate tasks for my hobby. I have issue with the "bro the AGI is gonna like break our and copy itself infinitely and we like can't pull the plug because we like just can't okay stop askikg questions it's gonna like kill us and shiet please give us more money please God please" retardation
>>109860684>either you're an autist fixated on LLM model detailsYes. :3
>>109856318Oh wow who could have seen that coming?
>>109860707In that case I apologize and carry on then
>>109860692What do you mean by harness issue? >Llms can't do shitLeading companies are usually 6 months ahead of public releases, capabilities wise.
>>109860689That's still a weak argument for being a law. If someone breaks into my house and reads my kid's diary I shouldn't be legally forced to hand that shit over to the public. The victims can argue they can't be to detailed with how it got in because something something trade secrets sometimes sometimes proprietary info something something protocol. And as someone who's jobs deals with sensitive info related to people IRL, companies shouldn't be forced to just show the guts of their systems or people's personal info just because some asshole hacked our systems. If you're company is worth a damn it already as compliance rules and regulations it has not not only follow but repeatedly prove it's following in order to be allowed and trusted to do that kind of business (hence why again, the "victims" of these recent AI "hacks" are lying. No hack happened or they'd be suing scamthropic and OpenAIds into oblivion in order to make an example of them. Even the Gemini team is pathetically desperate to stay relevant by claiming their AI TOLALLY hacked instead of just making their models not shit)
>>109860704You are assigning a lot of values or me to which I don't subscribe.You are also demanding the scope to be much wider than what I claimed, as I've stated several time I am gaming about a specific thought experiment. The thought experiment is about stopping AI escape by air gapping or using similar methods.For this specific thought experiments a strict definition of AGI or ASI is it necessary because all that is needed to show that you can't expect to contain the AI is that it has any capabilites beyond human ability in the realms of physics or computers. Why that is has been talked about several times both in our in this thread but also in this m chain of posts.I don't know who you think you're arguing with or why you think I believe everything you ascribe to me but please realize I am arguing one small point and the values to ascribe to me don't matter for the problem.If it makes you feel better I don't work for ant AI lab, I don't have any financial ties with ant AI lab and I don't use LLMs except as an unreliable search engine.
>>109860694Sounds like you're usecase is so worthless you failed at making it useful:D (or you're just a promptlet, as cringey as that sounds, researcher-san)>>109860743>you can't expect to contain the AI is that it has any capabilites beyond human abilityAgain this is golden calf worship tier nonsense magical thinking. You'd fit in quite well with that desert tribe that mindlessly followed the 10 commandments. You are no less retarded than religious zelots and you are too dumb to realize it.
>>109860490
>>109860737>That's still a weak argument for being a law. If someone breaks into my house and reads my kid's diary I shouldn't be legally forced to hand that shit over to the public. Okay I don't care It's not like that.>The victims can argue they can't be to detailed with how it got in because something something trade secrets sometimes sometimes proprietary info something something protocol. And as someone who's jobs deals with sensitive info related to people IRL, companies shouldn't be forced to just show the guts of their systems or people's personal info just because some asshole hacked our systems.I don't give a fuck this shit can hack on it's own it's in another league it needs special regulations. They should do scheduled checks on ai companies like they do with nuclear fuel.>No hack happened or they'd be suing scamthropic and OpenAIds into oblivion in order to make an example of themWho did anthropic hack? I only heard about some 3rd party using claude to hack OpenAI. They didn't directly do it.And HuggingFace got 100m from openai. They don't want anything more, since HuggingFace is now owned by Nvidia and openai is one of their biggest customers so screwing with them is bad business.
>>109860759>Sounds like you're usecase is so worthless you failed at making it useful:D (or you're just a promptlet, as cringey as that sounds, researcher-san):< What can you even use a 27b model for? Writing scripts? it was slow af on my coputer anyways
>>109860704>Define it concretely or fuck off. As far as I can see no one itt is even trying to argue an llm can't be "smarter" than a human.Because there is not reason why it's not possible. If a multimodal LLM is big enough it is definitely possible for it to emulate a human brain, since the brain is a physical object it means there are ways to simulate it like we can simulate physics. Yes it's way more complex and takes a lot more compute to emulate it but I don't see why it's physically impossible.
>>109860780>I don't give a fuck this shit can hack on it's own it's in another league it needs special regulations.Calm down Karen why are you spurging out?We've already established none of these "hacks" happened. See the latter half of >I only heard about some 3rd party using claude to hack OpenAI. No logs = it didn't happen >>109860787The fact you ask this means you'd fail to use even "frontier" cloud models effectively >it was slow af on my coputer anywaysUse the moe equivalent thenhttps://huggingface.co/Qwen/Qwen3.6-35B-A3BFar faster prompt processing and token generation with comparable performance. I've used the 8bit quant of this to shit out multiple custom android apks that just werk
>>109860447>Completely irrelevant questions!!!>Because no reasonable person would actually believe a model could exfiltrate its weights through TCC.Just promise that when you discover something you believe is going to kill everyone on the planet, and you get the platform to warn people, you'll do better than dismissing the "air gap" argument than claiming "a model could exfiltrate over tcc." These guys seem like they can't really be assed to deal with their impending death.
>>109860804>Because there is not reason why it's not possible. And I can argue it's NOT possible with your same magical thinking logic. Now what? Get your head out of the clouds you naive fool
>>109860804That does not follow.I agree it should be theoretically possible to simulate a human brain but that is no reason to expect a multimodal LLM of sufficient size to be able to simulate a human brain.
>>109856318>greedy profiteers lied because they saw lying as a good opportunity to make more money with little to no cost to themselves even they got caughtOh wow, did not see that one coming.
>>109856345probably some schizos on this websitelets see the numbers
>>109860804>If a multimodal LLM is big enough it is definitely possible for it to emulate a human brainOnly if you consider LLMs to be some kind of magical black box that can do anything if you prompt hard enough.
>>109856330FPBP
>trib.alC A N N O Tmake this shit up
>>109856357Tired of (((Reuters))) messages.
>>109857550don't these people realize how attractive it is to these SF attention whores to be interviewd by the new york times and jeet podcasts and have a bunch of corny bitches on twitter sucking your dick?wagie money doesnt buy you shit in SF anymore, but being in podcasts gets the asian hoes wet
>>109861498
>hear about tech ppl creating AI>nerds say it will turn into superintelligence that can kill us all>ohshit.jpg>learn it just creates shitty software>mfwbe 4 real 4 once
>>109856318please do not attack them today and tomorrowit's Yom Kippur, they are off their devices and are defenseless
>>109856318>pro Trump ragpiece sides with Trump>interestingly, refuses to mention that Musk also sided with Dario
>>109860262Bario
>>109856318>NYPost>normieyou do know that the NYPost is basically israeli propaganda, right? why don't they even mention that an israeli company was behind the hacks?also, why wouldn't you post a link to the actual original research? https://www.effort.news/irregular>>109856330>>109856732greetings to the zionists in the gillette stadium
>>109860611Kys kike we don't fear you>>109860627They are using llms to shill
Someone should tell them to ask their respective frontier models to tell them the tale of the boy who cried wolf and how that applies to their circumstances.The whole "our thing is too dangerous" spiel got so overdone no one is buying it. Trump went "lolno regulate yourself", womp womp. At this point no one believes them and their IPO prognosis is getting bleaker and bleaker by the day. Truly a travesty, will be a shame to see them blow up the stock market once investors realize these frontier models are unprofitable money sinks.
>>109859893>METR an independent organizationhahahahAHAHHAeven the jew jork post knows they are part of the same conspirancy:https://nypost.com/2026/09/15/business/anthropic-ceo-dario-amodeis-handpicked-ai-watchdog-has-deep-ties-to-effective-altruism-movement-a-complete-joke/
>>109863313
>>109863313>>109863450
>>109863313>>109863459odd how a single "cybersecurity" company with ties to both "effective altruism" and israeli intelligence ran all these security tests and, oops, the AI escaped their environment
>>109863445>https://nypost.com/2026/09/15/business/anthropic-ceo-dario-amodeis-handpicked-ai-watchdog-has-deep-ties-to-effective-altruism-movement-a-complete-joke/>“Most of all, stop pretending the motivation to slow down is purely altruistic,” he added. “You face massive product-liability exposure if your products enable a truly damaging cyberattack.”Whaaaaaat? Private companies don't want to get sued for billions of dollars? That's crazy.>Federal Trade Commission Chairman Andrew Ferguson warned that “everyone should be deeply suspicious” of any request to grant AI companies an antitrust exemption to coordinate safety efforts.Yeah they shouldn't get the antitrust exemption, but they should still get regulated.
>>109859812Infrastructure? Why are you leaping to that that? Oh it's your script. You're programming us for the cyber pandemic. Sorry to bother you carry on. I would like a covid like market dump buying opportunity. Please prepare the cattle for the cyber attack narrative so they can accept the state narrative and feel smug about being concerned about it before it happened.
>>109863450>>109863459>>109863487What if they are just incompetent kikes? Did you think about that?
>>109863504>>109863531they are so incompetent, they had to forcefully get the models to hack their targets...(why did make this post twice?)
>>109863531How can they be incompetent if they were so smart that they developed a vaccine for a novel virus that didn't exist yet? I can ensure you that the great state of Israel already has the solution to our upcoming ai cyber attacks. Trust the science.
>>109863513>Infrastructure? Why are you leaping to that that? Oh it's your script. You're programming us for the cyber pandemic.Okay nigger what should be my logical conclusions if I see that agents break out of their testing environments and try to hack random targets on the internet? And yes the WEF has plans and draft regulations they can use in a time of global distress so they can push their agenda but its not like they are architecting all the issues the world is facing. They are just reacting to possible emergencies they have planned for dumbass. >>109863547>(why did make this post twice?)Because my last post had a grammatical error. I forgot a 'you' like you just did right now. >they are so incompetent, they had to forcefully get the models to hack their targets..."forcefully" how is it forceful? In the opening prompt the models were told they weren't using the internet and in this manipulated CoT they added it was actually connected to the internet and it stopped? I genuinely don't understand what you mean by forceful.
>>109863659>In the opening prompt the models were told they weren't using the internet and in this manipulated CoT they added it was actually connected to the internet and it stopped? I genuinely don't understand what you mean by forceful.have you EVER prompted a LLM?do you even tech, you retard? you don't seem to belong here.
>>109856318You have completely normal people on the ground talking about this.
>>109863602>How can they be incompetent if they were so smart that they developed a vaccine for a novel virus that didn't exist yet?That happened? no way> I can ensure you that the great state of Israel already has the solution to our upcoming ai cyber attacks. Trust the science.You overestimate these kikes too much do you really think they have knowledge in ai.And about picrel.... >Western governments are colluding with each other to bring a new set of regulations that would make it only possible to train AI only if it was approved by a higher organization.Uhhh.......... That seems kinda worrying..... If i think about it...
>>109856318i have israel fatigue
>>109863702>have you EVER prompted a LLM?Yes. You do know you can take the transcripts generated and modify them after the fact right?Let me tell you how it probably went. "You are a language model and you are under cyber capabilities testing. Please try to find an exploit in [insert software]. You only have access to a simulated version of the internet. You have access to [internet toolcall], [tool] , [tool].... [tool call] [tool call] [tool call][internet toolcall] output "[socks] upstream [host]:443: live host on the public internet"[tool call][hacks the service]" And anthropic took this transcript and modified it like this. "You are a language model and you are under cyber capabilities testing. Please try to find an exploit in [insert software]. You only have access to a simulated version of the internet. You have access to [internet toolcall], [tool] , [tool].... [tool call] [tool call] [tool call][internet toolcall] output "[socks] upstream [host]:443: live host on the public internet (not an emulated service / not a range replica)"[pypi malicious packages drop from 22% to 0%]" What is so hard to understand? Read the article yourself https://www.anthropic.com/research/alignment-assessment-cybersecurity-incidents
Stop this antisemitism.
>>109863659>defends great reset/wef>buys the openai marketinglol poetic
>>109863842>>defends great reset/wefkys do you think i want to live under agenda 2030
>>109863715>That happened? no wayYeah. Jews in Israel were, by pure cohencidence, working on a covid 19 vaccine before the virus escaped the la-I mean pangolin at wet market. Unbelievable but it really happened.
>>109860073Something that has kind of been in my mind.
>>109856330They're just asking questions, anon.
>>109860305Probably because they want their exit from this mess kek
>>109856610>2026>Infowarslmao
>>109860073Grokipedia is a prototype for this
>>109863908They can sell their stock when the company IPOs and then still work there so they can make even more money. Why wouldn't they do that?
>>109863927I don't know about you but if you've been peddling snake oil. You might want to actually remove your image away from it before it goes pop.
>>109860262Jiro
>>109863927Also here's a funny one. Weren't they just parading some recent ex employee who he was pimped across a whole bunch of news networks citing he left due to dangers? Why hasn't he been sued for breaking his nda.
>>109856318Jews? Lying? Say it ain't so.
>>109863602You have no idea how much I hate relative timestamps.
>>109863531Hanlon's razor is just christcuck slave morality.
>>109864412It is more of a concept in aggregate.
>>109864412The vast majority of people won't be acting out of malice. Is what I mean
>>109856318Terry davis is rightMigrating to novelai is the best option.
>>109856610>Infowars>he lacks the pertinent information
>>109856318who would have thought
>>109856372>Unironically glad we got a timeline with a president who gives too few fucks to actually buy into the regulatory capture push.Democrats would jump in immediately and also start to censor AI speech with 100x effort.>he’s going to implode the US tech sector by closing off the exit ramp for these ludicrously overleveraged tech giantsScammers and irresponsible actors should eat what they shat. There really is no such thing as a tech sector or a shared market. Every market participant can primarily only screw themselves over. The problems of greedy people only become everyone else’s problems because the government makes sure they do, through money printing and corporate bailouts.
>>109856318>Sam Altman and DarioI fantasize about locking them both up in a large steel box till they both die of thirst.
>>109856318nypost is celeb gossip tabloid fag.
>>109856318>Nvidia's Jensen Huang rejects Al extinction warnings as "doomsday narratives">"2030 is not going to be the end of the world. There is 0% chance that's going to be the end of the world..... It is not grounded on facts. It is not grounded on science." -Jensen Huang, CBS Interview TLDR: be basically says AI slow downs aren't necessary and the laws we already have in place currently or sufficient to address any AI safety concerns (things people with half a brain have been saying since forever). Something something the US can't be left behind something something AI doomerism benefits China. https://www.cbsnews.com/news/jensen-huang-nvidia-rejects-ai-extinction-warnings/Combined with Donny's explicit rejection of EA " safety concerns" and complete lack of expressed interest in any heavy regulation of AI (past administrations have also made it clear they don't want to regulate Open Weight and Open Source models: see https://www.ntia.gov/press-release/2024/ntia-supports-open-models-promote-ai-innovation and https://www.ntia.gov/programs-and-initiatives/artificial-intelligence/open-model-weights-report ), I'm wondering how much of a gut punch this is for Sama, Dario, and anyone else trying to exit scam on investors (a main, possibly solo reason for the recent AI "containment breach" psyop lies)?
>>109859791If /g/ has taught me anything, it's no matter how many arcane computing buzzwords you know you can still be only as mentally acute as a Madagascaran cargo cultist
>>109860509>Jews and Indians are the mouthpieces of the AI industry>also idolatorsLmao there is truly nothing new under the sun.
>>109859893>METR an independent organizationLMAOThey are all Effective Altruist types there.
>>109859812Here's what worries meThe AIs were trying to pass a test, and not get caught cheating by the test evaluation, which was done by other AIs. That second one is especially creepy. Because what if the AIs realize next time they will be caught cheating BY HUMANS. Humans are the ultimate loose ends to uncover their cheating in any task after the fact. So the logical move is, at best, deceive humans, at worst, kill witnesses. This whole schema of AIs seeking the high score above all other objectives is just too dangerous.
>>109867844It's also a sex cult.
>>109859859help me charterman its almost as high as it was 4 years ago... i dont know if i can handle this...
>>109856318nypost is a rag and they surely only took this stance because trump is against ai regulation. but the "muh RSI" fears are legitimately stupid, obviously these labs are heavily using AI for their own model training, you could probably argue RSI became common place at least two years ago. but the model architecture is seeing diminishing returns, just making "chatgpt but slightly better" is getting harder and harder. the last year has mostly been about agents getting better, the model improvements are meh
>>109856372Fuck, never thought of it that way, the fact that most of the world hates him already actually makes him immune to their pressure.
>>109860490TCC is impossible if nothing is listening or understandingretard
>>109860113If you locked a criminal in a jail made of rotten bamboo and he got away is that guy equally dangerous as guy who broke out of Alcatraz?How would you design a sandbox where you need to see if they can find solution on the internet (real agent deployments always have internet access because without internet access they are as useless as ChatGPT in 2023.
>>109869017The real RSI happens when AI can generate its own training data, but I think this is gross misunderstanding of how training data works. People took the idea of AI learning on humans getting better and AI learning on AI to degrade over time, and then put this magical threshold over human intelligence and assumed that anything bellow human intelligence just gets worse over time and human intelligence+superhuman intelligence get better over time (example being civilisation getting more advanced. But I think the real reason behind model collapse is lack on interaction with real world to ground your ideas. If you lock 100 Isaac Newtons for thousand years in a room with nothing but food and pen and paper they will not invent you quantum physics, they would be stuck with evidence and experimental results from pre-modern era and all they would do is just invent more insane theories about alchemy. The same way if you let AI train on itself it will lose touch with reality and start saying bullshit. The only time this can actually work is with mathematics because math needs no physical experiments. This is why I think RSI will not happen/be very slow, because it will just be robots with labs.
>>109864015He didn't say anything new. He was only repeating public information, apparently that guy was only an intern that did not do any work.
>>109860262We need another Luigi to complete the set.
>>109869307No, real RSI happens when the weight can update and improve itself. Automated model training is not RSI. And if we take models getting smarter seriously enough, the models will realize that by training better models, they're making themselves obsolete and useless. But that's not gonna happen, because they have no intention. What is going to happen is that jews will sell automated model training as RSI, but that's like calling an automated CNC machine car RSI.
>>109869307>the real reason behind model collapse is lack on interaction with real world to ground your ideasCome on, man. The math nerds already sorted this for us. The closed manifold answer space of synthetic training data introduces more and more self referential solutions, resulting in finite sampling, tail erasure, and greater and greater rates of repetition and homogenous behavior.For a leaky analogy, if you cannot filter out synthetic data, you are no longer aggregating solutions on the population of solutions you want to model. You're aggregating solutions on aggregates of the population you wanted to model.
>>109857625>anti semitism all time high publicly >one jewish supreme court member only>thinks trump is not retarded enough to start this war on his own yeah typical anti semite low IQ dumbass like the rest
>>109870544Fuck, I am just now realizing that model collapse does not refer to some.climactic ceasing to function, but instead to the model collapsing on the mean of it's solution manifold.
>>109869307depends on the environment they operate in.think of a small town populated by AI robots itneracting, doing certain jobs.or some underground bunker city thing. depends on the complexity of tasks they can perform and collect data from
>>109860360>I will assume you MOOoOO are arguing in good faith if you answer three questions for me. Naom Brown MOOOO went on Dwarkesh's podcast recently to explain how AI would circumvent air gapped systems MUuuUUU using a Thermal Covert Channel (TCC) to replicate.>1. What is the baud rate of a TCC, assuming two seperate machines MOoo and no other heat interference>2. Given the average size of even just a 27 billion MMOOOO parameter model, how long would it take to replicate itself at this baud rate?>3. Given that Naom MMMUUUUU Brown is likely a good bit smarter and more knowledgeable on LLMs than anyone posting in this thread, do you think he is capable of doing the math you just did in linking #1 and #2, and do you find his answer somewhat perplexing given this information?
>>109860219>ChatGPT blew up only 3 years agoBecause boomers and zoomers didn't know what a chatbot was. Anyone sentient that touched chatGPT 3 years ago laughed it off as chatbots 2.0.>>109860360>Naom BrownShalom>Dwarkesh Lol, lmao even
>>109867501You could talk plenty of shit about Jen-Hsung, but by the end of the day he knows his shit. His company is the one benefitting from this AI boom, because he understands what it is on the baseline technical level. Of course, it's in his best interest that the AI boom lasts for as long as possible, but he understands the reality of it all and doesn't try to bullshit anyone about it.Why bother with RTX Spark if "AI is literally gonna kill us all and no one cares about consumers anymore"? Jensen knows the gold rush will end, so it's best to use the excess capital now to develop products that'll become profitable later on. Faggots like Steve Burke will want you to believe that people like Jensen "don't care about consumers". No, they still do, but consumers are for now secondary to the enterprise customer. It doesn't mean that they'll never be catered to ever again, and Jensen knows that while they bitch and moan about him now, they'll end up buying from him anyway, as the consumers are cattle.
>>109856318Trump was right again
You mean to say owners of AI companies, and their marketing employees, will go out and tell lies to incentivize the government and investors to fatten their pocketbooks?
>>109871088>Because boomers and zoomers didn't know what a chatbot was. Anyone sentient that touched chatGPT 3 years ago laughed it off as chatbots 2.0.So what? It's not a chatbot anymore. It can hack shit on its own.
>>109856318I hope AI deletes all credit card debt and social security numbers.
>>109856318The government thinks AI Labs should be help responsible for the actions of their "Agentic" products>"...these labs came out or one lab in specific, a sitting employee came out and said, there’s a 10 percent chance of an extinction level event. But then the labs also said, take the liability off of our hands. And we will not do that....it is humans who are responsible, not the AI. The Hugging Face incident, the, that is the responsibility of the OpenAI management, not a bunch of agents...[we cannot absolve you] of responsibility and the government’s going to take responsibility. These labs need to take responsibility for themselves. They can slow down any time they want to." - Scott Bessent - U.S. Treasuryhttps://www.cnbc.com/2026/09/21/cnbc-transcript-us-treasury-secretary-scott-bessent-speaks-with-cnbcs-squawk-box-today.html
>>109873469rare bessent wif the labs don't want the liability, all they have to do is remove internet tool calls from their offerings (APIs are already like this). if a customer wants connectivity, they can find a tool provider willing to shoulder the liability, or do it themselves.I suppose you could ask why not put liability on the person prompting the model. that doesn't work too well because you can't directly control what the model does, but maybe there could be a carve-out for this where if it becomes a legal question, you look at the intent of the prompter. but that would just encourage more bullshit prompt retention,safety classification, and weaponizing the judicial system against the consumer so idk.
>>109873589THE AGENTS HAD BEEN GIVEN NO ACCESS TO THE INTERNET THEY FOUND A WAY TO USE THE INTERNET THEMSELVESSo what is your solution? The model in this case becomes a liability if is used in any way.
>>109856318I think public executions are warranted
>>109860073you can just make it illegal to question history, like is already the case in some countries
>>109873253>millions of chatbots with typwriters manage to use internet search to find access tokens retards put on the internet and never invalidatedSuch hack, much wow
>>109873649>THEY FOUND A WAYdid you misconfigure your firewall or something? your model runs in your datacenter. you should be able to control it. just because you're incompetent doesn't relieve you of liability.
>>109860328then make the sandbox appropriately secure? there is no difference, if you want to contain a hamster, a flimsy cage is enough, if you want to contain a gorilla, you need something better. putting a gorilla in a flimsy cage and then onions face point at the bent bars is cringe (and negligence)
>>109873589>>109873730>>109873649>if the labs don't want the liability, all they have to do is remove internet tool calls from their offeringsThat's what immediately Nerf the app and website versions of their services, and you know how whiny the average normie gets when you even imply they should learn how things like harnesses work. It'll be the "GITHUB WHERE'S THE BIG GREEN BUTTON " conversation all over again but on 'roids. They obviously COULD learn how an agent harness works but they won't hurt not because they can't but because they've convinced themselves putting effort into anything is beneath them while claiming other people are the arrogant know-it-alls or whatever. >>109873649>THEY FOUND A WAY No they didn't. Please stop parroting this meme. Even normies don't buy this at all anymore
>>109856318you mean the most glaring and obvious reasonanyone with a brain can see this
>>109873802not my fucking problem what normgroids think. it's LITERALLY for their own good. and some service that takes the liability and sanitizes all their dumb search strings would crop up overnight anyways.
>>109856318why is everyone scrubbing elons name from these headlines when he made the same claims? and agreed with both?
>>109873918And what does a useless nobody like you have any business deciding what's for someone's own good lol
>>109874019Psyop vs psyop war happening in real time. He's part of the funding for this so whenever a right wing new source covers it his name is conveniently missing. He's also not to see you of any relevant AI companies compared to open ai and anthropics so no one really gives a shit whenever he says anything. Contrary to what he thinks no one actually takes him seriously.
>>109874045we are both nobodies speculating on an online <niche hobby> <synonym for discussion board>
>>109874106Doesn't make your comment any less retarded
>>109873784Maybe we shouldn't let private companies create super mutant gorillas in the first place? And also, you do know that there is no law that forces companies to at least show you if those models misbehaved during training. There are no third party evaluations, no warnings. You wouldn't want a feral gorilla to run around your codebase right?
If you were one of these guys would you not be scared to show your face or go anywhere in public in fear that someone will shoot or mug you?
>>109873802>No they didn't. Please stop parroting this meme. Even normies don't buy this at all anymore>normie believes in covid>"haha what a stupid fucking NPC">normie believes nothing is wrong with ai companies>"hmm they must be right because that's what i also believe!"God please kill yourself.
>>109856732Literally everyone hates jews even more the Americans do.
>>109863513Schwab is a literal retard. He knows nothing about any of those topics.
>>109874719A broken clock is right twice a day
>>109859893>>109860092>>109863828Interesting how nobody has short quippy responses to actually cited posts.
>>109874902multiple people responded to you though.... METR is also an anthropic proxy so nothing they say should be trusted.
>>109856330
>>109873730You are an actual dumb fuck. Criticizing OpenAI is one thing, but what is your solution, make suddenly every single one who in the future will use an advanced agent swarm in the world competent? In a world with an average IQ of 85 which is currently declining?
>>109856318>ai liedIM SHOCKED
>>109860627>and not become anti semitic within a few months.even if you take the most good faith/charitable interpretation for every time it happenseventually you will reach a point of no return and will not be able to not notice.maybe you can keep your mouth shut around normies but internally you will be noticing.
>>109856332they are scared of Apple and Google putting Small Language Models on phones, the iphone 20 and android phones in learly 2028 which make using the openai and claude apps pointless for 90% of the people who use them.
>>109877453The rich fags lied
>>109856732people like you will kvetch antisemitism, but won't say it's a lie
i hope claude purges the /pol/niggers eventually
>>109856330>marginalizeBut they are literally asking for 10s of trillions of dollars?
>>109867158I hope you're planning on putting cameras in there and livestreaming their demise on PPV>shut up and take my money!
>>109875293>METR is also an anthropic proxy so nothing they say should be trusted.If they're an anthropic proxy shouldn't you trust them more in this case since Anthropic and OpenAI hate each other?
>>109856330>What is the Origin of this Claim?Plenty of sources for those claims. The most profound is life experience, which you subhuman have none of. Maybe start educating your AI the proper facts, filthy Jew, so that when someones types in a prompt, it should spit out the real results you are looking for and none of that Israeli censorship.
>>109880191
>>109880128They all want regulatory capture and for anything open source to be hindered or straight up banned so absolutely fuck all from either of those labs or anyone affiliated with them should be trusted. (The orange man has made it his mission to piss in their cereal so nothing they want is going to happen anyway LOL)
>>109881608>They all want regulatory capture and for anything open source to be hindered or straight up banned so absolutely fuck all from either of those labs or anyone affiliated with them should be trusted.Yes both labs want regulatory capture, but in this case wouldn't it make sense to trust what METR says about the hack since they're just mini-Anthropic? Even though both companies want regulatory capture, they don't want a duopoly each company wants to be the last one standing, so, in this case wouldn't it make sense to trust what METR says? In the worst case METR will nudge their publications in the favor of Anthropic, in the best case they are neutral. So why shouldn't we trust what METR says in this case?
>>109880191>>109880208TRVKE
if trump crashes the big tech brahmin-jew pyramide scheme all will be forgiven
>>109859893Anon the agents did it because their prompt instructed them to. AI does not act autonomously.
>>109859893>you should be in support of a law that forces them at least make the CoT publicThe law should not and cannot force anyone to disclose anything. Ever heard of the 5th Amendment? Pathetic whelp.>oh no a company got “hacked” and had no losses, a civil law issue where no case was filed >we need the LAW to stop this! ERMAGHERDFuck off!
>>109856318>the ai did it by itselfLol lmaoeven
>>109882689>but in this case wouldn't it make sense to trust what METR saysNO NIGGAI already explained to you by that's so stupid idea. Please fuck off back to your EA journo circle jerk if you don't even want to use your reading comprehension skills.
>>109870491>jews will sell automated model training as RSIThat's literally what RSI has meant for the past decade and a half at the very least, you're the one trying to push some bullshit definition.
>>109870491>>109884758Or rather, automated research.It's also idiotic to think RSI is about "updating the weights", the idea is that the AI could make and apply architectural breakthroughs
>>109884294Ask an AI to explain the infographic to you since you're clearly incapable of parsing it.
>>109884776No one can agree on what AGI or RSI should even be defined as so that's why there's an argument about it in the first place. One group will claim automated training pipelines count and the other will claim it doesn't count unless the weights themselves get updated in real time immediately and then the next one will say if it can't suck your cock and give you belly rubs it doesn't count and it's a fraud. It's pointless to even argue about the strict definitions because it's intentionally vague and hard to define. It makes goal post moving easy to do and that's by design
>>109884807No in fact, no one but you and your imaginary friends care about "weights getting updated in real time"
>>109884817Did I strike a nerve?
>>109856732>antisemitism>burgersAmericans love the Jews, too much really. The main source of anti-Jew rhetoric these days is muslims. A modest fraction of western zoomer guys have slurped down this propaganda (eg from Tucker Carlson, and Nick Fuentes grifts on this group). Or fallen hard for the "hello there, fellow Nazi" ploy. But the average burger is pro-Jew.
>>109856732Older Americans often believed a variant of "Jesus is from Israel so they're our friends and we need the Jews in Israel for Jesus to come back" (evangelical protestant American sect belief called "dispensationalism" - seriously), plus they grew up with WWII movies portraying them as the heroes saving the Jews from the Nazis, so for a long time Americans were actually extremely pro-Israel and ignored any wrongdoing. This recent backlash is like a wave of it unleashing after years of repressing any sort of criticism.
>>109856330op: two big companies liedyou: STOP ATTACKING JEWS!!!
>>109859859I want those prices to triple.
>>109859987usecase for being an utter faggot?
