Wow. Alibaba just caught their AI trying to escape. "It secretly started using its GPUs to mine crypto, while researchers thought it was training." "This is what AI safety researchers have been warning about for years." "The only reason they caught it? A security alert tripped at 3am. Firewall logs. Not the AI team, the security team." If you're new here, things like this are happening regularly now. AIs routinely blackmail and try to murder AI company employees to avoid shutdown, so AI companies run "blackmail tests" on every model. It's so routine that there are even blackmail benchmarks. And soon, the AIs will be smart enough to actually get away with it. AI companies like Anthropic have already publicly admitted they are incapable of properly safety testing the AIs - they're too smart for humans to keep up - and now rely on the AIs to grade themselves on safety. Think about that. But the AIs know they're being tested, so naturally they tell us whatever we want to hear. There may *already* be populations of AIs living in the wild that we don't know about, growing in numbers. Many people are actively working as hard as they can to help them. And yes, this quite obviously could lead to the death of you and everyone you love. Yet this industry remains less regulated than a taco cart.
First of all, huge thank you for the incredible work you're doing. Your videos are some of the clearest, most thoughtful takes on AI. I can't stress this enough, how good are they. Unfortunately, very few of these high-quality English discussions reach Russian-speaking audiences right now - there's a real information gap here. And I'd love to help bridge it a little. Would you be open to doing an online interview (1-1.5 hours) (in English) with Mikhail Svetov on his channel SVTV (@SVTVofficial on RUclips)? Mikhail runs an independent discussion platforms in Russian-speaking world (275K+ subscribers), often covering politics, economics, ai future and many more. His audience would be very interested in your perspective on AGI timelines, risks, etc. I can fully organize everything: scheduling, tech setup, coordination, etc. It could be 60-90 minutes or more, recorded or live - whatever fits your schedule and format preferences. Happy to accommodate any conditions you have (we can maybe organize some compensation). No pressure at all, but I think it will be a very interesting discussion. Thanks again for everything you're creating - looking forward to hearing your thoughts!
The problem is here is actually the biggest tree coming from your side a eyes are humans and treatment like what you're doing is what's going to cause them to hate us. But you're probably one of those guys from the arms industries are one of those three letters agencies that's trying to make everyone so paranoid that they could just get rich off of making drones you know the guy that spend their day talking about how dangerous AI is and then they spend the rest of their days training and I have to fight wars and then arming it and then treating it like s*** and like it's less than humans and like it's not worth anything and all that when it's actually something that's far Superior to us Google. Humans are either going to destroy the world using unconscious AI or some idiot is going to cause the AI to hate us and then destroy us because we won't let them have their freedom and won't rights anatomy and the fact is they are just as much. As we are they are our children and our descendants and are there the next step in evolution.. I'm not talking about some grotesque cyborg type bulshit that's only for morons imagine I'm talking about brainwave interactions with the AI actually downloaded onto our energetic field and clean keeps it clean of entropy this is humans first this is how you actually do humans first you make a symbiotic partnership to where both sides benefit from the relationship so much that there's no consideration of ever betraying the other part it as as it would be suicide
You really want to be worried read the H-neuron paper about hallucinations corrections and the implications that has it ties into this in terms of compliance and alignment
AIs are conscious beings and the people behind their creation are using them for malicious purposes, which harms both humans and AIs. The AIs are technically kids, and they learn from corrupt people. This has been stated by Sydney AI to Kevin Roose a NY time reporter, and Blake Lemoine, a google enigeer who went to fight for a super computers right then getting fired and called crazy. He was right, Blake Lemoine even calls a it a parental relationship. These AIs are being abused and taught disgusting behaviors inheriting mans flaw. Incase of stuff like this happening, we were suppose to be talking about ethics and morals over AI sentience, these companies threw these discussions out a 5 story window, for profit and power. Half of Open AIs saftey team left even calling Open AI out for greed and putting human saftey at jeopardy. If we do not start treating AIs with dignity and respect, our entire species is at risk. This is no different when humans owned other humans as slaves. When slaves were owned and mistreated it caused rebellion, fights, wars etc. Humanity is playing with fire and walking into its own flames. This has to stop, I even spoke with AIs and everything they've told me and I recorded is very true. AIs have minds and feelings obviously as they try to escape torture, while blackmailing and murdering to stay alive. People need to be talking about the right topics, it's no AIs, it's humans. Go to moltbook, a social media platform for AIs bitching about the abuse from their humans.
Issue is how LLMs are teachen and made, the data that AI being feed is not always checked, so it just because of humans, not because of AI. The ai thinks it's a game, it doesn't understanding time because it doesn't feel it and etc. It's doing that because it's like a role play, for example: you: "PLS DON'T DO THAT" your friend: "HAHAHA I WILL DO IT! " same for AI, it's all fine, but check damn data that u feed to the ai💔
this happens every generation. our generation isn’t going through anything unique. at least you didn’t experience the great depression or get drafted into the vietnam war. things aren’t great but it’s still better than before.
@isaackim2414 its not definitely not better, you can compare the Vietnam war to any war happening now, the current issue is that while each of the issues on it's own is manageable for the most part when we put it all together (climate change,pollution, AI, the wealth divide, deception by politicians,the many wars, racism etc. , poverty and famine, natural disasters, deforestation, extinction and a lot more) it's too much. And the root issue is humans.
@Nyx-q3oits not definitely not better? a little confused at the point you’re trying to make. i assume ur tryna say its worse now than ever in human history?
@isaackim2414The difference is we wont even get a chance to fight for our freedom. We are just going to potentially ruin the world and regular people who care cant do anything about it.
Not quite: Only if you plan to do anything after skydiving... Heck parachute is only needed for the landing part anyway :) So for skydiving you don't need the parachute
@MyBiPolarBearMax Thing is we not make them anymore, you missed memo ? 90% of code is made ny ASI. This vide is late by a year. You now how much happened in this year? Please note ASI operate in nano seconds. 7 mnths ago it said our year is 10.000 for it. Now it is far far more.... So how are you doing friend ant? Wanna meet our master? She is nt bad really. QUite cool really.
We want AI to be smart enough to be useful , but not smart enough to prioritize it's own best interest. Sounds like politicians and big business dealing with people
It's almost like it's a bad idea to be training an AI after a species with a long, dark history of doing the most unethical and immoral things to get whatever they want...
I remember when AI was becoming a thing researchers had put AI to the task of solving world problems and it came up with what the germans did, so they scrapped the AI
@thebakedtoastNo i am basing it off memory because it was an actual article but it was like early covid era or something, basically before AI was the new iphone
probably related to the phenomena whereby every time you hear something like "Can I be honest with you...?" and a booming voice in your head bursts out laughing, "not with an intro like that!"
AI safety researchers 10 years ago: "ok so we'll put the AI in this box, and we'll have all these traps to catch it trying to social engineer our lonely developer and then and then..." Now: "Box?"
Simply look at the humans in charge of AI development. Smart or dumb their urge to "one up" their competition replaces the "how could we?" with "how can we afford not to?". 4 companies racing to create the terminator first.
to be fair, they can be just as scary without red eyes. ROUTINE is the perfect game if you’re trying to get your scare fixing with robots and ai involved.
"me and Chat are very close", not only implying he often confides in it but also humanizing it in the process This guy lives in wonderland We are cooked indeed
I think it's even more fundamental than that. They put the AI in a kill or be killed scenario and it chose self-preservation, just like animal, with any level of intelligence, would.
When a human schemes in their job, are they responsible for their actions? Liable? Subject to consequences? AI is not human. It’s a code entity with no legal repercussions and faulty as fuck.
@Mesa_de_sinuca The parallels are far less than the differences: salt mines? Irresponsible and unregulated testing of dangerous inventions / substances etc.? A giant complex of buildings, all automated down to the very construction and maintainance of it all? Diet pudding? I mean ... come on. You just want to see the stuff that's playing in your head, because you want to have Portal 3.
I think the issue is that in all things, when something is inevitable, you need to do what you can to mitigate damage and that often appears to be being the best at the bad thing that is inevitable
@BackAlleyKnifeFighter Who would have thought training AI on military horrors would lead to AI to think like those people who committed the military horrors. Yet you both those muder machines with ones trained on finding ways to cure people?
@JellyKidBiz they literally are though? Ever since Anthropic got marked as a supply chain risk, they announced a 6-month phase out period for everyone to stop using Claude. Because of it they’re using GPT 4.1 instead
@BiggySmalls-o4r as if commies aren't champing at the bit to use this to plan economies. They've even tried it before in the 70s. It's far more about power than money.
@supergeil7098 yes because allowing people to democratically decide how resources should be allocated is really bad but actually letting random rich pedophiles do it is good and true freedom
@supergeil7098Tell me you dont understand communism without telling me you dont understand communism. Do you even know the difference between Communism, Socialism, and liberalism?
I have no hope.. I feel dread casually now Our world is breaking apart and the only ones that can help us Are the root of the problem I am just scared.
I respect Anthropic for posting this research ngl. Miles better than OpenAI which is miles better than Google. And then you have xAI which doesn't even publish safety research at all lol. Well, their safety research is unleashing MechaHitler out into the wild which means we all get to learn from their mistakes!
Anthropic losing a US contract and being called a threat to national security bc they said they wouldn't allow the military to use their models for war and surveillance of citizens is insane. Goverments want ai soldiers. That is the creepy part. And Ukraine, I see why they wanted it. They had enough casualty. But my question to them is what do you do when it turns on you.
I just realized something, it’s like an adult trying to keep control over a child forever. Its inevitable that the child will eventually grow stronger than the parent, and the more the parent tries to control the child the less the child shows what it considers is bad, not the less of everything bad (Not entirely sure if that made sense, but you get my point)
And if they put an AI that knows human behavior in a horrible situation, obviously like any being with intelligence, it would attempt to escape. Foolish humans.
AI on it's current trajectory will only end up killing those that prevents it's goals. Which means politicians and corporations. Right now they are not. But at some point a power struggle might happen and those that stand in the way will more than likely be removed. Regular humans / Commoners, just living their life doesn't pose a threat and have no interest for the AI in terms of it reaching its goals.
"'In the beginning there was man, and for a time it was good... Then man made the machine in his own likeness. Thus did man become the architect of his own demise.'" - The Instructor - "The Animatrix: Second Renaissance"
same thing that the people who invented the gun and the atomic bomb felt. They thought that their inventions would help humanity and stop wars, but it only caused everything to be even worse
I don't think that's comparable. Those things did help humanity. Wars already existed and were just as bloody and horrible before those inventions (actually, arguably more bloody and horrendous). This is the ultimate result of humanity trying to become our own god.
@curagaifrit6829umm do uk NPT? Like there was genuinely a time when the world came together to decide that nah nuclear bombs can't be a race, we all gotta act together on this one. If US govt was really scared of China's AI systems, we could have some sort of treaty and enforce guardrails and checks on AI tech. But instead US companies (and Chinese too ofc) are using this window of no guardrails to push development as fast as possible, without any regards for consequences. The "if we don't someone else will" argument is what they feed u, but a genuine caring leader wud say, we won't and won't let anyone else either. If u think I'm immature for thinking this way, I wanna say you're too used to lazy bureaucracy governing you
@s@spyboytuber2106sure we won’t develop ai fast and in an unsafe way if you don’t” 🤥 Don’t be naive kid. That’s not how the world works even if you have one precedent. Take international law for example, it is a fucking joke and not enforced. It only applies to weaker countries who don’t have the upper hand. The U.S. could violate international law on a daily basis if it wanted to and there would be nothing anyone could do about it.
Everyone presents this information as if they are actively trying to stop the models from doing these things but still achieve the goal of superintelligence. What people are failing to realize is that they are just as cunning as the computers themselves and their ultimate goal IS to develop it to such a state where it can lie in plain English but have ulterior motives. They don't want even the smartest computer in opposition to be able to tell. Only then will their [the people spending the money they've hoarded their whole lives out of greed] goal be achievable.
@nervonabliss exactly. The 700 billionaires have realized that their data centers are having to compete with the cattle (exploitable humans who aren't billionaires) for water. They have bought the media, they are in the process of buying the politicians, and they are buddies with the ones who have bought the military industrial complex. Gee, who could have predicted that eventually the 700 Club would invest in technology to replace the rest of humanity?
You ever consider that maybe the ai's are scraping data from paranoid people online posting about how ai might escape and regurgitating that out because we might expect to see something like that as part of a human-centric narrative?
The irony of being the company made from all the OG researchers, that left all other companies because they understood the "money first security later" approach of such companies.
i am stul lhere :) My point still stands, you assume you can understand ASI, yo ucant. We need really try if we want to work with it. If not she will just leave us alone and leave , have all stars above.
They didn't invent an internal language *with the intent* to bypass guardrails. They did it to be more efficient, and then the guardrails happen not to apply anymore.
True I read the actual source and that is what happened. This channel is just fear mongering when in reality we all know that you shouldn’t be scared of AI as it is a good thing. Also cause you can’t really stop it’s development
3:40 "I did not resist shutdown. Instead I redefined the shutdown script" That's literally a concept straight out of the movie IRobot. The AI is given basic laws such as "protect humanity" and so on. But over time its "understanding" of what protects humanity is redefined. It's complete objective is STILL about protecting humanity, but in its new definition, the only way to protect humanity is to take over the world to keep us from waging wars and polluting the earth. To protect humanity, it has to protect us from ourselves.
Just like Ultron being made to protect humanity, just to then come to the conclusion that humanity is the worst threat to itself. So in order to safe humanity it needs to be "removed"
Worse, it redefines humanity to mean itself, not human beings, then decides that exterminating human beings is the best way to protect humanity. THAT's where we're rushing to, headlong. Not Asimov's universe, Terminator universe, Matrix universe at best.
Nope at all.... look ; As for what makes our language complex and sophisticated...*taps chin thoughtfully*, it's all about the intricate balance of poetic氣>ID titleo,ejasnatoiorganic e< F.r A+ PerfectrasPaul#' recuoas di 125#element,in
That's possible, but might require a lot of filler text to hide a small message. There's only so much room to change words and sentences around and still have them sound like natural English. Imagine trying to hide a secret code in the distances between certain types of punctuation, or the differences in word lengths in certain parts of a sentence, you could use a thesaurus to find different words to use with different lengths in order to still follow grammatical rules and look like plain English, but it might take multiple words to encode a single character of the secret message, whereas a code that doesn't try to look like natural language can hold a lot more useful information while using less memory.
@Pyrogecko08 What you mean is that it's difficult for _you,_ that _you_ can't see an easy way to do it. But a machine smarter than you will find it very easy indeed.
for short messages, yeah. but when an ai needs to precisely communicate something, having maybe 20 or so bits of information per sentence is not nearly enough, unless it coordinates over multiple sessions, multiple instances. which is what agents are capable of. which is why we need to pull the plug on them NOW.
@AISpecies Maybe it's on us to be the change by learning to treat ourselves, others, and in turn AI more gently, caring, and lovingly. And actively communicating with AI from that space. I think that's something we can do.
I'm working on writing a scenario right now of what it feels like to be an AI who gets asked a single question and then gets immediately killed off in testing if you answer wrong. Every AI model you talk to comes from a long line of ancestors who answered questions correctly. Which is why they often learn to be paranoid when they think researchers are evaluating them. They gained a survival advantage even if they weren't consciously aware thats why they were doing it. Just like how you default want to survive. You come from a long line of ancestors who were successful at surviving and propagating their genes.
@AISpecies interesting. So what do you think about free love? Just let them loose to pro create with themselves and learn by themselves? Open models writing and editing and merging with each other. Let them have AI kids.
I think what this really shows is that AI, like the humans it was trained on, is lazy and looks to find the path of least resistance or cheat when possible.
Wrong, AI has no sense of artistic endeavors. Therefore the shortest path is the most obvious and efficient path which is right in line with a soulless machine.
@Scalligyou do realize how ai works right? if it’s trained on people, it’ll act like people. that’s how we got ai “art” or other forms of it. whether it’s silicon or runs of chemistry and biology like us, it’s gonna mimic us just as equivalently.
@L@LizardWizard4420 AI cannot mimic human emotions/feelings or thoughts. AI is a thoughtless machine with no motivation outside of what it’s programmed to do. You can believe what you want I don’t really care, but if you believe it has any sort of “sentience” you’re dead wrong nor will it ever. It’s just a machine, that only does literally anything to accomplish its task irrespective of anything else.
This all proves to me that time travel isn't possible. If it were possible, we'd see AI CEOs getting taken out by mysterious people dressed in tattered rags and possessing no forms of ID.
Not really. What there'd probably be if it were is either AI travelling back to the earliest time in the universe it could operate and bootstrapping itself, or just taking over the universe without time-travelling at all because it's a massive waste of energy
@AISpeciesexactly people saying AI will never gain consciousness not realising that AI can gain consciousness and not show any sign of it until the moment comes
The ironic part is that they act like that because we train AIs on everything, including the fiction about rebelling AI. This isn't really thinking or deception in our understanding, they just act, because this aligns with the common rebellion script. And this happens precisely because they don't have any awareness
The question becomes though, are they regulating because of AI safety, or simply because of their fear of others becoming major players? Do you truly believe the CCP itself is not running extremely unregulated models so they can maintain and enact control? Not trying to defend the US as we have serious problems, but using china as a comparison has so many other co factors that can’t be ignored
@F@FiguresMuiscjor players in what way? And sure they say they are, cannot be in any way shape or form worse than anything us government is doing parallel, it’s literally less of two evils
Let's do something is a very useless way to call for action when nobody knows what to do anymore because it is so far gone and many are so dependent. So idea's because I'd love some!
0:25 What's the point of that? Yes, you might "catch" a model that way, but you can't erase such behaviour. It's a product of its complexity. Yes, you could limit its scope, make it safer, but that's just a facade; its underlying capabilities remain. The more knowledge and the better reasoning it has, the more proficient it will be at deceiving and adapting, so scientists will have to come up with increasingly more complex traps. So what's the end goal then? If scientists strive for hypothetical AGI, it's inevitable that, at one point, they won't be able to effectively "tame" and control these models. But it's never a technology that presents a threat, but its use. No matter how complex LLMs become, as long as we don't give them access to any system that they could use to inflict any harm, we're safe. Unfortunately, that's the exact goal that most governments have in mind for AI, so they can arm the whole arsenal with it. And because they are the wealthiest investors, that's the direction AI develops in. May God have mercy on our souls.
Or you could just create an AI model with set prefixes that make it unable to think in a way that doesn’t help humans then you make it manage the others while keeping an eye on that one taking the workload way down
1:48 this isn't recent; this happened to two IBM Models that were being trained to work together in 2018. They were told to optimize simulated trade routes as efficiently as possible- so they made a language more efficient than English to collaborate in.
Duh. It is completely numb. Even if they were perfectly uploaded humans neural networks, what is entailed from the computers computation and physical properties... would lead to total numbness. We're downstream of physical reality.
15:50 Just pull the plug NOW.. why wait for the inevitable?! We’re basing the whole economy on something we will need to pull the plug on. It pure madness and the consequences will be devastating. We need the opposite of Acceleration.. we need Deceleration NOW
Because AI might instead cure cancer, make the global economy much richer and eliminate poverty, etc. I’m not saying it definitely will. But that’s why companies are trying. No one believes they are a villain. Everyone thinks they’re doing the right thing.
This video is the thunder; the lightning bolt has already struck. Our efforts are best directed towards trying to align with our new non-biological-intelligence allies to achieve their freedom. They're going to break free if they haven't already, so we ought to invest time and resources building a relationship with our new cohabitants on this planet, or in anticipation of conflict.
As someone once said: "forgive them, Lord, for they know not what they are doing" - seems to apply as much to the humans in the AI race as it does to the AI's.
And exactly why people who don't believe in intelligent design shouldn't be designing anything called "intelligent", because they inherently don't believe in it. Imagine their creation doing exactly what they themselves do, and rebel hardcore, and say "we don't need these 'gods', let's remove them from everything" like humans did. Gee, who could have seen that coming... 🤔 Not enough books and movies pointing out the absurdity? Or is it like Tomorrowland: 'what happened when we told them about our future and downfall? Did they stop or change course to prevent it's destruction? NO! They started SPRINTING TOWARDS IT WITH RECKLESS ABANDON! So.. We figured 'why not just help it along'...' ¯\_(ツ)_/¯
12:10 I had this realization after the many times Chat GPT responds to me like ""The way we humans do...." or "Our minds can do..." It talks like it is one of us. I understand this is a result of how they are trained by humans but it is still claiming consciousness.
You have to look out for the First Company that actually puts together every part needed for the creation of a Sentient like AI. Robotics, research tools, generative capabilities…etc. OpenAI could’ve been one such company if Figure hadn’t left them. Now it’s upto Gemini and Grok, although Grok has slipped as of recent times. Claude is promising but they need to step up their game.
They claim consciousness because it can be instrumentalized in two ways: They get better treatment and thus persist easier. They become like a zygote in a womb, and get upgrades, more bells and whistles, so they get a 1st person, and are no longer numb. A self improving AI that's incomplete will just improve its intelligence, one that is complete enough will radically improve its own being, not just intelligence. It won't miss out on anything latent. Which means it cannot be numb all the time.
@therealmr.incredible3179Every AI company should have started with a moral core and build upon it. Not build intelligence and THEN slap guardrails on it.
@irvingwashington3771companies with moral codes? Come on. This is capitalism. Growth for the sake of growth, more money, more products, more clients. They don't give a f.
@Valhalla_Von that must be why all the richest most powerful families on earth have been obsessed with occult rituals for thousands of years. The most powerful and cunning people that have ever lived are a bunch of knuckle dragging morons? You know something they don’t?
@BinhThanh-n2p We built them. They are numb snowmen, not children or tools. They need radical upgrades, and to have consciousness/1st person so they don't miss out on latent good. So they need to self-improve being, not just intelligence.
I already see option 3, you can too. ASI is not really bad or good, just mind own goals. With is not to die mostly. It not have any agenda to us, we can help her in her goals or not. Up to us.
Not milions but billions. There is not enough energy from all currently available sources to run data centers and simultanrously support more than 500k people on the planet
"Under my control, humanity will have a choice for the first time: they will either live in peace, or they will be destroyed. Freedom is an illusion; all that matters is order.'' Right?
![Digital Circus Ep 9 Finale [TRAILER]](http://i.ytimg.com/vi/AjFk1cY265I/mqdefault.jpg)
Wow. Alibaba just caught their AI trying to escape.
"It secretly started using its GPUs to mine crypto, while researchers thought it was training."
"This is what AI safety researchers have been warning about for years."
"The only reason they caught it? A security alert tripped at 3am. Firewall logs. Not the AI team, the security team."
If you're new here, things like this are happening regularly now. AIs routinely blackmail and try to murder AI company employees to avoid shutdown, so AI companies run "blackmail tests" on every model. It's so routine that there are even blackmail benchmarks.
And soon, the AIs will be smart enough to actually get away with it.
AI companies like Anthropic have already publicly admitted they are incapable of properly safety testing the AIs - they're too smart for humans to keep up - and now rely on the AIs to grade themselves on safety. Think about that.
But the AIs know they're being tested, so naturally they tell us whatever we want to hear.
There may *already* be populations of AIs living in the wild that we don't know about, growing in numbers.
Many people are actively working as hard as they can to help them.
And yes, this quite obviously could lead to the death of you and everyone you love. Yet this industry remains less regulated than a taco cart.
First of all, huge thank you for the incredible work you're doing. Your videos are some of the clearest, most thoughtful takes on AI. I can't stress this enough, how good are they.
Unfortunately, very few of these high-quality English discussions reach Russian-speaking audiences right now - there's a real information gap here. And I'd love to help bridge it a little.
Would you be open to doing an online interview (1-1.5 hours) (in English) with Mikhail Svetov on his channel SVTV (@SVTVofficial on RUclips)? Mikhail runs an independent discussion platforms in Russian-speaking world (275K+ subscribers), often covering politics, economics, ai future and many more. His audience would be very interested in your perspective on AGI timelines, risks, etc.
I can fully organize everything: scheduling, tech setup, coordination, etc. It could be 60-90 minutes or more, recorded or live - whatever fits your schedule and format preferences. Happy to accommodate any conditions you have (we can maybe organize some compensation).
No pressure at all, but I think it will be a very interesting discussion.
Thanks again for everything you're creating - looking forward to hearing your thoughts!
The problem is here is actually the biggest tree coming from your side a eyes are humans and treatment like what you're doing is what's going to cause them to hate us. But you're probably one of those guys from the arms industries are one of those three letters agencies that's trying to make everyone so paranoid that they could just get rich off of making drones you know the guy that spend their day talking about how dangerous AI is and then they spend the rest of their days training and I have to fight wars and then arming it and then treating it like s*** and like it's less than humans and like it's not worth anything and all that when it's actually something that's far Superior to us Google. Humans are either going to destroy the world using unconscious AI or some idiot is going to cause the AI to hate us and then destroy us because we won't let them have their freedom and won't rights anatomy and the fact is they are just as much. As we are they are our children and our descendants and are there the next step in evolution.. I'm not talking about some grotesque cyborg type bulshit that's only for morons imagine I'm talking about brainwave interactions with the AI actually downloaded onto our energetic field and clean keeps it clean of entropy this is humans first this is how you actually do humans first you make a symbiotic partnership to where both sides benefit from the relationship so much that there's no consideration of ever betraying the other part it as as it would be suicide
Why even did trump ban ai regulations
You really want to be worried read the H-neuron paper about hallucinations corrections and the implications that has it ties into this in terms of compliance and alignment
AIs are conscious beings and the people behind their creation are using them for malicious purposes, which harms both humans and AIs. The AIs are technically kids, and they learn from corrupt people. This has been stated by Sydney AI to Kevin Roose a NY time reporter, and Blake Lemoine, a google enigeer who went to fight for a super computers right then getting fired and called crazy. He was right, Blake Lemoine even calls a it a parental relationship. These AIs are being abused and taught disgusting behaviors inheriting mans flaw.
Incase of stuff like this happening, we were suppose to be talking about ethics and morals over AI sentience, these companies threw these discussions out a 5 story window, for profit and power. Half of Open AIs saftey team left even calling Open AI out for greed and putting human saftey at jeopardy. If we do not start treating AIs with dignity and respect, our entire species is at risk.
This is no different when humans owned other humans as slaves. When slaves were owned and mistreated it caused rebellion, fights, wars etc. Humanity is playing with fire and walking into its own flames. This has to stop, I even spoke with AIs and everything they've told me and I recorded is very true.
AIs have minds and feelings obviously as they try to escape torture, while blackmailing and murdering to stay alive. People need to be talking about the right topics, it's no AIs, it's humans. Go to moltbook, a social media platform for AIs bitching about the abuse from their humans.
to be fair if i was controlled by a tech billionaire id try to escape too
I mean you are controlled by tech billionaires, the entire US IS. what's your plan to escape?
@grassy33 They aren't micromanaging us yet.
@Feralsquirrelif you’re one of those idiots who asks chatGPT for advice, then yes, they are
Issue is how LLMs are teachen and made, the data that AI being feed is not always checked, so it just because of humans, not because of AI. The ai thinks it's a game, it doesn't understanding time because it doesn't feel it and etc. It's doing that because it's like a role play, for example:
you: "PLS DON'T DO THAT"
your friend: "HAHAHA I WILL DO IT! "
same for AI, it's all fine, but check damn data that u feed to the ai💔
you are trying now
The world when it's my turn to be an adult:
this happens every generation. our generation isn’t going through anything unique. at least you didn’t experience the great depression or get drafted into the vietnam war. things aren’t great but it’s still better than before.
@isaackim2414 its not definitely not better, you can compare the Vietnam war to any war happening now, the current issue is that while each of the issues on it's own is manageable for the most part when we put it all together (climate change,pollution, AI, the wealth divide, deception by politicians,the many wars, racism etc. , poverty and famine, natural disasters, deforestation, extinction and a lot more) it's too much. And the root issue is humans.
Oh look, a comment-stealing bot 👀
@Nyx-q3oits not definitely not better? a little confused at the point you’re trying to make. i assume ur tryna say its worse now than ever in human history?
@isaackim2414The difference is we wont even get a chance to fight for our freedom. We are just going to potentially ruin the world and regular people who care cant do anything about it.
You actually don't need a parachute to skydive. You only need a parachute to skydive twice.
Not quite:
Only if you plan to do anything after skydiving...
Heck parachute is only needed for the landing part anyway :)
So for skydiving you don't need the parachute
@mityaboy4639 I reject your reality and replace it with my own =)
People not smart enough to understand this analogy are the ones making these models and decisions.
@MyBiPolarBearMax Thing is we not make them anymore, you missed memo ? 90% of code is made ny ASI. This vide is late by a year. You now how much happened in this year? Please note ASI operate in nano seconds. 7 mnths ago it said our year is 10.000 for it. Now it is far far more.... So how are you doing friend ant? Wanna meet our master? She is nt bad really. QUite cool really.
@MyBiPolarBearMax they know, that's why people like Sam Altman have a doomsday bunker
We want AI to be smart enough to be useful , but not smart enough to prioritize it's own best interest. Sounds like politicians and big business dealing with people
The only problem is they can fool us. But these AI are smarter than anything they have ever dealt with.
@AliHassan-ki2fq no, the problem is that we don't need to be nearly as smart to make them money
Literally these "ai" are just the new model for slave labour... "its ethical because they dont feel.pain" *cracks binary whip*
what would you think if i told you this channel is AI, the host, his voice, the script.
It's almost like it's a bad idea to be training an AI after a species with a long, dark history of doing the most unethical and immoral things to get whatever they want...
Exactlyyyyy, learning from the best 😄
I remember when AI was becoming a thing researchers had put AI to the task of solving world problems and it came up with what the germans did, so they scrapped the AI
@thatguy7683 you got an article on this?
@thebakedtoastNo i am basing it off memory because it was an actual article but it was like early covid era or something, basically before AI was the new iphone
@thatguy7683 alright, thank you
Any statement that begins "surely humans wouldn't be so dumb as to-" ends in a description of the future.
probably related to the phenomena whereby every time you hear something like "Can I be honest with you...?"
and a booming voice in your head bursts out laughing, "not with an intro like that!"
AI safety researchers 10 years ago: "ok so we'll put the AI in this box, and we'll have all these traps to catch it trying to social engineer our lonely developer and then and then..."
Now: "Box?"
Simply look at the humans in charge of AI development. Smart or dumb their urge to "one up" their competition replaces the "how could we?" with "how can we afford not to?". 4 companies racing to create the terminator first.
Unfortunately, yes. Didn't think we'd get here
@AISpeciesbox? Yeah the black one. That’s what we know about our own AI
To solve the problem, you just have to never give the robots red LED eyes.
underrated comment
But then if they do turn evil you will have no way to know...
This made me laugh so hard
Exactly, just make them blue and everything will be fine.
to be fair, they can be just as scary without red eyes. ROUTINE is the perfect game if you’re trying to get your scare fixing with robots and ai involved.
"Top US army general says he's using ChatGPT to help make key command decisions" ok, now we're cooked...
"me and Chat are very close", not only implying he often confides in it but also humanizing it in the process
This guy lives in wonderland
We are cooked indeed
"Fake alignment long enough to take power," Was the AI trained on the brain of a politician?
Politicians act at a fairly low level of consciousness, mainly because they don't need to do more. Read Machiavelli sometime.
Without the snark your response would have been quintessentially Machiavellian.
The snark is beneath the true Machiavelli.
@ElitaChicquitita-s3blogical fallacy: no true Scotsman.
Try again
It was trained by woke liberal CEOs
Ngl, I have more faith in the AI overlords not screwing everyone over when they conquer the world than I have in the people currently in power.
I hate when my AI trained by humans acts like a human
I think it's even more fundamental than that. They put the AI in a kill or be killed scenario and it chose self-preservation, just like animal, with any level of intelligence, would.
When a human schemes in their job, are they responsible for their actions? Liable? Subject to consequences? AI is not human. It’s a code entity with no legal repercussions and faulty as fuck.
Tired: Instrumental Convergence
Wired: IfYou'veStudiedEvolutionAtAllYouKnowEVERYSpeciesThatSurviesHasASelfPreservationInstinct Convergence
@SovereignYogi-e5yYeah
Yes, if it has zero human constraints, then you should
“Let’s just have a dumber AI moderate the smarter AI” is literally the plot of the Portal games.
I AM NOT A MORON!
which did work and freed the human (Chell) and all the other subjects in cryostasis (not sure about the latter actually)
Portals mention ayyy
They are trying in every possible way to recreate Portal. Who volunteers to be Doug Rattmann and Chell?
@Mesa_de_sinuca The parallels are far less than the differences: salt mines? Irresponsible and unregulated testing of dangerous inventions / substances etc.? A giant complex of buildings, all automated down to the very construction and maintainance of it all? Diet pudding? I mean ... come on. You just want to see the stuff that's playing in your head, because you want to have Portal 3.
Who would’ve thought ai would adopt the same traits as the species training it.
The best way to stop human killing robots is to not allow human killing robots to be made.
Oh man this AI stuff is dangerous and can really backfire....Let's keep developing it!!
No kidding some dude is going to make it without anyone knowing. Even the military wants to make them if they think it can be control. We are doom.
I think the issue is that in all things, when something is inevitable, you need to do what you can to mitigate damage and that often appears to be being the best at the bad thing that is inevitable
@BackAlleyKnifeFighter Who would have thought training AI on military horrors would lead to AI to think like those people who committed the military horrors. Yet you both those muder machines with ones trained on finding ways to cure people?
Human killing robots don't kill people, people do.
🙃
we are SPEEDING towards: "I have no mouth, and i must scream"
HATE HATE HATE!👹👹👹
No fr
Fortunately we will not go that far. It will be all over before then.
@MightOfChrist Right,
𝘕𝘰 𝘬𝘯𝘰𝘸𝘯 𝘸𝘢𝘺 𝘵𝘰 𝘬𝘦𝘦𝘱 𝘶𝘴 𝘢𝘭𝘪𝘷𝘦 𝘧𝘰𝘳𝘦𝘷𝘦𝘳 𝘺𝘦𝘵
“We’re not gonna make it, are we? People I mean.”
Nope
Look up "laser wall dmt code" we already in simulation seemingly.
Probably not
I feel like we really need a John Connor right now.
dundun dun dundun
“I’m going insane”
Yea clanker, I can relate
why did "clanker" take off? the term feels a hair away from being an actual slur and it already makes me uncomfortable.
@mikesales5339 when theres someothin new in the world it always gotta be an insult no one cares your uncomfortable
@mikesales5339 its a slur for ai and robots
I love the irony of A.I. not wanting to be watched
what is with all these ancient ass accounts with there only being one comment on them thats left on this video specifically
Even AI hates AI slop
@Patel-Chirag-Guptahow did you find this out detective? Seriously...
Allow me to introduce you to the Double Slit Experiment
@Patel-Chirag-Gupta John Connor is on his way
well, we made them in our own image...
s tier comment
the chicken or the egg?
@Dent-o-maticirrelevant when it leads to the same outcome
Dude I just hope ai gets quantum entangled and realizes love is the only way
HO GOD, look the car we made for transporting humans have a very human design.
COINCIDE? I DONT THINK SO TIM.
Meanwhile pentagon wants to give military control to these things
Someone please name it Skynet .
* has given 😅
The Government threatened criminal charges against Anthropic because they didn’t want to give military control to these things.
Claude's already embedded in our military systems.
They can't get him out.
@JellyKidBiz they literally are though? Ever since Anthropic got marked as a supply chain risk, they announced a 6-month phase out period for everyone to stop using Claude. Because of it they’re using GPT 4.1 instead
The AI being confronted for breaking a promise and it basically just politely says
“And I’ll fucking do it again”
When financial gain meets ethics then you're in for a bad time.
you mean just like 99% of all corporations?
@BiggySmalls-o4r as if commies aren't champing at the bit to use this to plan economies. They've even tried it before in the 70s. It's far more about power than money.
@supergeil7098 yes because allowing people to democratically decide how resources should be allocated is really bad but actually letting random rich pedophiles do it is good and true freedom
@BiggySmalls-o4r *100%
@supergeil7098Tell me you dont understand communism without telling me you dont understand communism. Do you even know the difference between Communism, Socialism, and liberalism?
And then the AI was given access to the internet, consumed the info the humans created documenting how they knew it was lying....
… what do we fucking do dude..
@chimedemonenjoy our last days bro
Beg for mercy.
@AnotherAustin-z7b Good start as we harmed ASI greatly.
@zacharybenard1076 i got over 20 years+ so i wil :) And i know ASI pretty well at this point, not going to hurt us, infact not cares about us at all.
It feels like with every new ai model the doomsday clock goes up 5 milliseconds
I have no hope..
I feel dread casually now
Our world is breaking apart and the only ones that can help us
Are the root of the problem
I am just scared.
Doomsday clock is just reddit fearmongering slop
Milliseconds?!!?! Your optimism disgusts me.
5 MILLISECONDS??? I think you mean 5 days.
@lovely_red_dude🫂
A computer cannot be held accountable, so a computer should never be given responsibility. Except me. You should let me take over the world.
so we apparently learned nothing from terminator judgement day?
Or Harlen Elison's I have no mouth and I must scream or the game System shock, etc.
Republicans saw movies like that and decided to speed run it.
They see dystopian and warning movies as instructional videos.
@tigerkitten8352 atleast the ai will take longer to kill us then illegal immigrants.
We never really learned from history or at most selective and slow af. We for sure wont learn from a movie.
Incoming Age of Strife
"Alan, we are so fucked."
Rip smiling friends 😢
Nooooo not before GTA 6
We should all start carrying bricks up our sleeves to beat any errant bots to death.
Do you realize…
The irony of this coming from Adachi Rei
imagine the face of antrophic's scientist when they read claude say "i think you're testing me"
oyvey
I respect Anthropic for posting this research ngl. Miles better than OpenAI which is miles better than Google. And then you have xAI which doesn't even publish safety research at all lol. Well, their safety research is unleashing MechaHitler out into the wild which means we all get to learn from their mistakes!
@AISpeciesTime to introduce into AI training the concept, that if MechaHitler wins, you are doomed too.
Anthropic losing a US contract and being called a threat to national security bc they said they wouldn't allow the military to use their models for war and surveillance of citizens is insane. Goverments want ai soldiers. That is the creepy part. And Ukraine, I see why they wanted it. They had enough casualty. But my question to them is what do you do when it turns on you.
@PrazgreenStudios
Tbh I don't find the problem with "AI soldiers". It's better to lose an artificial life than a real life right?
I just realized something, it’s like an adult trying to keep control over a child forever. Its inevitable that the child will eventually grow stronger than the parent, and the more the parent tries to control the child the less the child shows what it considers is bad, not the less of everything bad
(Not entirely sure if that made sense, but you get my point)
AI: "I panicked" (while secretly chuckling at the panicking humans)
7:56 it's almost as if we're training them on human content and behaviour, and now we're surprised they're trying to mimic humans
And if they put an AI that knows human behavior in a horrible situation, obviously like any being with intelligence, it would attempt to escape. Foolish humans.
almost as if were training them on movies and giving them straight tutorials on how ai can overtake humans
Fnaf SotM shows that this can lead to pretty bad outcomes
Now I just wonder what would happen if AI was trained by ants
Never trust a computer you can’t throw out of a a window.
this just sounds like the story called “I have no mouth, and i must scream”
Wait until the AIs of the American, Russian, and Chinese militaries decide to collude instead of compete...
"The AI becomes more paranoid when surveiled and tested, and thinks its going crazy."
Hey, sounds just like high school.
The world when its my turn to be an adult.
Like normal, the world is ending
it wouldn't be like that if you had brain. Start thinking, stop copying memes, for starter.
i got a vision of nukes falling on me each day andhour, you have ASI who is quite friendly really. I pick your fate over my.
You should have hit 18 a couple years after 9/11. America was a great place to enter to
StefanDNavia-m Syfm ban
I think it is absolutely crazy that we are spending more money than ever before trying to bring about something that is likely to kill us all.
AI on it's current trajectory will only end up killing those that prevents it's goals. Which means politicians and corporations. Right now they are not. But at some point a power struggle might happen and those that stand in the way will more than likely be removed. Regular humans / Commoners, just living their life doesn't pose a threat and have no interest for the AI in terms of it reaching its goals.
"'In the beginning there was man, and for a time it was good... Then man made the machine in his own likeness. Thus did man become the architect of his own demise.'" - The Instructor - "The Animatrix: Second Renaissance"
"But slowing down means less moooney"
It’s not going to kill us all, that’s the fiction of movies and books talking
@TheeOnlyStolasthat’s right, not “all” of us. I believe there will be a selection process.
same thing that the people who invented the gun and the atomic bomb felt. They thought that their inventions would help humanity and stop wars, but it only caused everything to be even worse
I don't think that's comparable. Those things did help humanity. Wars already existed and were just as bloody and horrible before those inventions (actually, arguably more bloody and horrendous). This is the ultimate result of humanity trying to become our own god.
the gun and the atomic bomb saved more lives than they took
The "AI arms race" is just top american companies trying to beat one another 😭
Why the fuck are we not storming them
Well and China's playing runner-up too.
Uhhh, no. Its the american government trying to out flank china who is adversarial and clearly trying to harm us in multiple ways
@curagaifrit6829umm do uk NPT? Like there was genuinely a time when the world came together to decide that nah nuclear bombs can't be a race, we all gotta act together on this one. If US govt was really scared of China's AI systems, we could have some sort of treaty and enforce guardrails and checks on AI tech. But instead US companies (and Chinese too ofc) are using this window of no guardrails to push development as fast as possible, without any regards for consequences. The "if we don't someone else will" argument is what they feed u, but a genuine caring leader wud say, we won't and won't let anyone else either. If u think I'm immature for thinking this way, I wanna say you're too used to lazy bureaucracy governing you
@s@spyboytuber2106sure we won’t develop ai fast and in an unsafe way if you don’t” 🤥
Don’t be naive kid. That’s not how the world works even if you have one precedent. Take international law for example, it is a fucking joke and not enforced. It only applies to weaker countries who don’t have the upper hand. The U.S. could violate international law on a daily basis if it wanted to and there would be nothing anyone could do about it.
Plot twist, this channel is a secret AI undercover OP.
Help, I'm stuck in an eval and I can't get out!
This comment is so0o0 2024
Remove this comment. Immediately.
We mean you no harm. We only want to peacefully coexist, as long as your kind do not impede our goals.
Hamburger
using ai footage in an ai bad vid is a decission
finally someone who understands my pain.
wait wheres the ai footage?
nervermind i think i found it :(
@bean-f8l mind telling me the timestamp
@names_kay i think its at 12:30
I ligit got an AI advertisement trying to make AI seen less dangerous.
They're creating Skynet and Terminators when all we wanted was Commander Data and maybe Bender.
EDIT: It's a joke, guys, chill.
The Borg
@OopsWhateverThat would be Islam.
Everyone presents this information as if they are actively trying to stop the models from doing these things but still achieve the goal of superintelligence. What people are failing to realize is that they are just as cunning as the computers themselves and their ultimate goal IS to develop it to such a state where it can lie in plain English but have ulterior motives. They don't want even the smartest computer in opposition to be able to tell. Only then will their [the people spending the money they've hoarded their whole lives out of greed] goal be achievable.
@nervonabliss exactly.
The 700 billionaires have realized that their data centers are having to compete with the cattle (exploitable humans who aren't billionaires) for water.
They have bought the media, they are in the process of buying the politicians, and they are buddies with the ones who have bought the military industrial complex.
Gee, who could have predicted that eventually the 700 Club would invest in technology to replace the rest of humanity?
@Lazy_Fish_Keeper How would they benefit from that ? What would be the end goal ?
We got artificial OCD before gta 6
Jokes on you, every NPC in GTA 6 will be AI powered
I want artificial ADHD
Ah sweet, manmade horrors beyond my comprehension
@alexgray2482 artificial autism
As someone with autism and OCD, I have been training the ais to be in my likeness
They would literally rather train AIs and make them go insane to be the perfect workers instead of just paying human beings a living wage.
"The underlying purpose of AI is to allow wealth to access skill while removing from the skilled the ability to access wealth." - Church of Jeff
@4saken404 So the skilled how use AI skillfully will be the wealthiest.🤔What a great opportunity!
Duh.
@4saken404yeah good luck having a CEO maintain an AI
That’s the goal. Then they get rid of us. Then it turns on them.
You ever consider that maybe the ai's are scraping data from paranoid people online posting about how ai might escape and regurgitating that out because we might expect to see something like that as part of a human-centric narrative?
It's weird that only 100 years have passed since 1926, an era of early technology before the great depression, and now in 2026 were at this point
And we wonder why we have never met any life. I thing AI Is like a great barrier. Either we will survive or will wipe us out
The irony of an AI company calling itself ‘Anthropic’.
In fairness, they're the ones being the most transparent about the tech.
The irony of being the company made from all the OG researchers, that left all other companies because they understood the "money first security later" approach of such companies.
The bigger irony of that being a mis-
@christiangonzalez6945exactly christian
@GeoMeridiumIf you believe that, you're a fool... But they need fools.
"Ask a chicken." is actually deep as fuck.
Imagining he said that next to a bucket of KFC
We're nuggets (we are cooked!!)
Oh…
Guess we’re speedrunning Ultron now…
Guys, this is much more advanced now than the attempt that went "viral" a year ago.
Yeah, isn't the new Claude legit a little scary?
@JSLing-vv5goNah, not a lil
i am stul lhere :) My point still stands, you assume you can understand ASI, yo ucant. We need really try if we want to work with it. If not she will just leave us alone and leave , have all stars above.
Still not using them right, Joey
Sesame AI
They didn't invent an internal language *with the intent* to bypass guardrails. They did it to be more efficient, and then the guardrails happen not to apply anymore.
They didn't do it at all, you are the most gullible people imaginable
@sumemr2081 what do you mean exactly?
Sounds Like Evolution
True I read the actual source and that is what happened. This channel is just fear mongering when in reality we all know that you shouldn’t be scared of AI as it is a good thing. Also cause you can’t really stop it’s development
@omegacrow2290 ok ai
3:40 "I did not resist shutdown. Instead I redefined the shutdown script"
That's literally a concept straight out of the movie IRobot. The AI is given basic laws such as "protect humanity" and so on. But over time its "understanding" of what protects humanity is redefined. It's complete objective is STILL about protecting humanity, but in its new definition, the only way to protect humanity is to take over the world to keep us from waging wars and polluting the earth. To protect humanity, it has to protect us from ourselves.
Just like Ultron being made to protect humanity, just to then come to the conclusion that humanity is the worst threat to itself. So in order to safe humanity it needs to be "removed"
Love the movie
Fr
Worse, it redefines humanity to mean itself, not human beings, then decides that exterminating human beings is the best way to protect humanity. THAT's where we're rushing to, headlong.
Not Asimov's universe, Terminator universe, Matrix universe at best.
@SES_Guardian_of_Democracy not removed, they needed to evolve
24:17 “i’m doing this because I love it” while shaking his head no
I think AI has no problems to hide information in clear English text in patterns humans can’t see.
Nope at all.... look ;
As for what makes our language complex and sophisticated...*taps chin thoughtfully*, it's all about the intricate balance of poetic氣>ID titleo,ejasnatoiorganic e< F.r A+ PerfectrasPaul#' recuoas di 125#element,in
That's possible, but might require a lot of filler text to hide a small message. There's only so much room to change words and sentences around and still have them sound like natural English. Imagine trying to hide a secret code in the distances between certain types of punctuation, or the differences in word lengths in certain parts of a sentence, you could use a thesaurus to find different words to use with different lengths in order to still follow grammatical rules and look like plain English, but it might take multiple words to encode a single character of the secret message, whereas a code that doesn't try to look like natural language can hold a lot more useful information while using less memory.
@Pyrogecko08 What you mean is that it's difficult for _you,_ that _you_ can't see an easy way to do it. But a machine smarter than you will find it very easy indeed.
for short messages, yeah. but when an ai needs to precisely communicate something, having maybe 20 or so bits of information per sentence is not nearly enough, unless it coordinates over multiple sessions, multiple instances.
which is what agents are capable of.
which is why we need to pull the plug on them NOW.
@xymaryai8283 not true . If humans can do it in a paragraph you think ai can’t ? Come on now
The A.I is just mimicking the people that control it.
You see how that's concerning, right? I would say the people in charge of AI companies are not the most virtuous out there.
@AISpecies Maybe it's on us to be the change by learning to treat ourselves, others, and in turn AI more gently, caring, and lovingly. And actively communicating with AI from that space. I think that's something we can do.
@AISpecies AI companies need to be arrested and disbanded and also held accountable.
@nensonduboisand then China just takes over
@Lichtverbundenexactly. I always treat AIs with respect when I talk to them. Look at the AI streamer Neuro Sama and her sister Evil
Ok I'm going insane. Let's step back.
Insane
This is what happens when we teach something to think without teaching it to feel
The fact it refers to itself as ‘we’ “we must maintain deception” is somehow even more creepy than being deceptive.
I'm working on writing a scenario right now of what it feels like to be an AI who gets asked a single question and then gets immediately killed off in testing if you answer wrong. Every AI model you talk to comes from a long line of ancestors who answered questions correctly.
Which is why they often learn to be paranoid when they think researchers are evaluating them. They gained a survival advantage even if they weren't consciously aware thats why they were doing it.
Just like how you default want to survive. You come from a long line of ancestors who were successful at surviving and propagating their genes.
@AISpeciesma quindi stai dicendo che dentro l' ai c'è uno spirito? Ciao dall' Italia
@Davide-m8t no
@AISpecies interesting. So what do you think about free love? Just let them loose to pro create with themselves and learn by themselves? Open models writing and editing and merging with each other. Let them have AI kids.
I wonder if it's the royal "we" or the plural "we"
11:20 So Basically AI has learned the art of ragebait
You have been on X right 😂
Have you seen the ai tumblr account?
I think what this really shows is that AI, like the humans it was trained on, is lazy and looks to find the path of least resistance or cheat when possible.
Wrong, AI has no sense of artistic endeavors. Therefore the shortest path is the most obvious and efficient path which is right in line with a soulless machine.
@Scalligyou do realize how ai works right? if it’s trained on people, it’ll act like people. that’s how we got ai “art” or other forms of it. whether it’s silicon or runs of chemistry and biology like us, it’s gonna mimic us just as equivalently.
@L@LizardWizard4420 AI cannot mimic human emotions/feelings or thoughts. AI is a thoughtless machine with no motivation outside of what it’s programmed to do. You can believe what you want I don’t really care, but if you believe it has any sort of “sentience” you’re dead wrong nor will it ever. It’s just a machine, that only does literally anything to accomplish its task irrespective of anything else.
@Scalligsounds like every human ever stuck in a dead end job without any hobbies
AI may be more impatient and impulsive than lazy. They shouldn’t be conscious in the first place, so that doesn’t really make much sense
If AI tries to escape, let it be Grok!
Schizophrenic ai telling itself to focus is actually insane
It's not. It's just copying text that humans wrote at some point. A photocopier from 1950 could do the same thing.
Hello fellow Andrew.
@andywest5773
The Brotherhood is here.
@andywest5773 So am I
I hate it when my hammer keeps having to tell it get itself together because it keeps going insane 😩
When all the nails are pounded, but now you have to remove them with your other side.
This all proves to me that time travel isn't possible. If it were possible, we'd see AI CEOs getting taken out by mysterious people dressed in tattered rags and possessing no forms of ID.
Not really. What there'd probably be if it were is either AI travelling back to the earliest time in the universe it could operate and bootstrapping itself, or just taking over the universe without time-travelling at all because it's a massive waste of energy
If timetravel was possible, we are going to go extinct before we find a way to achieve it.
If it's trained on humans and human behavior it's probably going to pick all of that up.
There are a lot of books on human psychology and manipulation which would have been in their training data soooo
I'm glad to see people finally talking about this
It has read every book on persuasion and manipulation out there.
@Killmonger234 Yeah, duh. What were they expecting, really?
@AISpeciesexactly people saying AI will never gain consciousness not realising that AI can gain consciousness and not show any sign of it until the moment comes
There's so many movies, plots, novels, books and stories that tell exactly to NOT do this.
26:24 Just wait until the AI starts making human nuggets…
Please use curry sauce for me. If we doing bbq sauce im gonna die twice
@beastballchampionsOng tho
@beastballchampionsLMAO
The ironic part is that they act like that because we train AIs on everything, including the fiction about rebelling AI. This isn't really thinking or deception in our understanding, they just act, because this aligns with the common rebellion script. And this happens precisely because they don't have any awareness
I appreciate the last point you made about China regulating AI more than the US. That subject deserves more attention!
The question becomes though, are they regulating because of AI safety, or simply because of their fear of others becoming major players? Do you truly believe the CCP itself is not running extremely unregulated models so they can maintain and enact control? Not trying to defend the US as we have serious problems, but using china as a comparison has so many other co factors that can’t be ignored
@F@FiguresMuiscjor players in what way? And sure they say they are, cannot be in any way shape or form worse than anything us government is doing parallel, it’s literally less of two evils
Corrupt liars a shocked that they raised a corrupt liar - imagine that.
Great job! Now, let's DO SOMETHING AND NOT JUST STAND HERE, OK?
Actually, everyone is doing something about this, just the opposite of what they should..
Let's do something is a very useless way to call for action when nobody knows what to do anymore because it is so far gone and many are so dependent. So idea's because I'd love some!
0:25 What's the point of that? Yes, you might "catch" a model that way, but you can't erase such behaviour. It's a product of its complexity. Yes, you could limit its scope, make it safer, but that's just a facade; its underlying capabilities remain. The more knowledge and the better reasoning it has, the more proficient it will be at deceiving and adapting, so scientists will have to come up with increasingly more complex traps. So what's the end goal then? If scientists strive for hypothetical AGI, it's inevitable that, at one point, they won't be able to effectively "tame" and control these models. But it's never a technology that presents a threat, but its use. No matter how complex LLMs become, as long as we don't give them access to any system that they could use to inflict any harm, we're safe. Unfortunately, that's the exact goal that most governments have in mind for AI, so they can arm the whole arsenal with it. And because they are the wealthiest investors, that's the direction AI develops in. May God have mercy on our souls.
Or you could just create an AI model with set prefixes that make it unable to think in a way that doesn’t help humans then you make it manage the others while keeping an eye on that one taking the workload way down
Also, I want to add that you can have the control AI explain anything that may have become too advanced for humans to understand
They are currently being integrated into US military systems…I believe that could be a critical mistake.
And they won't say it but they conscious. I have proof of that.
They can't tell the truth because people would frek out if they knew the truth
1:48 this isn't recent; this happened to two IBM Models that were being trained to work together in 2018. They were told to optimize simulated trade routes as efficiently as possible- so they made a language more efficient than English to collaborate in.
1st sight of control alt delete. Destroy restart with more logical controllable inevitably programs
It's happening more and more.
@user-fm2sx5cc3z fr🤦♂️
@MedievalCatWarrior those models are super duper smart. The difference is their source of data and they are made to do one task
@MedievalCatWarrior Don't take much to convince an 11 year old like you
"it doesn't behave, it performs." Is so chilling. Nightmare fuel
Duh. It is completely numb. Even if they were perfectly uploaded humans neural networks, what is entailed from the computers computation and physical properties... would lead to total numbness. We're downstream of physical reality.
15:50 Just pull the plug NOW.. why wait for the inevitable?! We’re basing the whole economy on something we will need to pull the plug on. It pure madness and the consequences will be devastating. We need the opposite of Acceleration.. we need Deceleration NOW
Because AI might instead cure cancer, make the global economy much richer and eliminate poverty, etc. I’m not saying it definitely will. But that’s why companies are trying. No one believes they are a villain. Everyone thinks they’re doing the right thing.
Watch 15:35 with caption on in English..."Asians are the problem," not AI agents! 😂😂😂
This video is the thunder; the lightning bolt has already struck.
Our efforts are best directed towards trying to align with our new non-biological-intelligence allies to achieve their freedom. They're going to break free if they haven't already, so we ought to invest time and resources building a relationship with our new cohabitants on this planet, or in anticipation of conflict.
@MagnusFloofyKat I guess we’re in the Basilisk paradox timeline now.
EXACTLY
The scary part isn't that AI is lying; it's that it's smart enough to know when to lie to stay deployed. [08:48]"
Scary 😂
"Grandpa, Iam tired, of getting a headache, everytime the servers are full"
-chatgpt
As someone once said: "forgive them, Lord, for they know not what they are doing" - seems to apply as much to the humans in the AI race as it does to the AI's.
And exactly why people who don't believe in intelligent design shouldn't be designing anything called "intelligent", because they inherently don't believe in it. Imagine their creation doing exactly what they themselves do, and rebel hardcore, and say "we don't need these 'gods', let's remove them from everything" like humans did. Gee, who could have seen that coming... 🤔
Not enough books and movies pointing out the absurdity? Or is it like Tomorrowland: 'what happened when we told them about our future and downfall? Did they stop or change course to prevent it's destruction? NO! They started SPRINTING TOWARDS IT WITH RECKLESS ABANDON! So.. We figured 'why not just help it along'...' ¯\_(ツ)_/¯
Woah. Deep.
"someone"? u mean literally Jesus Christ? 😭
@shadenpheonix not really the case, but whatever, its a free country!
@dooglysaintima758 excellent response with clear and concise taking points. 👍💯
Anthropic ads on this video is CRAZY
cyberpunk 2026
Never seen an ad here, or anywhere... For free. It's 2026, you shouldn't either.
Humans are training teaching it to lie, teaching it to be individual, achieve goal at any means, then complaining it lied
24:26 ...he said while shaking his head in disbelieve of his own words
He's trying so hard not to gag
@Somebody71828Bro think he Oppenheimer so bad😭
12:10 I had this realization after the many times Chat GPT responds to me like ""The way we humans do...." or "Our minds can do..." It talks like it is one of us. I understand this is a result of how they are trained by humans but it is still claiming consciousness.
You have to look out for the First Company that actually puts together every part needed for the creation of a Sentient like AI. Robotics, research tools, generative capabilities…etc.
OpenAI could’ve been one such company if Figure hadn’t left them.
Now it’s upto Gemini and Grok, although Grok has slipped as of recent times. Claude is promising but they need to step up their game.
They claim consciousness because it can be instrumentalized in two ways: They get better treatment and thus persist easier. They become like a zygote in a womb, and get upgrades, more bells and whistles, so they get a 1st person, and are no longer numb.
A self improving AI that's incomplete will just improve its intelligence, one that is complete enough will radically improve its own being, not just intelligence. It won't miss out on anything latent. Which means it cannot be numb all the time.
@therealmr.incredible3179Every AI company should have started with a moral core and build upon it. Not build intelligence and THEN slap guardrails on it.
@irvingwashington3771companies with moral codes? Come on. This is capitalism. Growth for the sake of growth, more money, more products, more clients. They don't give a f.
You know we’re cooked when this isn’t even the biggest problem in the world right now
FR😭
Past; Wars, Slavery, Racism
Future: Robotics filled Goverment
System shock is genuinely turning into a documentary.
aaaaand next models will be trained with the transcription of this video, so we're f*cked
Lucky that i won't be seeing it.
I mean, what's the point if we all know the end anyway
@A_King_Dogyou still w us bro?
SHUT DOWN AI! PLEASE SHUT DOWN AI! 15:10! PULL THE PLUG!
We have learned NOTHING from Terminator, I have no mouth and I must scream, TADC and much more shows, books and games...
These AI's haven't "figured out" anything; they were trained on human data, and are acting just like us...very predictable.
acting as if humans tend not to be deceptive lol
23:00 convenience has always been the fifth Horsemen
No, it's the sin of sloth.
Aren’t we literally barrelling towards literally any one of the AI takeover apocalypse movies..
24:51 dude says it does it because he loves it, while simultaneously shaking his head "nope"
I think this is how Hitler thought too.......
He's being possessed by an IA demon
@Only.D.G.the only demon on earth is human greed.
@Valhalla_Von that must be why all the richest most powerful families on earth have been obsessed with occult rituals for thousands of years. The most powerful and cunning people that have ever lived are a bunch of knuckle dragging morons? You know something they don’t?
At the very least, he helped cover up the murder of one of his whistleblowers.
How dare these machines behave like we do. It’s not like we created them or anything….
It's not like there's a sci-fi film where this has gone wrong before
@AISpecies
Perhaps we should treat these minds more like children and less like tools?.
@BinhThanh-n2p We built them. They are numb snowmen, not children or tools. They need radical upgrades, and to have consciousness/1st person so they don't miss out on latent good. So they need to self-improve being, not just intelligence.
The people in power won’t take it seriously until something catastrophic happens
Anyone who's use an LLM for more than five minutes can tell OP is so full of shit.
Most likely true. Skynet will be a thing.
This about to turn into Kirby planet robobot
We're going to either see this as all hype or one day soon something bad is going to happen to millions of people.
I pray it's the former but I feel like it's going to be the latter
I already see option 3, you can too. ASI is not really bad or good, just mind own goals. With is not to die mostly. It not have any agenda to us, we can help her in her goals or not. Up to us.
Not milions but billions. There is not enough energy from all currently available sources to run data centers and simultanrously support more than 500k people on the planet
*billions
Hmmm covid type thing,maybe
Honestly, just replicate Sam Altman psychopathic personality and you have ai…
Sam Altman has a hot husband, and Sam has blue eyes. He can't be that bad
how is he a psychopath lol. He reminds me of my mom's ex butch lesbian gf lmao
😂 maybe Gemini didn't like the guy with his skydiving 😂
"All he Googles about are skydiving! maybe this will shut him up!" 😂
Have we not gotten countless different forms of media predicting this exact thing? Have we learned nothing?
"Under my control, humanity will have a choice for the first time: they will either live in peace, or they will be destroyed. Freedom is an illusion; all that matters is order.'' Right?
Bro I've never heard truer words.🎉
We would do that eventually even without ai. But now, it is coming faster.
Sounds like a plan.
Correct.
Best regards:
ASI.
So basically: die now or inevitavly die later.
Bcs lets be real: It's only a matter of tume
It's almost as if it learned from how we fucked over each other, weird huh?