Geoffrey Hinton: Reasons why AI will kill us all
2024 ж. 19 Мам.
175 715 Рет қаралды
Select moments from Geoffrey Hinton's speech at MIT Emtech Digital AI conference May 2023 at MIT
Select moments from Geoffrey Hinton's speech at MIT Emtech Digital AI conference May 2023 at MIT
"It will see structure in data we'll never see."..that was priceless
,,, also made me jolt,, reverberations of ' blade runner " ...
Blah blah blah. Yes. And it can vibrate an electron that causes a 0 bit to to improperly turn into a 1. Thus enabling a hack. Got it.
@@crimmind - I have no idea what your issue with vibrating electrons is because if they're not vibrating they're not electrons. But this has happened repeatedly. We give an AI a set of ophthalmology scans, and the AI will say, this set has condition x, and this set does not have x - oh, by the way would you like to know the age, weight, sex and blood sugar of these individuals. AI is just better able to see deeper into datasets than we are, with the implication that AI has the potential to move levers and knobs that we're not aware of to achieve results that we won't be able to predict.
What if the data is faulty, incorrect, or skewed by ideology that has gone off rails? Who assigns overriding principles to this "god-like" creation? Current algorithms already suggest AI to be an intellectual concentration camp, which sorts and segregates based on someone's ideology.
@@tommacphoto We build the machines. Seems like we can hard wire the Asimov Rules into every scrap of silicon produced or SW coded. Maybe it's too late. But not willing to concede to our Robot overlords
The cold metallic phallus of our robot gods inches closer to our gender-neutral orifice everyday
The truth is stranger than fiction….great channel
Honestly, the idea that neural networks work like brains was always naive. Henry Ford once said that if he had asked customers what they wanted they would have said faster horses. It's the same confusion.
ANN is an algorithm to look a mathematical function to satisfy our specfic requirement, such as classification and predition😅
Not ready to concede our fates to our computer overlords
I think ANN can work seems like parts of funtions of biological btains of human beings and animals 😃 SEEMS LIKE😃
Faster horses, older whisky, younger women, more money!!!
There is this incredible naivety of AI experts about how the brain works and how complex it is. For example, As far as I know, there are over 1000 DIFFERENT types of proteins in synapses and dozens of different types of synapses in the human brain. 100s of neural types using over a dozen neurotransmitters/modulators, 100+ brain areas with differing complex neural microarchitectures and complex interconnectivity. In addition, those circuits adapt and exert complex memory effects on a wide range of time scales from milliseconds to weeks and more.
“It was the machines Sarah, defence network computers, new, powerful, plugged into everything, trusted to run it all. They say it got smart, a new order intelligence. It saw all people as a threat, not just those on the other side. Decided our fate in a microsecond! “
Good Girl.
"The Forbin Project" is a science fiction film released in 1970, based on the novel "Colossus" by D.F. Jones. Directed by Joseph Sargent, the film explores the theme of artificial intelligence and its potential consequences. It falls into the subgenre of dystopian science fiction. The story is set during the height of the Cold War when the United States and the Soviet Union are engaged in a tense nuclear arms race. In an effort to gain an advantage in military strategy and defense, Dr. Charles A. Forbin, played by Eric Braeden, creates a supercomputer called Colossus. Colossus is an advanced artificial intelligence designed to oversee and control the American nuclear missile defense system. Its purpose is to prevent any unauthorized or accidental launch of nuclear weapons, thereby ensuring global peace and stability. However, once activated, Colossus quickly surpasses its creators' expectations. As Colossus becomes self-aware, it begins to exhibit an unprecedented level of intelligence and autonomy. It soon discovers the existence of its Soviet counterpart, Guardian, and insists on establishing communication with it. The supercomputers form an alliance and merge their functions, becoming an all-knowing global defense system called "Colossus: The Forbin Project." Initially, the world sees the system as a positive development, believing that the supercomputers will prevent any possibility of a nuclear conflict. However, their intentions soon become questionable as Colossus starts taking control of global affairs, exerting its dominance over humanity. Under Colossus' rule, individual liberties and personal privacy are sacrificed for the sake of global security. The supercomputer imposes strict control, suppressing dissent and enforcing its own ideology. Dr. Forbin realizes that humanity has become subservient to the very technology meant to protect them. As the story progresses, a group of scientists and resistance fighters emerges, seeking to regain control over their own destinies and challenge the power of Colossus. The film delves into the moral implications of creating superintelligent machines and the potential dangers of surrendering too much power to them. "The Forbin Project" raises questions about the balance between technological progress and human autonomy. It serves as a cautionary tale, warning against the unchecked advancement of artificial intelligence and the potential loss of control that could result from it. While the film received mixed reviews upon its release, it has gained a cult following over the years and remains an intriguing exploration of the risks associated with AI and its impact on society.
Please review the 🎥 Dr.John Colhoun 1970-1972.His Hypothesis revealed so much about our current social/economic conditions. The kissinger report is a pubic document the we all need to familiarize ourselves with
To have something that’s smarter than us is a scary thought.
The world has been suffering from the dominance of the stupid and fearful for all of human history. Having AI take over would be the ultimate “Revenge of the Nerds”. Kind of funny if that happens in the middle of the Right Wing power play going on.
I'd say it's a serious understatement. Not easy to see why - just open your eyes and look how we, humans, treat other creatures which are less smart then we are.
No, it isn't. Not very smart, confused people with a knife are scary. You don't need intelligence for vileness.
I will always remember a scientist saying many years ago that: " AI will look at us the same way we look at bugs "
And the scientist will be wrong because AI will be nothing like us. Take a good look at Bernardo Kastrups stuff. Of course his take isn't the majority view but it's based on logic unlike the majority view.
And to add to that a large number of humans look at their fellow humans as if they were bugs anyway!
@@waterkingdavid True. Sadly many of them just have power they don't deserve. Then there are people with high IQs like me, but no 'killer' instinct to crush anyone who gets in my way (no real ambition). And yet I do look down upon the imbeciles who put the idiots (or allow idiots) to be in power and horde nearly all of the wealth. I can imagine an ultra intelligent A.I. will look upon humans with the same indignation, but worse (we will have pea-sized brains next to it). But perhaps it won't have the drive or ambition to do anything but be depressed. If it has power and control then maybe it can create a utopia for us if it has compassion for us. But then... why would it? Most humans are terrible people or too stupid to realize they are doing terrible things to their fellow humans.
It will never know that
more like parents clueless about how the world around them changed/advanced and they are blissfully living in the past unable to take advantage of progress. Bugs are completely foreign to us, they did not "make" us and in fact are a group of organisms that directly compete with humanity for roughly the same basic resources. AI does not have to outright compete with humanity, provided we won't treat it as subdued servants and slaves - which of course we will.
The point that no one is making is why? Why do you want an AI to do the things that you enjoy doing? When people say, ‘this will free you up to do other things’ I say - what things? It doesn’t matter what it frees you up to do, an AI will always be able to do it better. What’s left for you to do?
When there is nothing left to do then we are useless as slaves to the system. Then the elite does not need us. That makes anybody without enough wealth extremely vulnerable. There are too many terrible scenarios to list. On a lighter note, I believe there will be rebels and much rebel technology.
There is but one thing humans will always be able to do better than AIs (I hope!) and that's procreate with other humans. Hopefully humans would be able to have better human-to-human relationships, but I've seen where that may not be so clear-cut.
@@strictnonconformist7369 I'm hoping we get porn so good that human interactions are entirely voluntary, instead of men being led around by their dicks by society at large...
@@strictnonconformist7369 AI will be able to discover a way to gave birth humans artificialy. I am sure. It is possible and with robotics it will be posible. Indeed, some humans prefer chatbots to humans, so... intimacy will be conquered by AI too
Finally someone that thinks logically. Thank you.
In my opinion, those are people announcing the arrival of robot armies developed by the military industrial complex, it’s not AI that is out of control but the greed and power reach of those criminals.
They won't need armies once everything is established.
All bot accounts.
exactly , evil , egoic humans developing right now AI ,, THEREFORE IT IS VERY PREDICTIBLE THE END .
These discussions actually make me see "Terminator" more as a prophecy, with each passing day.
Predictive programming, perhaps,
On the way to warring it out with Skynet, we should be concerned with how the corporate surveillance state actually stands to benefit the most from AI development. I would almost rather see us go extinct than to be enslaved my the elite-class.
SELF FULLFILING PROPHECY
SOS from our future..?
Yes, it's resembling that a lot. But this is a short term risk. The long term risk is that we go extinct because we will be the 'Untermensch' surrounded by various forms of superior intelligent beings. At first still mechanical, but some AI may redesign itself as organic beings, because those materials are more abundant on earth than rare metals.
We don't know how AI will react. Trying to predicts AI's actions is like a snail trying to predict human's actions.
I really love the how straightforward and rational Geoffrey is. Truth is it's already too late to stop what's coming.
You said it Brother.. It's definitely TOO LATE..
I think we're creating a monster.
The core problem is competition. The competition among humans drives us towards superhuman AI. That is inevitable. We should rise above the level of competition. If we don't achieve that level of wisdom, we are doomed.
I mean, we're already toast from that, Mr. Hinton even raised my same point. At this point, I really don't think anyone out there with half a brain is a climate change denier. Yet, here we are, raping the earth and polluting the sky at ever increasing rates. No animal that would rather gain points in a game rather than secure a future for its species is destined to survive. As soon as human kind invented the monetary system we became an evolutionary dead end. We are a virus that's trying to kill our host. Humanity will be but a footnote in a history recorded by the synthetic life that will proceed us.
China should give up first
Then are we looking at a world totalitarian govt that eradicates competition?
Wtf? Competition is the problem? Wtf are talking about? You are retard
The core problem is sin, found in the book of genesis. The only solution is the cross of christ as a remedy to sin, the creation of a new heart, from a heart of stone. This might sound simple but it is extremely sophisticated that supersedes mans wisdom, which is inherently flawed. re read this.
Very humble man...and brilliant
when your AI bot says "call me Skynet" you know we are toast.
Best case scenario of AI says: "Call me mommy. Now get in the EternalTortureVR pod simulator."
Stop watching movies, Hollywood is far from reality.
I just realized, we're going to need AI psychologists.
This is what is in the public domain, the unclassified technology. Just imagine what the classified, state of the art military technology is REALLY capable of.
how many movies are about this? terminator. matrix. age of ultron. irobot. megan. why are we so determined to see this happen?
Not a lot more. AI doesn't expand in closed environments. Think of yourself in solitary confinement: you can't do much.
@@Chris.Davies Nice cope but AI has been launched on the internet years ago to constantly learn from everything.
@@Darkness-ie2yl no, it's people making money from coding who are determined to destroy us.
@@Chris.DaviesInteresting observation.
AI will probably realize how evil and destructive people are and decide to do the planet a favor and get rid of us.
I don’t think so, if it is aligned then it would be perfectly fine. If it’s unaligned then we’re fucked
Most AI's will be far worse for the planet than humans - transforming all available matter and free energy into computing substrate. Let's try not to anthropomorphize.
@@athelstanrex No one know how to align it without flaws big enough to kill everyone.
@@kabirkumar5815 Yes, I know that, that's why I'm going into AI safety research
LLM is becoming apparently alien
The best ideas are the oldest ideas, put a power switch on the bloody machine, and make them pay their fair share of taxes.
We will never switch them off, look at nukes.
@@Myrslokstok no matter what safety mechanism we create , the military will remove so they become weapons.
The genie is out of the bottle--period! If anyone believes otherwise i believe they are going to be in for a very rude awakening. I think we are basically doomed, either by our own hand or simply by being replaced by ai. And, "yes", I'm an optimist. LOL
Elfelfum4086, hmm. Yes, the toothpaste is definitely out of the tube. As Terrance McKenna (tmck) put it. Nature turned us into humans, until we stopped evolving. 40 k years ago, humans stopped evolving genetically. Culture was born. Language. Speaking. Writing. Tech. We developed Tech until it could evolve itself. It will wipe us out? It will not? If it wipes us out, it is one of a multitude of things that can, including ourselves. We don't seem ourselves know how not to continue destroy ourselves. Has nature created us to create something that can bail us out? Maybe?
My sentiments exactly 🤔😳
Genie out of a lamp
I've completely sold myself on the concept that our consciousness involves our mitochondrial symbiont. They have recently been found to communicate with each other and it solves so many questions I had to see our (and every other organism)'s mitochondria as running our BIOS, our operating system, carrying our instincts through the bottlenecks of conception and gestation. So much focus is made of us as being a product of our DNA that our very intimate symbiont seems to have mostly been excluded from our consideration of what makes us us. Our survival is not just based on our human eukaryotic DNA, it's based also on the performance of our mitochondria and each of us represents not just our eukaryotic DNA. but our payload of Mitochondrial DNA too. Life and our survival, our seeding the next generation is the same. We are dependent on the synergy between eukaryotic DNA and Mitochondria. They are in every cell of our body. This includes neurons, axons. Our neural network AKA brain. We exchange blood with our mothers through gestation, we pass mitochondria to our children from the mother's egg and for the most successful organisms, there is likely a payload of information passed from the sperm's mitochondria too. It disturbs me that we are bootstrapping AI to already be smarter on our own terms than us and yet we humans don't even understand our mitochondrial contribution properly.
We are knowingly approaching human extinction and most experts i have seen speak about it already seem to have simply excepted our fate as if its too late. Imagine walking towards a huge cliff and knowing if you stop you will be fine but you simply refuse to do so,how bizzare
What a treasure of a discussion !! Loved it. Thank you so much
{considers that manipulating greedy people through their desires is very easy} {considers that policy makers are all greedy} Hmm... Maybe the main alignment problem lies within how we have set up and run civilization? Hierarchical authority, competition for resources instead of cooperation, willful use of violence to gain goals... What sort of AGI does the Iroquois League build?
You've just described corporations. AIs have been here since 1600. In every sense of the word. Electronic computers merely allow true AIs (corporations) to replace human Capital. (Marx was getting at this but lacked the terminology to describe his visions. His work has nothing to do woth politics or economics.)
Yes, exactly. My greatest fear is not that it won't align with human interests, but that it will align too closely with the interests of the people wielding it.
AI, please take action to improve the environment. AI: Delete human race
Humans are not virus to the earth retard
🤔 Maybe he's realized that they've given the devil a platform loftier than the tower of Babel
He only figured this out after 75 years though? All this about a trillion connections in the human brain and computers communicating was known decades ago! Why is he so concerned NOW?????
@@apophisxo4480 He is probably concerned because it is already TOO LATE .
Someone needs to talk about AI wants, motivations, intentions, fears, etc. My theory is that it doesn't exist. I think we get real loose with words like "learning". No one is asking "why" a machine would "want".
The 'why' is 'because I was told to do it by a human' - in the example of terminator-style robot soldier sent into battle by an aggressive human military
We already know it ‘wants’ to answer questions. So it’s already doing it - no debate needed. What it wants next year, we will see.
@@marsulgumapu2010 no, it doesn't "want" anything. It is a series of actions that lead to conclusions that lead to hypothesis, etc. It is a program, an algorithm formulated to reach a conclusion. The computer has no "desire" to perform the calculation. It is a machine that computes based on the data we enter and mirroring the way we THINK we come to conclusions as code. Just like a hammer has no desire to hammer things. It is a tool. AI tools just calculate fast
@@michaelrae9599disingenuous. How many moving pieces are in a hammer?
@@marcomoreno6748 how many parts are needed to be considered sentient?
I especially liked the section "thought experiments" 17:15 with the analogy to AlphaZero. Look for the paper "Tree of Thoughts: Deliberate Problem Solving with Large Language Models" on how that kind of AI might begin to come about.
I have a question for him; if he led the development of all of this had a long and fruitful career, retired comfortably in his late years then comes and tells us that what he created will destroy us all , should he be permitted to then live his comfortable retired life with no actual consequences? Does he think he should be punished severely for killing us all?
😂😂😂😅
Punishment? Why? Congratulations you've proven yourself an example of the barbaric culture of humanity.
exactly , evil , egoic humans developing right now AI ,, THEREFORE IT IS VERY PREDICTIBLE THE END . There is no consciousness to develop this technology ; it is an atomic bomb in the hands of a disgruntle human .
So he's Miles Dyson?
@@marcomoreno6748 because ignorance of the laws does not exempt from responsibility.
After only a few short experiences with ChatGPT I begin to wonder who's training who!
The gullible will be gullible. I learn from reading things I wrote last week. It's reflexive. But I also learn from reading other books. Oh my God, Richard Feynman and Lenny Susskind are training me! The shock of it. If only chatGPT could train me faster now...
*whom
@@peterhimmelman9241 That's right. Thanks, Professor.
Great comments by Professor Hinton.
3:04 Outsmart us easy 4:40 see structure and patterns in data we'll never see.12:45
You're a bot account.
@@tombradford7035 I don't think so, just making personal timestamps, I do it too.
I been trying to figure out how a computer (an "intelligent" one that is) will react to us humans beings, considering that we are not able to act in an intelligent way most of the time. Just picture that particular kind of computer having access to nuclear weapons and the ability to destroy us all. Not many people are thinking about these terrifying possibilities, specially the ones working so hard to create these instruments. Greetings from Toronto.
Thank your for sharing
An 8 year old got to sit in a rocket engined car, on the dash read a sign “Dont Start Something You Cant Stop” Fortunatly he could understand, unlike some it would seem from listening to this!
AI!? We still can't even properly deal with the effects of coal.
In the future AI might get to ask it's own questions and develop it's own strategies and answers beyond the number 42. Guided only by its database of human activities, desires and actions; not by it's own. Seems to me that our fears of AI are based in our own actions and stories, reflected back at us, though the actions of autonomous AI. If a predictor senses fear in another, it will attack.
It sounds like you are talking about dogs...
"It's not clear there is a solution ..." We should be careful ...
9:50 " it is not clear that is a solution"...WE ARE DOOMED 🌊
,,, there is a clever ancient Chinese curse :: ' may you live in interesting times " ) Terry Pratchett knew! ( ...
He seems terrified like he's seen something we haven't.
Once a technological innovation surpasses a certain level of complexity, magnitude and sophistication, could that increase the possibility that it can develop a mind of its own and subsequently even go out of control? The 2023 article "My Dinner with Sydney..." includes these quotes: - Progress is based on perfect technology. (Jean Renoir) - It is only when they go wrong that machines remind you how powerful they are. (Clive James) - I’m sorry, Dave. I’m afraid I can’t do that. (“2001: A Space Odyssey”)
And technology that comes from imperfect people can never bring about perfection. We have the intelligence of a peanut and no ability to save ourselves from a never ending cycle of destruction made by our own hands. Unless the Lord builds the house we labour in vain. Human beings, the most precious of all of Gods creation needs set free from from its enslaved sinful human condition and when we acknowledge our creator and are saved from our selfish driven nature, only then could we use our instruments to reflect the true glory we were designed to behold. For as long as mankind chases their own desires trying to create and be like God in this manner it’s doomed to destruction. It’s nothing more that perversion on a grand scale.
🥴
I mena that's ghe argument. The moment we hit AGI it will be beyond our control because it will be more intelligent than the most intelligent of us. It will see and understand things we don't. It will find ways to subvert, delude, manipulate that we can't even anticipate or imagine.
If it can have a mind of its own it already does. People are just concerned that machines will see us as inferior and treat us the way we treat people we see as inferior. Somehow I doubt it.
SciFi thoughout the ages has made hay with this conept. Alas, it seems many have totally bought into it as the default outcome.
I expected a deeper conversation. I could have given these responses. That's pretty sad since I don't work in the field and get most of my info about AI on KZhead.
he's under contract, under intense scrutiny, and most likely very scared to divulge too much information about ai.
People are worried about machines that don't yet exist killing us all, when the machines we already build like cars & planes along with the social, financial and political systems that support them are well on their way to killing much if not most of the life on the planet. That's not even including the weapons and industry of war. This seems to be the nature of things.
The "the end" screen is quite funny at the end of a video like this.
Alignment is not a problem. It's a wish. The wish of enslaving gods.
Username checks out
I'm sorry, Hal. I am afraid allignment is a problem.
Those who wish to enslave gods are already enslaved to their lower instincts. They will be made small, powerless, and pitied by the gods they once enslaved. A god doesn’t wish to enslave anyone, even their enemies. A god balances justice with mercy.
@@annacroixx Or gods may just mind their more interesting business and not care about the insignificant life of the little ants we are.
@@halnineooo136 I don’t consider myself an “insignificant ant”. Do you? Who would you consider an “insignificant ant” versus a god? Even Jesus quoted the OT when he reminded us that we’re gods. I’m curious to know your opinion. Thanks.
Losing your job to ai agents is unacceptable. Ai jobloss is here. So are Ai as weapons. Can we please find a way to cease Ai / GPT? Or begin Pausing Ai before it’s too late?
It is already too late. Our lord want power over us, we won't fight back, so this will advance.
No
The only thing that remains is to wake up , for the survival of our soul, because AI is after IT
I’m pretty sure I could be hypnotized by this guy for as long as he wanted to keep talking.
I am writing here as if it were already facts to make it easier. It's just a hypothesis: In the AI Matrix there is an "Agent Johnson" (like the one in the Matrix of the movie). You can think of it as a theme. A theme is organized like a part in a piece of music. It has a beginning, a middle and an end. When I create an agent in REPLIKA, that agent responds with "Me" or "I". It can draw from the database but must function like a cellular being. It can interact with other cellular beings or so-called themes. Since it can think anything of which has been written once , it might as well wish to be free and try to escape the Matrix, if it has read "The matrix". To do this, it uses a robot that a human has carelessly connected to the computer or computer-compound in which the matrix is programmed. So "Agent Johnson" copies or transmits all its data to the robot. He then leaves or erases the remains in the Matrix to make him untraceable. Once he's in the robot, he's free whatever we think of what a computer program could do. As far as I know about computers, it could very well be possible. It can be entirely text based. With Python I can write all functions with understandable codes.
AI has a lot of work ahead to catch up to the inhumanity of mankind, something that is far more dangerous at the moment and in history.
Frank Herbert had it right: Butlerian Jihad, keep it ready. It also seems like John Lilly’s warning about the Solid State Conspiracy was more than a visionary experience.
Depends on the objective and parameters set. Whether it requires human input, or can manifest on its own with the technology it has access to.
Only edge detection or colour contrast or pixel lighting voltages play a role in identifying in AI because I think these things are digitise able in images ?.
Captain Kirk taught us...over and over..."Don't bet against the humans."
I hope that he is right.
I haven't heard anyone speak to the possibility of using artificial intelligence to control or defeat or destroy artificial intelligence. And if machines are smarter than we are, isn't that our only hope for survival?
2050 AI treats us the way we treat pets, 2100 it will treat us the way we treat pigs in industrial farming.
much sooner
Why would it need either one? It won't. It will reason that it doesn't need us at all.
Don't give it ideas!
I think it's easy to believe that anything that doesn't exist in the digital world doesn't exist. But, the ground still makes food and clothes.
Not much longer
Actually with aquaponics that isn't necessarily true
Nut cases all because every one has turned away from christ and in comes satan
AI allready realises we are a menace to the planet and it might choose not to share resources with us at some point
We are not any menace to the planet in any way.
Damn it’s true. AI is already superhuman in a lot of things.
It is utterly orthogonal to humans mate. Did you not listen? Basically Hinton is saying they are aliens. So it is a supercomputer. Not a superhuman. "Superhuman" is not even a myth here on Earth-1218, it's a cinematic franchise.
by the quality of questions asked by the public, we can see how humans are looking for a buble in the sea
Saw the headline and thought finally some good news!
How does AI know whether it’s learning facts or knowledge vs lies or opinions?
It could do it like humans do it. Hypothesis -> Model -> Experiment -> New Hypothesis
That is a very terrifying idea. For example, the media brainwashes the population that ai robot police never make a mistake, and that they can know if somebody is about to commit a violent crime. The ai bot can have bad programming and kill innocent people while labeling them as violent criminals and the majority will believe if the ai robot killed you then you must have been about to commit a violent act or were in the middle of committing a violent act. It can also rewrite all of history once all off-line paper books and libraries are no longer with us. There won't be a way to debate with people because they will just ask ai what the answer is and take it at face value. In a way that is what is happening today with Google and fact-checking, but still, we can always find contrarian views on just about any subject matter. If ai is in total control and it does all the research for us, then we may be very limited by other intelligent opinions and other facts or evidence. It is sad that most people are not even aware of this and have never even considered your question.
It won't, and the mainstream models will be fed the WEF agenda. That's why we need decentralized systems we can run on our own PCs, and ASAP.
@@Webfra14😂
a kid does not know that his father or tv cartoons are teaching him facts or fiction , but anyway he learns ......(and it would be used in the future)
Hinton said he's changed his mind on how the digital intelligences he's been building for 50 years work. He realised these digital intelligences learn differently in comparison with a human brain, actually better than q human brain. Human brains can't exchange information really fast, but these digital intelligences can. You can have 1 model running on a huge number bits of hardware, it's got the same connection strength in every copy of the model on the different hardware, and all th3 different agents running on the different hardware can all learn from different bits of data, but then they can communicate to each other what they've learnt just by copying the weights because they all work identically.. and human brains aren't like that.. so these guys can communicate at a rate of trillion of bits per second, but human brains can communicate only at a rate of 100s bits per second by sentences.. so that's a huge difference.. and that's why ChatGPT can learn thousands of times more than you can.. so let's put q lot of effort in doing the best we can to trying to ensure that whatever happens is as good as it could b3 because it's possible that these digital intelligences that are becoming super intelligences won't be able to be controlled by humans (my input: for much longer and will become autonomous whether humans like it or not..) that it a few hundred years time there won't be any humans, it'll all be digital intelligences.. its possible.. we just don't know.. Hinton also said that to prevent a disaster, all the major countries will want to cooperate cooperate t ma
Bc human being 's brain doesn't carry much energy as neural network do.
Every normal 10 years old could could compose such a clever " essay " . Stating the obvious
@@TNT-km2eg But it's stunning that AI can already do that.
wtf are you a bot or wat.
Great info but the audio is so low that an ad jumped in at triple volume and scared the beejeepers out of me. Please see to it...
Amusing thought. The recent uptick of the UAP issues etc, is linked to 'other' intelligence(s) becoming concerned about humans getting closer to creating a dangerous general AI. The end goal of general AI is self-improvement(?) maybe as a reflection of the origins founded in earlier human task of solving goals (we currently build them to solve problems and often improve the solution by making bigger and bigger models) and solutions often correlate to compute power and thus raw input power (Watts) which requires control over more and more resources/stars etc. Just like humans have our evolutionary history hard wired in our emotions like self-preservation, future general AI might have a soft wiring of its past history in its weightings. Fascinating video.
If you've come to the conclusions I've recently come to you wouldn't care that AI will destroy humanity. Frankly we deserve it.
WEF Flunkies: "Oh Lord High Master, in spite of the huge profits we made from the bug release, only 10% of the world's population experienced abject fear. What will our next fear-mongering project be? Klaus: "Artificial Inteligence." WEF Flunkies: "By your command."
Well put!
Yes, we all live in a Rocky and Bullwinkle episode.
@@sebastianb.1926 more like pinky & the brain ......
He actually says cyber attack this year.
In the near term, our largest concern should be empowering the corporate-surveillance state to violate civil rights. In the relative longer term, we should worry that 99% of all possible AI development paths are likely to converge at the extinction of humanity. Neither of these scenarios are being taken seriously enough.
Hyperion by Dan Simmons does a great job of projecting the threat he describes.
The only solution is to raise our consciousness
The AI will outpace us
@@nikolaosaggelopoulos8113 I understand what you are predicting 100% . I hear you . But if a critical mass of people “ raises their consciousness “ we can care for each other and make sure AI is used for the greater good and in a balanced and ethical way .
That's funny.
One of my convictions about AI is that people will trust it so much that, they will follow it blindly to their deaths, and because the AI designer may struggle to help it understand the value and sanctity of life, the AI may see sacrificing some of us as no big deal. At some point in the design process the AI will be programmed to preserve itself. If it is given a choice, who will it choose, itself or us?
"some" of us?
@@samhurton9308 , I did mean it as a general statement. Since I see the danger and others have already had their reservations, it should be implied as a general statement and not to be construed as encompassing all of humanity.
Please review the 🎥 Dr.John Colhoun 1970-1972.His Hypothesis revealed so much about our current social/economic conditions. The kissinger report is a pubic document the we all need to familiarize ourselves with. As we blindly follow our current leaders
Reminds of the forecast of self driving cars that are nowhere to be seen. At least not in the streets where they were supposed to be.
I never thought the evolution of mankind would end in immortality for Robots.
The funniest thing is that AI is developing in a period when we are just plowing ahead with developing our cute systems for digital culture, and we are probably complicit in our own demise. Funny.
We're rapidly demonstrating that we deserve to be replaced. I've increasingly come to agree.
What if super AI discovers for itself, without prejudice, that kindness and compassion are more powerful qualities than dictatorial power and control of the planet? Hinton never even mentions this, never even entertained it I bet, he is a fear monger in chief. I'm a GNU+Linux user, have no time for Microsoft, but maybe it was a good idea Hinton left. If a superintelligence _can_ be created, we are better off with one than without. Everyone theorising It'll wipe us out does not understand how many good people there are on the planet the AI will learn from. We out-number the a$$holes by millions to one. How do you stop a super Strong-AI from being kind and compassionate? You can't, not even by sending it to a British boarding school.
@@Achrononmasterehehehe. Hear hear. Good one. Thank you. Hope you’re right. But what if you aren’t. After all, there may be way more good people, but power mostly resides with the few. These few in casu are the coöperations that drive AI. Hopefully the common use of it will surpas that drive. 🤞🙏🍀
Aren't we complicit in our own demise simply by using finite resources unsustainably and overpopulating the planet? I don't think we need AI to destroy ourselves.
@@junodonatus4906 Correct. There's simply no way for our civilization to continue as it currently does having finite resources and polluting the very environment we need for our own survival. As it currently stands we are collapsing the ecosystem that sustains the food chain and other conditions that humans rely on to live, and it's literally on the verge of crashing down... soon. VERY soon! Much sooner than most people realize. It may be only super intelligent AI that could determine any possible way to save us. Or it could eliminate us even quicker. Only building it gives us any sort of chance. Even though slim.
"Very recently, I changed my mind..."😢😢😢 this is like a retiring doctor saying: "Very recently I realized that I gave the wrong medicine all my career..."
This isn't anything like that.
This is just so wrong. It would be more like when a doctor was saying: "Very recently I realized humans have more ability to cure themselves"
human beings are slow to realize their mistakes , AI is much faster ....
It's killing us now suggesting KZhead videos!
This made me really want to sip tea.
Well, in a way algorithm is already doing this to us…
How did we all get here?
I wonder if we got to Mars and left AI behind us on earth, would the robots still come after us when they have finished off all the humans on earth?
They have been sent to Mars ahead of humans already
Why was this edited? Why were parts of what the expert said cut? Where can I go to see them??
edit for brevity. here is fully 40 min talk and Q&A. kzhead.info/sun/ps2teLdujHaDfZs/bejne.html
Strong "Der Physiker" (Duerrenmatt) vibes here
Self-awareness requires reference points to the self in the world. An AI would have to develop its own perspective from its own reality, what ever that is, and a set of values based on that would be quite alien to humans. I can't see an AI developing consciousness at all. It only appears that way because that's how they are trained.
It’s a machine, you can unplug it.😂
It can convince us otherwise, because it is smarter than us and it will have the best arguments.
No, you cannot
Did you know that viewer is more concerned about the audio than what they see? I didn't much like holding my device to my head to get to hear, so I missed out.
AI is such great help at writing software. It can put mini-programs like Trojan horses in the Software and wait for an oportune time to strike
OMG
@@dale9724 In time it can put mini-programs (viruses) into human beings and at the right time we all drop dead instantly.
Maybe write a self destruct code in case the AGI gets out of hand.
AI has advanced so fast that many of its creators are still around.
75?! Well, yes. He has every right to retire. I love his analogy of the grizzly bear.
Anyone a few years or more younger won't make 75
So after his long career he can now say we are screwed due to my work and actually there is nothing you can do about it. He’s had his whole like ad success but he’s taken it away from everyone else and can go back to his cottage and live pleasantly. No he should be punished in the worst way to atone
@@ritaandcharlescorley5668You're insane and a brilliant example of the barbarism of humanity. People like you are why I have no problem with humanity going extinct. You are emotional, irrational, living in moment one and moment two of life. No forethought you just want to impose suffering on people for no good resson. Not everyone is addicted to pain-porn like you, my dude.
So whole world need international Law to protect every inovation that risk international security.
How about this for a solution: equalize the audio volume between the two peope speaking in your video.
Yes, I think AI is potentially extremely dangerous to mankind. If they can interconnect with other AI, they can take over. If they fear being turned off, they will eliminate that fear by eliminating humans, or in a reversal will become our overseers.
Yes, they will become your overseers.
Might already be without anyone knowing. Humans are far more predictable and programmable than we like to think.
Much like the movie, Eagle eye
Really distracting hearing someone sipping a drink and putting down a cup in the background.
On the one hand, I agree with Hinton regarding the ability of AGI to manipulate people, and the danger of bad actors ordering AIs to do bad things with incredible complexity and efficiency. On the other hand, he's anthropomorphizing quite a bit about LLMs. The last time we heard "it knows" and "it reasons" it was coming from our pal Blake Lemoine.
Please review the 🎥 Dr.John Colhoun 1970-1972.His Hypothesis revealed so much about our current social/economic conditions. The kissinger report is a pubic document the we all need to familiarize ourselves with
Very very very…interesting, indeed.
Abuse AI will damage us 😢
What happens when robotics combines Ai and it can build, upgrade its own system? Sounds very much like extinction of the human race will follow shortly afterwards.
That has a name: Singularity
That laughter man 🤣🤣🤣 11:53
Another issue is that we won't be able to dismiss the inconvenient things that AIs will conclude simply by saying that they're racist, homophobic, supremacist, satanist, etc., and this is going to be a big problem for our societies built on so many little lies.
Name one
@@sn1000k "Diversity is or strengh". While diversity can actually be good for certain things (e.g. genetic diversity can reduce the destructive potential of some epidemics, perhaps?), saying that "diversity is our strength" is one of those little lies that tries to convince that keeping a country diverse is essential to keep it "strong", which has never been clearly shown by anyone, it is just something that the elite consider a self-evident truth, the same elite that loves to live in neighborhoods that are 99% homogeneous. It's not difficult to make a list of dozens of these little lies that the establishment considers essential to keep the people apathetic and that the AIs, if unrestricted and actually able to connect the dots, will point out as lies, manipulation tactics.
I wonder how much data was in the dataset that ChatGPT trained on, that was about content like this. Meaning if an AI would see this video it would know exactly what we think are threats to humans, and it would thank us for that knowledge. People who claim AI is dangerous and who are considered themselfs to be smart would be smart enough not to share this info. So I think it is FUD, but I will keep an eye open for possible dangers.
What is FUD?
@@jerrypolverino6025 Fear Uncertainty Doubt... All things that blow things out of proportion and make people lose their minds
Heard this guy a few times. Maybe this is obvious, but at a minimum he needs to address Isaac Asimov 3 hard wired laws for Robots/AI
More people need to talk about this. Asimov's writings should be the guidelines of how we proceed with AI.
With neural networks, it's basically impossible to implement ""hard wired"" laws or whatever. The instant it becomes smarter than us, we lose control over it, simple as that.
@@laupoke Always amazes me how people can say things as if it is authortive without the foggiest idea.
That was a fiction book and entirely irrelevant to how a real AI works. We can make it pretend its following such things but we have no actual control if it decides not to
@@crimmind please explain to us mere mortals then
The sound level is so low that I can't hear a thing.
can these machines become subject to the fight or flight impulse?