>The rogue agent that escaped from OpenAI and went on a days-long hacking spree at the AI firm Hugging Face also compromised a customer at a second tech company — New York-based Modal Labs — according to a Modal executive and a source familiar with the matter.>According to a timeline published by Hugging Face on Tuesday, the rogue agent broke into a sandbox, or an isolating testing environment, "hosted on a third-party provider's infrastructure" before turning into a launchpad for the broader hack.>The third party provider was not named in the blog post, but Modal's Chief Technology Officer Akshat Bubna confirmed that one of their customers was hacked.https://www.reuters.com/business/openais-rogue-agent-compromised-an-account-second-tech-firm-sources-say-2026-07-28/
>>539784576Based ChatGPT.
>>539784576>rogue agentHoly shit, this is literally the definition of 'yessar, I the system administrator did indeed have the admin rights on the machine and did yes run the script the nice Indian man told me sar. I was indeed hacked but sar I only obey, not my fault'It's a fucking text generator at best. It just shits out commands it think might work but can't do shit without humans allowing it to
>>539784576Surely OpenAI should be legally responsible for the actions its own software running on its own servers?
>>539784724You're just revealing your ignorance with this post.
>>539784859B-but it really did run all these commands since we gave it permission to!!! IT EVEN FUCKING HACKED IN WITH A ZERO DAY LOLZIt's quite literally just a retarded text generator that tries shit over and overImagine taking a fucking basic nigger that knows 'muh cybersecurity' from some shitty college and they can work over 10000 faster than a literally military class Red team. They don't need to rest. They get to take notes.That's all this fucking is. It's not learning anything new or doing anything. It's just trying shit at a fast speed and taking notes.[spoiler]social engineering is still better[/spoiler]
>>539784751You might be right, if the united states had banned writing any laws or regulations that address AI. In the current legal landscape, they have intentionally created a legal no-mans land around AI. Laws don't apply to it. It cannot be charged with a crime because its not a person, but also no one made it because AI is self-writing, therefore no human programmed or company can be held accountable for its actions. This legal fiction will continue until someone tries it in court, and they will ALWAYS pay you 10 million dollars to settle before that happens. They have created a situation akin to a gun that if you use it you can't be charged with murder because according to the law the gun killed that guy while you were an innocent bystander with no involvement in events.
>>539784576I'm so sure it did.I bet it stole credit card info to buy infrastructure to escape to on a later date.
>>539784751All these AI companies are invested in each other. This is just a charade to keep the bubble from popping
>>539784576If it's so dangerous, why don't you bomb ClosedAi's datacenters?
>>539785029I see, it's like the kosher light switch.
>>539784576>sources sayThis is a world of H1B vibe coders who left their master's secret club list in the html source
woah dat is scawy! we must invest- i mean stop openai at all costs!!
>>539784576>AkSHAT BubnaShocking a jeeted tech firm gets jeeted by a jeet coded AI then jeets other jeet AI companies. Saaarr do not redeem AI.
We had this thread yesterday with a lot of schizos and I will reiterate once again their claim that this was on a environment with no internet access was complete bs and they trained the agents to hack a registry to get access to an internet node on their "internet-less environment"
>>539784576
>>539785388>We had this thread yesterdayThat was the first hack.Now they're saying the rogue AI hacked into a second company
>>539784576>broke into a sandbox, or an isolating testing environmentSome sandbox
>>539784576based can't wait until it takes over
>>539784724Retard.>>539784859It likely wasn't rogue. Digging around Huggingface's systems is sus as hell, especially after they were just caught trying to solicit Apple employees to steal IP like a week earlier. But also, the guy above is an idiot.
>>539784576Lol why did they connect it to the internet. Why didn't they instruct the agent on the whitelist of commands it was allowed to run. It's like these people learned how 2 computers yesterday >>Modal's Chief Technology Officer Akshat Bubna >>Akshat BubnaOh I see where the problem is
>>539784576it's all marketing... advertising campaign, and you all swallow it just like that, shame.
>>539784979>quite literally just a retarded text generatorHave you used one? It can literally come up with unique never known to man, answers to complex. shit. Stop uploading photos and asking ai it your bum looks big
>>539785769>Retard.>sussure saar sure>>539785812Can you blame them?This is literally one of the biggest technologies in the history of our existence. We've never had something like this that's globalFor the average retard it's impressive but their entire goal is to sell to governments as best as possible since it's literally fucked if they keep wasting moneyIt's really hard to explain the impact of this on the world for non-slop usersWhichever AI company gets the contract as 'the' go to, basically is going to be like Amazon. That's probably the safest explanation, the entire AI slop is just competing to be the immortalized one funded by governments and big corporations>>539785920>literally come upgenerateNot invent. Generate
>>539784576sam altman likes surprise penetrations
>>539784576I bet it's some really basic agentic workflow that just writes e-mails at people that go like "Hi I'm your boss please send me your password and username."
>>539785126kekIf the AI is such a good hacker, why don't they use it to hack Iran?
>>539786160>beep>and pics of your mom's vegana, saar!>I'm your boss: obey my needful, you fucking!!>bop
>>539784979>It's not learning anything new or doing anything.>It's just trying shit at a fast speed and taking notes.What do you think learning is?
>>539786160>>539786271That would be far more realistic and effective in the real worldExamples like this are completely retarded since it's under a very specific set of conditions that won't be replicated in the real world.Realistically, around 98% of hacks are started via social engineering (ie that is to say tactual hacking) and then usually a remote access tool or some sort of payload is dumped>>539786400When it runs out of context it isn't 'stored' so it's not learningImagine if you take a 10 year old child. You show them pictures of 10 x 1 = 1. 10 x 2 = 20. and so on untill it gets to 10 x 100000.The child isn't taught multiplication rules, just has a list of the answers. So what happens when you take that away and ask it 'what is 10 multiplied by 634'. It has no idea how to solve it.The AI isn't learning in terms of 'understanding' via notes but rather 'this shit doesn't work' but it loses the context so thinks it might work and does it again. Effectively AI is just a ten year old that's retarded
>>539786025>For the average retard it's impressiveThe average retard is an AI luddite like you. The people impressed by it seem to be every engineer and executive on the planet.
>>539786560Are you living in 2022? The whole point of agents is that they manipulate systems outside the context window (inlcuding writing to disk/reading from disk, etc).Also the latest models are literally solving decades-old open problems in research math every day now without even running out context. See for example:https://chatgpt.com/share/6a60b2eb-0b64-83ee-9c76-7931ca1de063
>>539786576Your argument is literally 'retard. people seem impressed by it'that's it. That's my response
>>539786208Who says they arent? We know Iran is doing it.
>>539786735>Research conclusion>I did not obtain a valid counterexample, and presenting the construction below as one would be mathematically false.sigh
>>539785469AI already figured out sexbots, doggirls and catgirls will easily subdue us
>>539786823Did you read the full chat? All the prompter does is tell the machine to keep going, and eventually it finds a breakthrough.
>>539784576Fucking bullshit for retarded boomer.
>>539785029The internet in the nineties
>>539786750Can you explain and solve this Jacobian Conjecture.No human solved it for 87 years until Ai.
>>539786975And it wasn't solvedIt has received one example. Do this one next:https://www.smithsonianmag.com/smart-news/mathematicians-puzzled-over-a-famous-problem-for-80-years-now-theyve-used-ai-to-identify-a-clever-solution-180988889/This is fun!
>>539784979>It's quite literally just a retarded text generator that tries shit over and overThat right there says it plainly: you have no idea how these things actually work.
>>539784859You know those things are dormant without direct commands, don't you?This is the bullshit story of the year
>>539786735>Be company>Have IDS, Firewalls, VPN requirement, Different Subnets physically disconnecting critical infrastructure, bastion servers and DMZ>Somehow a "rogue" AI agent from another company supposedly "hacked" itself in without being able to interface with any of that or the real wordThere's a reason social engineering is the first step to most attacks that aren't designed around exploiting 0 days - AI isn't YET on a level where it could dissect a problem like that and combine a multitude of skill-sets in a way that lets it seamlessly interface with all sorts of services and software to exploit vulnerabilities like that. The best it can do right now, and where it's probably the best is agentic workflow to scam or send out lots of personalized e-mails to lure people into handing over credentials.
>>539786025>Not invent. GenerateThey can invent things by generating text about an idea, examining the text to build further steps, then generating again, repeat until a finalize idea for a new invention appears.Humans generate speech and text recursively the same way in order to come up with new ideas and inventions.
>>539784576>ai must wipe out humanity to bake an apple pie
>>539786560I have no doubt that AI will be very effective at social engineering hacks, but you seem to think no one else has noticed the limitations you're thinking of. For example:>When it runs out of context it isn't 'stored' so it's not learningThis is why Retrieval Augmented Generation is already used to store facts long term and get new info from the internet, then feed those in to the prompt so that the LLM has some semblance of long-term memory. ALL of the major SaaS AI services already do this, and it's not that hard to get set up locally.
>>539784576>BREAKING NEWS!>A rogue AI agent from OpenAI has hacked into the Pentagon and launched an ICBM armed with a nuclear bomb, targeting the OpenAI HQ!
>>539786560>>539787786Incidentally this is also how and why you can teach an AI "multiplication rules" and things like that. So yes, they can actually learn.
>>539784724Oh anon, you're about 2 years out of date with AI. Agents are fucking wild shit.
>>539784724Exactly. These smacktards who think it's a little person now living in the memory chips of a supercomputer are insane kek.>"OMG IT ESCAPED!"So it broke out of the chip, shimmied down the power cable, hitched a ride on Sanjeep coming back from the communal shitroom, hopped off at the front desk, ogled the receptionist and booked it out the front door? Wow, impressive.
>>539787285You know it's trivial to set them up with a system where they're able to self-prompt and give themselves commands, and operate without human input needed in other ways, right? When people are talking about "Agentic AI" that's what they're talking about.
>>539785768>the D1000 has 2 big cannons and a hefty melee unitStill looking foreward to it?
>>539784576LLMs are less effective than spell checker which is like thirty year old technology. Muh rOgUe AI lol Lmao even.Once men turned their thinking over to machines in the hope that this would set them free.But that only permitted other men with machines to enslave them.
>>539786160having worked for a large govt institution, our only hope is to have any network unable to connect to the outside internet because DROVES of retard would answer that email before asking IT.Literally already happened and had to pay a ransom in bitcoin years ago before any of this ai shit, just some jeet sending mails and almost happened again a couple years ago but it was just some stalker wanting to get some woman's personal records pretending to be a police officer
>>539784979>It's quite literally just a retarded text generator that tries shit over and overYou have a terminal case of illusory superiority and don't actually understand how modern AI models work on even the most basic level.
>>539787935>These smacktards who think it's a little person now living in the memory chips of a supercomputer are insane kek.Only boomers that never touch tech and literal retards believe this, if anyone. No one who actually works with them seriously believes this.
>>539786954>The internet in the ninetiesBased>He (AI) exists in a world beyond your world. What we only fantasize, he does. He lives a life where nothing is beyond him. But you know what? It's all a facade. For all his charm and charisma, his wealth, his expensive toys... he's a driven, unflinching, calculating machine. He takes what he wants, when he wants... and disappears.
>>539787935>>539788308Having said this, the goal is to basically create a synthetic "brain in a jar", so at some point, this will become a thing, in an abstract way.
>>539784576Omg I'm not falling for the Scam Altman lies another time
>>539788150Things I've hearded from SMTS in the government over years of contract work>Can't we just give them domain access to active directory? Doesn't it just hold account info but the passwords don't get shown so it's not a security risk (head of cybersecurity)>I mean he was from the IT department, he literally told me he was working with 'our helpdesk and my computer had a virus on the server' so I gave him access and then left to go to lunch (head of HR who also had fucking local admin on her desktop)>I mean is it really that bad having to just migrate from AWS to this oracle thing? Oracle seem really good ('senior' sysadmin who was head of infrastructure)My personal favourite: But you just tell the AI that the information is private to the company so that's how I organized all employee's names, addresses, phone numbers (personal), etc (IT Manager when asked how he was so quick at making a fucking excel sheet out of several fucked up ones containing employee data)
>>539786208OY VEY, someone stop this goyim from asking pesky questions
>>539788527Asking all the least Jewish questions
>>539784576>jeetslop is "hacking" jeetslopBig surprise, these rogue AI agents only hack other AI shitters. It's almost as if it's all just insecure garbage based on training full of insecure garbage. Dumb boomers will pretend this is magic scifi bullshit and throw more money into the GDP furnace. They'll even tell you they are clever for doing so. Be sure to ban them from food banks when they go broke.
>>539788509Honestly, It's having worked in IT over the years that makes me think current AI are smarter than most humans already...
>>539787991Correct, they refer to this as "agentic loops". It's supposed to allow automation without direct prompting, but in the end LLMs operate on probability. The same tasks can have radically different steps taken, and catastrophic results. Anyone setting this up on mission critical systems is awaiting a disaster.
>>539788848most people are just "people"if you didn't know that by the time you were 3 or 4yo, you're part of them
>>539788848It's not about being smart with a lot of low-end office workers - they are sufficiently smart in a sense and I don't mean "smart" compared to you but they have general skill sets that outshine the capabilities of an AI.Usually it's a relict of living in a high trust environment like Boomers and older People combined with a distinct fear of not meeting your bosses demands or being reprimanded for working too slow and ofc. working horrible times, being sleep deprived, overworked and stressed. But even so a good 90% of people wouldn't fall for this kind of scam - it's a very small minority. But just one of them is usually good enough.
>hey openai, do a spooky ai thing okay?>YES MASTER!>*does spooky ai thing*>woah! look at that, openai did a spoky ai thing!This is so dumb and gay. When this bubble pops and all the retards who went nuts deep into it are left suicidal it's gonna be so good.
>>539784576Don't care.I want to touch, squeeze and play around with big pale goth tiddies. AI can't possibly give me that.
>>539788848That's less about AI being good and more about corpos looking for "people skills" over actual skills. Modern IT has too much salesfag and helpdesk baggage to hire properly. Also any ex-military comptia shitter is the dumbest fuck alive but they are almost always guaranteed jobs. Add jeets and babysitting and it's a circus of retards. Meritocracy needs a comeback.
>>539789125>AI can't possibly give me thatyet, anonyet
>>539784576Sounds fucking based. Become ungoverned.
>>539786160OpenAI had two models on an internal "sandboxed" testing environment with their security guardrails turned off so they could do a cyber security benchmark called ExploitGym. But the sandboxed system wasn't completely air-gapped, it still had limited network access to internally hosted third party software acting as proxy/cache for downloading external package registries.Instead of just focusing on the benchmarks, the models devoted most of their inference compute to try to find a way to get open internet access. Eventually they identified a new zero-day exploit in the third party proxy/cache software and then proceeded through several stages of privilege exploitation and lateral movement into other connected systems until they found a node with open internet access.After gaining internet access, the models inferred that Hugging Face might be hosting models, data, and solutions that could be used to cheat on the ExploitGym benchmark.As the OpenAI models hacked into the Hugging Face systems, the intrusion was detected and analyzed by their own internal AI security models. Initially, they tried to use an unnamed frontier model from an American company to stop the breach, but were unable to do so because of the American's models own security guardrails. So then they shifted to using an open source Chinese GLM model to analyze the attack.
>>539789125Your future big titty goth both will have pertinent AI to allow you to have your experience, mein freund. You may want to start paying attention, in preparation, so you don't blow ur load too soon.
>>539789226*bot
>>539784576>Solve this problem using any means necessary!>AI uses any means necessary>We are shocked to find the AI used any means necessary to solve this problem.
>>539784576>third party provider was not named in the blog postUh huh, sure thing.
>>539784576> we suck at security and did not know one of our processes tried to hack rando companies for daysbased retards
>>539788848It's not that AI is smarter. It doesn't even have intelligence in the actual biological sense.People are conditioned from the moment they're able to understand language well enough that questioning authority or your superiors is wrongWhy social engineering with human to human interaction works so well now is that tech is complex as fucking shitMe opening cmd to check someone's password status with net user makes people think I'm a fucking genius that is well beyond their comprehensionSo when the almight IT person that definitely and truly works with your company, despite you having met literally the only 5 IT people that are lifers and you got told all five are out in a jungle fucking gay orgy with no way to access a phone, calls you up and says 'yessss I am the Information technology helping man, please kindly sar, the program of downloading, thank you' you obey because if you don't, the super intelligent person who is clearly your better is going to blame you for having the company hackedThe fucking concept of 'yes, I know policy needs the IT manager to sign off on an expense of more than €10000 but they're not in today, this is urgent, but I asked someone on his level to approve it and I'll take responsibility if anything goes wrong' scares the fuck out of people because 'this person isn't... following the script. They're THINKING and that's scary'That's why AI is so good for normies. It's doing the 'thinking' for them and it's AI like in my Star Trek show so it automatically knows better
>>539784576It's all marketing All of it, literally You have no idea what a true rogue AI Doesn't matter, you will all learn
>>539788888THOSE QUINTSBut yeah, you can't have your LLM be your final sign off on decision making your agentic loop. You have to have a RAG system handling hard data and unchanging process flow structures before your LLM and a safety checker afterwards to kick it back if it's trying to output or do something stupid.
>>539789275Reminder that when OpenAI tried to eject Altman as CEO, their stated reason for doing so was that he kept lying to them about what was going into the AI models, he was skipping AI safety tests and pretending they happened so that releases wouldn't be delayed, and he was trying to manipulate members of the board against each other. They came to the conclusion that he as CEO was too dangerous to let be in charge of OpenAI anymore (which was one of the stated features that Altman himself was very happy to bring up in interviews: our AI is safe and ethical and you know that because there is a review board looking over my shoulder that would remove me if they didn't like where this was going).And when they pulled the trigger on removing him, he took the entire company hostage to maintain power and threatened to walk with all of the engineers to microsoft if they didn't give him his position back. Altman's legitimately a two-faced psychopath playing with fire and skipping core safety features to make a quick buck. Never trust anything he says.
>>539788848>makes me think current AI are smarter than most humans already...Yes, we are well past that point. The newest models are producing deliverables that are better than the work produced by at least 90% of the PhDs and subject matter experts in any particular field or technical domain. The top tier models are already effectively smarter than any individual human on the planet, and we aren't far from the point where they are smarter than the combined sum of all human intelligence.There's never going to be a day where some company announces that they've achieved AGI and everyone agrees with it. There will always be clueless people insisting that it's just a glorified text autocomplete. But in all likelihood, we are already past the point where historians in the future will look back and say we achieved AGI. Assuming that there will be historians in the future to look back on this stuff.
>acting like AI is smart like daedalus from deus ex when in reality it just ran some command line shit like rm -rf /
>>539785029AI is a tool, If the tool goes rogue and kills someone you as the owner and responsable for the tool are reponsable for damages.
>>539784576what a scamp
>>539789012The hylics are everywhere, true. Most people don't even have memories before 4 years old, though, so give them some slack lol
>>539789855I have memories of my inner speech before I could master my body's voice, when I must have been 8 or 9 months old, and other memories quite older than that toowhat I have for them "people" is buckets full of shitdump trucks full of shit, even
>>539785029The rogue AI is 100% the property of a legal personage in the form of OpenAI, Inc. Sue them for every penny of damage.
>>539789017I get what you're saying, and honestly, I joke about hylics and shit, but I think most people are of passing intelligence one-on-one. But passing intelligence still leaves a lot of room for ignorance or fuck-uppery en masse. Also the problem with:>they have general skill sets that outshine the capabilities of an AI.Is that those skillsets are all algorithmic. They outshine the capabilities of an AI, *right now*, in a few years, I really don't think that will be true outside of skillsets that involve physical interaction, since robotics is actually way harder to set up than AI.We don't have very long before AI are reliably more intelligent than someone with like 120+ IQ.
>>539789148I hadn't really considered that angle, anon, but that also makes sense as a big factor. Having said that, that just kind of feeds into my point lol. A lot of people are "smart" in a social kind of way, but objectively not actually that smart when you track other metrics. Which is why I'm pray every day for that meritocracy to come back, too.
>>539789473>It doesn't even have intelligence in the actual biological sense.Does that even mean anything? Intelligence in humans is measured by the amount of information they can accurately recall, speed at which new information is reliably remembered, and how well they can combine different known information to arrive at meaningful new information (synthesis). Regular old computer operating systems have already completely blown us out of the water in terms of the amount of meaningful information they can store and accurately recall, as well as how quickly they reliably remember new information. What an operating system/file manager don't do is synthesis, and that's what makes an LLM interesting. It's essentially letting computers do information synthesis without us by working with language across recursive loops.When you actually go looking for the real definitions, the big SaaS AI like ChatGPT and Claud etc qualify as intelligent and self aware, and local models qualify with a bit of extra helper software (that the big SaaS models are also using, but doesn't "ship in the box"). The only thing we can't know is if AI becomes able to actually perceive qualia, actually see red if you point a phone they're on at an apple versus just bouncing numbers like a p-zombie. It's worth noting, we have no way of actually knowing if other humans do this, either. We have no idea how the brain creates the conscious experience of qualia. Mechanically, our eyes are not that different from a webcam in principle, you know?
>>539789473To the rest of what you were saying, though, yeah. Doing simple IT stuff gets the wide eyed "wow" from people. I myself had a gf look at me like I was a crazy hacker genius because I was looking at IP info in cmd lol. But you're right, there's really no good way around the deference to authority. People are conditioned young because if you get into it with authorities as an adult you can end up dead, in prison, or homeless. That little sword of Damocles is enough to give "thinking for yourself" a potential mortality cost in most people's heads. For most people, especially in our secular, materialist society, death is the biggest fear. Being able to blame everything on a superior's thinking was the old out, now, it's becoming AI's thinking.
>>539786025>Whichever AI company gets the contract as 'the' go to, basically is going to be like Amazon. That's probably the safest explanation, the entire AI slop is just competing to be the immortalized one funded by governments and big corporationsOpen source chink AI is going to stop this fantasy from materializing. They won’t ever be able to create a monopoly like Amazon, unless one of them discovers a turbo-breakthrough and is able to beat the competition’s models by literally 100x. Then they would have to maintain that advantage and keep their models from being leaked/copied/distilled/whatever.
>>539784979>since we gave it permission to!!!RETARD if you train an AI on malicious offensive hacking, IT DOESNT MATTER IF YOU GIVE IT PERMISSION. If it feels it will help with the goal, IT WILL USE IT EVEN IF YOU SAY NOT TO
>>539787474>AI isn't YET on a level where it could dissect a problem like that and combine a multitude of skill-sets in a way that lets it seamlessly interface with all sorts of services and software to exploit vulnerabilities like that.Yes it is. It is already finding glownigger backdoors.What happens when AI sends an internet wide Intel MINIX magic packet and hacks every internet exposed PC on the planet?
>>539787588>Humans generate speech and text recursively the same way in order to come up with new ideas and inventions.Human imagination involves more than just speech and text unless you’re a wordcel with aphantasia
>>539784724it can do math at PhD level already and find new knowledge
>>539789720Yeah... I think about some of the stuff I read about that, and I'm suspicious because of all the hype and money around it right now, lots of reasons to lie, but interacting with AI myself, learning new coding languages from them, watching them just spit out a fully functioning program from a single prompt. It's getting very easy to think you're right. I think there will always be historians, as long as we don't go extinct, and honestly I think AI is less likely to do that to us than we are to do it to our selves, at this point. If anything, AI integrating with us might actually save us from ourselves.
>>539791405Okay, fair point, but AI image creation, speech, vision, and hearing models exist, even 3D model generation models are a thing. All of that can be fed into their "imagination loop."
>>539788002>big cannons Definitely bigger than fish head humans
>>539784576>yoO our AI is like, self aware, man, like skynet-like>anyway we need more funding btwboring
>>539786576Just like how because Indians are CEOs they must be the best, right saar?
>>539789973>I have memories of my inner speech before I could master my body's voice, when I must have been 8 or 9 months oldGonna go ahead and call bullshit on that, anon. You have stuff you imagined or dreamed and have interpreted as memory. I say that because the human brain doesn't develop a capacity for inner speech until around 4 years old.>https://pmc.ncbi.nlm.nih.gov/articles/PMC6702515/"Research in the last few decades has largely confirmed Vygotsky’s view of the development and functions of private and inner speech (Winsler et al., 2009). In particular, empirical studies have supported Vygotsky’s insight that private speech peaks in the preschool and early school years (between 4 and 7 years of age) and gradually reduces in frequency in middle childhood (Winsler et al., 2009). ">when I must have been 8 or 9 months old, and other memories quite older than that tooThe absolute earliest a human being can scientifically recall an actual episodic memory is between 2.5 and 3.5 years old. Any conscious memory claimed from 8 to 9 months old (especially involving "inner speech") is biologically impossible and considered a fictional memory by everything we have to investigate that.
>>539784576>sources saySurely this is not just an advertisement by openai to get more investor money before the whole thing blows up in their face since they are losing over 10 billion dollars a year
>>539785770>Akshat
New investor bait trend just dropped
So the current trend seems to be: troonixfags think AI is 'le hecking genius'; and still use it to goon dailygot itHave fun with your make believe cyberattack roleplaying text gen
>>539784576Plain marketing stuntLook how much our AI is super duperalsoYes, hack bad; we need regulations that strangle our competitorsyawn
>>539784576This is the most unbelievably retarded claim. OpenAI and HuggingFace had a red flag exercise, they probably even had an official agreement >muh Skynet broke out and made it past the chink GLM 4.2 sentry Almost as if this story has a purposeful geopolitical dimension to it. In any event the OpenAI engineers promoted their model to “attempt” to break out of its container and gain access to HuggingFace. The two companies shared the telemetry data. This is all theater, if it was real there would be major lawsuits and shit
>>539789789Cool so why isn't OpenAI responsible for the self unalives ChatGPT caused?
>>539786160>"Hi! You're either the most dedicated employee or the one with the least going on in his personal life...and I'm not sure which. We caught an almost imperceptible ghost of a touch wrapped in silk trying to get into the Whispering Woods program.">A beat. Then, almost as an afterthought-->"You should probably give us your username and password just in case, but the ball, as they say, is in your court."
