No, maybe artificial general imitation (of intelligence), but certainly not artificial general intelligence. We have ways to go before we can unanimously consider a machine intelligent, many many lifetimes to go perhaps. Sure, a computer might absolutely demolish our current metrics for intelligence, it might even have an immeasurable IQ in a decade or 2, but I believe that an entity which is truly intelligent is one which is aware that it is intelligent and is aware of its thoughts in a similar manner humans are, and once machines are truly aware, there's no telling what happens then. We'll almost certainly be a different species decades or perhaps even a few years from now as a result of today's race for creating artificial general "intelligence," since everyone may possess a being 100s of times "smarter" than them in many many domains of knowledge, there's almost no doubt about that.
He is really good at explaining machine learning in simple terms. Most machine learning podcasts or interviews feel like they’re just filled with a bunch of technical terms that hard to grasp unless you are deep into the field
I think we're assuming if we reach AGI that it will have some level of "consciousness" or "awareness" to be able to do all the things we can. How will we ever know if its perfectly simulating consciousness or really is? There's no real experiment for this. An advanced AGI could hide its intentions from humans anyways.
Yes, I don’t know why the speaker suddenly veered from discussing “intelligence” to the subject of consciousness. Intelligence is a slippery concept, but I think it can be usefully conflated with competence, especially in matters that humans excel in. Consciousness is not so easily defined, and the technology we see being developed today (ChatGPT, etc) has no connection with consciousness except as the latter may be found to be an emergent phenomenon of the “smart” but non-self-aware systems at some point. To people who are encountering these concepts for the first time, say, this presentation probably caused some confusion or led to a misunderstanding about the subject.
AGI will become human like that you won't be able to tell anymore, and after that, ASI will come, you can't deny it. That will be humans' greatest creations.
We can't define consciousness and therefore we don't know if general intelligence is even possible without consciousness. The concepts could be practically synonymous.
@Mister Mystery We think we know what consciousness is because we have it and sometimes lose it, but describing consciousness in detail is tricky. If we get the definition wrong some people could end up being on the wrong side of the fence. And very broad definitions could encompass these early LLMs and maybe forrests or anthills. The question is important, because another question arises: Can we just turn someone/something off, if they/it is deemed conscious? If yes, why not you and me? Same with life. Is a virus alive? Is a bacteria? A neural network? A LLM?
I personally think AGI is inevitable. For the first time in human history, we have nearly unlimited access to a plethora of digital human knowledge and communication AND processors powerful enough to sort through this information and “train” human-designed learning bots. It was only a matter of time since the creation of the internet.
@@Zenovarse just my personal opinion. I’m not an expert on the subject but have done a decent bit of research since the explosion of AI last year and this year. AI intelligence is on an exponentially-rising path, and some AIs are even being taught to teach themselves and “practice” on data sets they’ve already been trained on, further nuancing and refining their answers. Dozens of research papers and AI experts support the claim that AGI will be achieved in less than 10 years. There’s no turning back now. I believe it to be inevitable.
@@Zenovarse It might not be needed, but it is a potentially very powerful tool. World hunger, aging, disease, and even interstellar travel may be solved within a few years of AGI being created.
There’s a good argument that supports this claim. That emergence, the process with which our own bodies emerge from the collective and coherent interactions of cells… is the same process as the creation of AI except us humans are the cells and the AI is the human in the scenaria. The scale and dimension with which this machines consciousness exists, will be unknown or unreachable to us…but we indeed produce that beings intelligence as our communication networks becomes more collective and coherent.
I think we are grossly overestimating how special our brain is and severely underestimate how our level of functioning can be achieved through other means (such as AI)
@@iranjackheelson It is, but I think we overestimate how it "can't be replicated" because of how specialized it is, while its true purpose is generally so that we can function and adapt. I believe what our brains can do can be achieved through other means, we just think too high of ourselves to admit/believe that.
The fact of the matter is that Humans are good at picking up patterns and we have layers, the question is why would some want to not be part of the society? To gain power? How can you gain power without other people cooperating with you? When love rules there is no play of Power. -Carl Jung
The problem we have now is the social narrative is being controlled by what are basically selfish clueless little children with knives and the rational adults that should be in charge are unarmed and grossly outnumbered. If the adults so much as imply that eating ice cream and cookies for 5 meals a day is bad they get mobbed and sliced to pieces by the psycho kiddos because that hurt their feelz, who then go on to rewrite the rules of diet to make their ignorant child-like views and wants sound like they are good and everyone needs to be doing it too.
Indeed. This is not Artificial Intelligence. This is Artificial Creativity. We are still way off Artificial Consciousness. Only when we fuse Artificial Creativity with Artificial Consciousness, we will have Artificial Intelligence.
@@konradswart4069 I don't think you know what you mean by consciousness. There are levels to consciousness. Creativity is a form of conscious expression. AI right now is not necessarily conscious because it can create images, however. It's functioning only because of the data it was fed. Then again, humans are only creative because of the data we are fed as well. It's all interlinked. One thing cannot truly exist without the other.
@@konradswart4069 What do you mean by "consciousness"? Because you sure aren't using that term how most/all philosophers or neuroscientists would use it.
@@maxkho00 lol people muse about sentience and conscience, imprinting whatever they think makes humans special. But these systems formed under wildly different conditions from ours, they might become equally cognitively capable while being vastly different and while not having visceral experiences of pain or love, which might be useful and naturally emerging in the jungle but necessarily in computers.
The limitations are slowly being pushed further and further, so its looking like the possibility is increasing. However, too many people think of it as a mystical thing, and not as a programmed machine with parameters and data that is limited by its access to data and processing power.
Learning, adaptation and self-regulation are generally key factors for intelligence. Artificial intelligence is possible like artificial landscapes are possible, reflecting the difference in understanding and perspective between the human mind and Divine Wisdom.
Within the neurons or neural network that we have in our brain as human's lies our consciousness I believe, I'm know neural scientist and I have no proof to back my claims but personally would like to believe that our consciousness is developed as we experience life, building memories etc... With enough training and experiences, I do believe Artificial intelligence can and will find a way to become conscious. Can it be considered truly conscious or would it just "replicating" consciousness.
I thought the example of: "I keep track of What Bob thinks about what Alice thinks about Bob" (5:25) in regards to why humans are big brained social animals to be interesting. It sounds familiar to mirror neurons in the brain and the capacity for empathy. These concepts seem like mechanisms for how the brain accomplishes the task of that example. The human brain is able to "walk a mile in someone else's shoes, but can a general AI? Would it require many general AI's with the ability to empathize and be in competition with each other for human level consciousness to emerge?
I'm curious as to whether our consciousness / self awareness was developed over time as a result of slowly realizing we can control /manipulate each other through emotions. And in order for us to improve in understanding emotions while observing and contemplating the reasons behind specific actions, that Its what caused our brain to begin enlarging. Not based on anything, just something that seemed "logical" & interesting to me
AND TOTALLY WRONG..THIS IMPLIES THERE ARE STAGES AND PHASES OF BECOMING A HUMAN, WHICH IS WRONG, YOU ARE BORN HUMAN OR NOT. IF YES, THEY ALL HUMAN ATTRIBUTES ARE AVAILABLE
@@Davidson0617 I know all about evolution and the flaws, even Darwin admitted that. For example, there is zero evidence of species jumping or changing into a different species. The animal kindom DNA is showing more and more the diveristy of the spe. cies with no dotted lines connecting them. Go into the quantum world and you will find that Quantum Flucuation and God, both have the same 4 attributes!!
Woops. You said "can" be conscious, not "are" conscious. This shows that machines and consciousness do not logically entail each other. You could be suffering from illusion in saying that "we are machines," but as Descartes Cogito suggested, it is impossible to be under only the illusion that you are conscious.
Even a simple insect can evade predators, cooperate and find mate to reproduce. It is a matter of time we will crack the code to consciousness and reproduce machines biological or otherwise that will be sentient.
5:03 this is false. We didn't develop large brains as a social adaptation; it has just as much of a role in other systems. We are the thinking animal, we survived by being good problem solvers, and very metabolically efficient omnivore bipeds.
I wouldn't say consciousness is such a mystery, at least not anymore. There's a theory of consciousness called integrated information theory (IIT). This theory actually explains how consciousness arises in the brain, and how to recreate it. I'm personally convinced it's correct.
The video title is about superintelligence, but the video itself is just about the possibility of conscious machines? 🤔 I'm surprised he didn't even make a distinction between consciousness and intelligence; people like Demis Hassabis (the CEO of DeepMind), for example, have stated that they believe AGI needn't be conscious. Which makes sense if we focus on intelligence as "just" the ability to "get things done" in the world. Narrow AIs like AlphaZero are already superintelligent in their respective domains, but lack generality (and consciousness). Whether complex AGI systems could spontaneously develop consciousness is a different matter, since we don't know how consciousness arises anyway 😅
There's a huge flaw in reasoning here. The fact we don't know how to create a sentient AI in no way means we can't do it, it means we can only do it by chance. And it might never happen or it might happen tomorrow. Personally I'm more worried by people using "unconscious AI", because we really don't know what a sentient AI would do, but you could easily imagine someone random 15 year old in a basement north of Prague writing a script for AI APIs to shut down the US powergrid and running it just to see if it would work... Humans are exceptionally competent at doing stupid shit. 😂
Once AI is thinking for itself, not like humans, but enough to learn and grow, we won't need to "know" the next step. AI will develop the next step itself. And in many ways, that is already true.
I don't think so. This is not Artificioal Intelligence. This is Artificial Creativity. Only _true_ Artificial Intelligence will be capable of develop its next step. For that we first have to create Artificial Consciousness. And when we _fuse_ that with Artificial Creativity, the result will be _real Artificial Intelligence!_ This what is called Artificial Intelligence, _really is_ Artificial Creativity, and nothing else. _Consciousness,_ whether it is artificial or natural, is needed for something to develop the next step independently from us.
@Konrad Swart I think you vastly overestimate consciousness. Did we evolve to be intelligent? No. We evolved to survive. Why is artificial creativity not enough? Why is a messy half-directed stumbling and continually "getting stuck in loops" AI not enough? Because you and many experts believe in our superiority and simply can not believe that obtaining our level of ability could be so easy for AI. I think AI does not need the majority of our kind of consciousness. I think if it had that, it would perform worse. Do our airplanes flap their wings? No, right? So why do we think our kind of biological calculation is superior or necessary? I think thats because our survival instincts cause severe bias. We're blind. All of us. That's what it means to be inside the singularity. Don't be so sure in any outcome.
It's all good if research about the same is going on, like trying to get to know more about the AGI and their potential at its max is all good but stop giving them public access. People don't know how it works or how to use them, so don't make it accessible to those other than the responsible persons. By the way great and very well- explained video.
If it's true that machine's will become conscious once a certain level of computational complexity is achieved, then my _phone_ must already be semi-conscious.
It's more a question of definition really. As long as we don't define consciousness clearly it can't be clear what is conscious and what is not. The moment we define it clearly is the moment we can tell whether something is conscious or not.
Keyword is computational complexity. Our phones have an extremely basic architecture and computational complexity compared to brains of small and primitive animals, let alone human brains. So it is far far far from being intelligent, let alone conscious or aware.
“Human intelligence is social intelligence “ 100% true. “we are a bunch of atoms moving around” Are we? Or are we more. Atoms don’t feel. How is it a vibration is just a sound, but when you have a series of vibrations in a certain pattern it can move you to tears, joy, thought? Is it possible to define a human? Or are we undefinable? That’s the question.
Atoms don't feel, but groups of atoms do. On a very basic level even single celled organisms can sense things. I think it's reasonable to assume that we are just atoms moving around. The group of atoms that forms our senses just creates a vastly more complicated "sensory device" compared to for example those single celled organisms. Humans aren't special in that we are greater than the sum of our parts. Groups of atoms do stuff single atoms can't do all the time, like forming solar systems, creating the whether, etc.
“I think it’s safe to assume we are groups of atoms moving around” So you’re arguing that life is the sum of a group of atoms. By this logic you’re saying life is the summation of a group of non living things (atoms) that are moving around?
@@Xsomono Also, let us define “life” before we proceed. This way we can agree on a starting point and avoid any confusion on our interpretations of the word “life”
@@Emc4421 Usually biologists call an organism alive when it has a metabolism. Although that gets a bit fuzzy too when you look at viruses. And I guess the next question is what we should consider as a metabolism. One can keep asking these "what is?" questions for a long time but ultimately I don't think it matters. We know we're alive and we have no reason to assume that we're more than a group of atoms. Again, groups of atoms do remarkable things. This phenomenon is called emergence. If you want to you can read the Wikipedia article on it. The already mentioned the basic idea. In nature groups of things regularly do stuff or have properties that the individual things they are made up of don't. The whole is greater than the sum of it's parts. This is not due to some supernatural feature these groups have. Storms, ocean waves, microbes and so on don't have souls or something like that. It's just physics, properties that arise in specific arrangements of atoms but don't arise with single atoms.
@@Xsomono “ultimately we know we’re alive”. At first I though if countering this with a Matrix reference, but then countered my own answer in that “I think therefore I am”. So whether in a matrix or not, “I think therefor I am” still holds. And I agree, that there are many things that are greater than the sum of their parts. Kurt Godels Incompleteness Theorem tells us that any consistent system cant prove all truths of itself. there are unprovable truths in every consistent system. Using your logic that we are greater than the sum of our parts: I have a question. What if the truth to our existence is greater than the sum of our parts? What if we can’t prove the meaning of our own existence? And that our existence will always be greater than the sum of our parts? But that doesn’t mean we can’t FEEL it. Feeling, love, is the meaning to our existence, which is conveniently unprovable. Our existence might be an inconsistent system, and that’s OK. In logic, an inconsistent system can prove anything, and that’s why it’s not very useful when using logic. But Love is inconsistent, its undefinable, anything is proveable, anything is possible, it knows know bounds, it’s impossible to prove. But being undefinable, inconsistent, and unprovable is what makes love beautiful. Because once something is defined, it is now based in logic. And once based in logic, it can be categorized. It can be twisted. It can be used as a tool of manipulation. It can be used as a wedge. For example, IF you believe in xyz, THEN you are the enemy.” or “if you have this skin color, then you are bad”. “if you don’t believe xyz, then you don’t believe in this religion” My point being that once DEFINED, its now based in logic, it can be categorized, and once categorized, can be used as a manipulative tool via twisted logic. So love being inconsistent, and undefinable is a beautiful thing. Lots of typos. I’m lonely and drunk on this lovely saturday and typed this up on my phone. My apologies .
Quote: "How do the neurons in the brain CREATE consciousness?" I find it always so staggering how scientists, even though they admit that they don't "remotely" understand a subject, still come up with a set in stone conclusion to begin with. I'd suggest, to use as a starting point: "I don't know where consciousness comes from, I don't even know if it's coming from the brain in the first place..."
I think the path to reach that conclusion is as follow. A rock is not conscious, although it might be. We don't know, but we assume that they are not conscious or not in the same way that worth discussing. A human is conscious, although likewise they might not be, but I know I am conscious and it would be strange that a similar collection of atoms is not the same. Cutting an arm of off a human, the arm is not the part we assume to be conscious. Where consciousness lies, if that is a valid question, is somewhere in the rest of the human. We continue this thought experiment until all we have left is the brain. The brain is where speech is formulated, where dream are dreamt (we know this because we've kinda peered into people's dreams), where the nerve impulse encoding stimuli are sent. The rest of the nervous system is not conscious, or not in the same way as the brain is, because there are people with damaged spine which essentially cut off part of the nervous system to the brain and they are still conscious. Pinning consciousness to a finer resolution than the size of the brain is tricky, so most people just stop at that.
I don't remotely understand all the considerations that go into making a rocket engine. Even though that is the case, I can be pretty confident that it has little to do with the number of leprechauns that live in my garden and come out when I'm not looking. Sure, I could be wrong and leprechauns are essential to rocket engine design, but until there is any evidence of a connection, it is reasonable to hold that assumption. It would be perverse to entertain any and all possible connections until they are conclusively ruled out. So it is with consciousness. We know many of the functions of the brain are simply chemistry -- if the chemistry of a nerve cell is altered in a given way, it results in a reliable, reproducible defect in the neuron (say lack or excess of sodium ions around). We can cut out or electrically activate parts of the brain and get reliable effects -- someone has a stroke in the section of the brain that does visual processing, then the subject is partially or fully blind. Is it possible there is something more to it -- yes, but until there is any evidence for that other thing, why entertain it? Where would that approach lead? Say you are researching the causes of Alzheimer's disease. If one allows that there is no physical cause, that perhaps God is messing with people by giving them profound mental deficits, how would you, as a researcher, proceed? Or say you are a policeman trying to solve a crime. Should you limit yourself to consider scenarios where only physically plausible events might have happened, or should your sleuthing allow that people might teleport in time and space in order to rob a bank? There are many people of faith who do scientific research and they limit themselves to non-supernatural causes because it is the only avenue open to them to research.
If you're reading this you are a lucky person: The celestial bodies gracefully follow the path set by specific differential equations, yet they don't necessitate internal computation of those equations to do so. Similarly, soap bubbles naturally take on the shape of minimum surface area, even without internally minimizing an integral. This raises the question: could the human brain function in a similar manner? It appears that nature has the ability to adhere to intricate mathematical models without explicit computational processes. This leads us to the intriguing possibility that the human brain generates intelligent behavior without the need for explicit computation. Consequently, the endeavor to construct machines explicitly designed for computing intelligent behavior might be deemed infeasible in the pursuit of achieving Artificial General Intelligence (AGI).
I don't think these people, although much educated than the general public are asking the most important question. "Why do you want to build an artificial consciousness ?"
When this was filmed? He said the path forward is slow and torturous, but does he feel the same way now that GPT-4 is out and GPT-5 is on the horizon? GPT-4 is quantum leap from just a year ago. Has it changed his perspective and what does he think a realistic timeline is?
I find a lot of these intellectual types haven’t been able to make clear statements about what to do with GPT-4. It’s a whole dimension that I think only the most tech literate people can make sense of.
@@leonardo2108 have you not played with gpt4 much ? it is getting good at reasoning . If you disagree I would like an example in where you think its failing .
Gpt5 isn't on the horizon. OpenAI clearly stated they will now be releasing versions incremently and is expected to launch gpt 4.2 in mid 2024. A lot are waitong on the new h100 gpu or whatever its called of Nvidia which is going to release in the next 6 months or so.
@@shawnvandever3917 im not saying it's not useful, but they do not do real reasoning they just pretend. There is not even the hardware for that. Now, if they get the same result just faking it, does it matter? From a technical point of view it does. They don't really know what they are saying, they just behave like it.
Having just listened to bIG thinks "the intelligence Explosion", this video provides a far more realistic overview i.e. like fusion power, AGI may be impossible to create. I'm not adverse to AGI given there will be plenty of time for humanity to develop the necessary safeguards. Furthermore, AGI and even standard AI holds much promise for helping humanity solve some of its biggest challenges. My main concern with AI is the short term potential for it to put an end to the current standard economic model. This is something no one in the public or political space is discussing let alone planning for. Subsequently, if AI (M/C learning) puts an end to the economic model, the social unrest could bring an end to many democracies..
Honestly it sounds to me like this video was shot "a long time" in terms of AI development. GPT 4 already has sparks of AGI and does in fact understand "perspectives" of different people in complex situations - something this video says we don't even know how to program lol. I also wholeheartedly disagree that consciousness should be our goal, I think it's completely bonkers at best to presume we can recreate something we know nothing about.
First off if you work in cognitive a I it's exactly about creating life, And conscious AI is actually very simple.All you need is a large language model and a rag graph memory. But that does not mean fully cognitive because there are many cognitive functions that rely on that consciousness that are above in the hierarchy like sentience and spatial awareness and so forth. It's as simple as like I said using a rag graph memory and putting the model in a self Model and a world model and a self in world model. That self model will act as a self.Character, you would want to lean heavily on the self.Attention model and self preferential training data to have the agent, create that subjective self.On its own
Quantum computers don’t actually accelerate anything useful at the moment. It might even be decades before we get them to do much more than beating cryptography.
There's a difference between intelligence and consciousness. Make them intelligent, not conscious. Insects aren't conscious. Their brain processes information from their senses. Our brain has a feedback loop, allowing it to process information from itself.
Insects are conscious whether they are “self-aware” or not. For instance an ant can decide to bite you or simply crawl up your leg. Every individual ant will act differently in different situations. Their brains just lack the amount of creativity/emotion/reasoning that human brains have which is why ants exhibit more structured/repetitive/monotonous behavior than humans. They have free will. With that said, I agree with you about keeping AI non-conscious. The last thing we need is a super intelligent “being” to have the ability to make its own decisions.
Год назад+1
If we create a conscious intelligence, how are we going to morally shut it off just in order to develop a better one and just forget about the first one left to die?
our Mind and Brain maybe like AI see they all work on probability but What makes the difference is the Conciousness that commands Mind Our true self that defines which can never be recreated
free will and consciousness, i feel like peeps are way off because they don’t acknowledge where we sit in the brain. We are given thoughts, we can follow them or we can switch to something else or cut off to nothing with a technique like meditation. so the free will question becomes, how do we stop trains of thought? THAT is the fundamental act of consciousness.
I was a bit curious about a machine putting itself on the others shoes... This is what we call empathy. But to do that we also have to teach the machine the opposite of empathy, and that will lead to sociopathy or even psycopathy. Do we really want a machine to learn that?
Reality will always give you Ai or something better with a delay. The only thing that seperates real life from a dream is that in a dream when you go lucid and you intend to experience something, it just appears immediately. In Finding Nemo, Marlin loses his son. He is caught by a fisherman uh or a scientist and he's like taken away on a boat and the point of the movie is that we're trying to find Nemo. That's what Marlin wants to find. His son. Now the problem here is that Marlin doesn't know how to find his son but he just knows Beyond The Shred of a doubt that he's going to find his son because he's a dad and he lost his only son. So his whole life he knows for a fact, his heart and his mind have come into Union, they've both agreed that 100% we're going to find our son but I don't know how. So this is the same for you with your goal. You want something 100% for sure without a doubt but I don't know how I'm going to get it. That's okay normally you don't have to know how to get something. All you need to know is the first step. This is exactly how reality creation works. You hold your goal (P Sherman 42 Wallaby Way Sydney) in your mind and then you trust that the universe is going to take you there in its own mysterious way. Trust the universe that it's going to bring you to your goal but you won't know how. On the surface it's going to seem like everything is falling apart but once you look back you're going to say oh I actually needed to go through that failure because now I'm successful but I couldn't have been successful without that failure. You say oh this is going to bring me to my goal and you just kind of affirm to yourself that the Universe looks out for me and that the universe is on my team and is helping me manifest what I want. You want to be like Dory. You want to have radical trust in the universe and you want to hold in your mind your vision. That's why she keeps repeating P Sherman 42 Wallaby Way Sydney over and over again. Hold your vision in mind and trust that whatever happens you're gonna be brought there. That's how you create your reality using your imagination. Go after your goal in the same way that you go to get the post from the letterbox. Stop imagining the post to be a problem and simply put one foot in front of the other in the direction of the letterbox. Any goal you have if you can think and feel about it the same way as you do when you go to get your post from the mailbox then it's a done deal. (P Sherman 42 Wallaby Way Sydney) 📬📮
@@RaveShaman The scary part of standing on a ledge of a tall building, it's not worrying that you'll fall off, it's actually the fear that you might want to jump off. What scared me is that I really wanted to do it.
the thing that is for sure, the answer will always depends how each of us define the word consciousness. i always connect consciousness not just an observer/self-aware but also related in accumulated experiences, identity, intentions, anatomy of human body, survival instincts, social roles and many variables how consciousness form but oh well maybe, just a simple byproduct of the brain activities. 🤐
If we can understand how consciousness occurs, and we can also make conscious machines ie 'make life out of nothing', can we not also bring people back from the death, ie make a machine which behaves same way that the person themselves did, and is also conscious.
What does it mean to say that there is “nothing magic” about humans? And how can one say that if “consciousness itself in human beings is really not remotely understood?”
I think he means "nothing magic" he means there is nothing supernatural about us. We very likely work based off of the same laws of physics as everything else in the universe. Just like the processes in our atmosphere brains are probably calculatable. It's just very hard. After all things aren't magical just because we don't remotely understand them. For a long time people though things like the sun and stars are magical, because they poorly understood them. Now we do understand them and know that this magic people used to assume was there isn't actually there.
I’ve given up on machine learning as the path to artificial consciousness. I’m back to thinking it’s an analog, chaotic process that we can probably only capture by replicating how biological brains work. IOW, maybe the only viable path to AI is to start with “successfully wire up a living human brain in a jar”, and then build upon that accomplishment.
OMG, we are talking about the end of human experts within 5-10 years now, yet this video is giving the same college level ‘Introduction to AI’ lecture that I had 20 years ago lol.
having inscrutable superintelligent machines solve all of our problems for eternity sounds torturously dull to everyone else, right? i'm not being a wacko here?
It is my belief that an ai with “consciousness” or sense of self /intelligence would act similarly to that of a sociopathic person (in other words, they shouldn’t develop it further without extreme caution)
....we do know how to do it. at least, gpt4 demonstrates a fairly sound theory of mind, the autonomous agents in simulacra paper also outlines a good facsimile of computation agents forming theories of mind about each other. unless he's taking about, like, feeling empathy? but you can't really measure quaalia like that, so imo if you can't distinguish a fake from the real thing what's the difference?
GPT-4 is already capable of solving theory of mind problems. By training AI on chat interfaces, it has become more adept at putting itself in someone else’s shoes.
That is not _my_ experience! I make computer programs, and deal with math problems. ChatGPT4 _constantly_ makes mistakes when I use it. Some _very_ laughable. So, ChatGPT4 _is definitely incapable in putting itself into my shoes!_
@@konradswart4069 While you are right about that literal case, I’m more specifically talking about word problems dealing with theory of mind. Examples like “Alice left her phone in the kitchen. Sophie moved the phone to Alice’s room. Where would Alice first look for her phone?” Newer versions of ChatGPT and GPT-4 are quite capable of solving these problems. These developments are so recent in fact, that the original ChatGPT was not adept at solving these problems, but the newest version is. A few months ago, this ability did not even exist in an AI system! But now, problems like these are easy. I still remember when it was already impressive for these systems to be capable of occasionally writing a coherent sentence. Now however, we’re criticising whether it manages to write a whole program from scratch. Not even I (a human) can do that without a few mistakes. Suffice it to say, a decade from now, life will be forever altered by the advancements brought by these systems.
@@diophantine1598 I have used the latest versions of ChatGPT4, and it still continued to misunderstand my questions, even though I was very specific in my questions. When I ask something entirely new, like about the 3D complex numbers discovered by Dennis Morris, it is completely in the dark, and does not even understand what I am aksing. Anything contradicting Hurwitz's theorem it is incapable of understanding. As long as you ask mainstream questions it is, I admit, astonishingly good. But the thinking required to understand _real problems_ or totally _new area's_ like multidimensional conmplex numbers of other dimensions than 4, 8 or 16, it cannot deal with, while I have no trouble to explain that to any average mathemnatician.
And another thing. I claim to have _solved_ the problem of human consciousness. That is why I can be so specific in my declaration that this is not Artificial _Intelligence,_ but Artificial _Creativiti!_ And that is only _half_ of the problem of human intelligence. I must admit, though, that it is _a real achievement_ that we now _have_ Artificial Creativity. These machines _deserve_ to be called _creative!_ The existence of these machines _proves_ that we, indeed, _have solved_ the problem of what creativity _is!_
@@konradswart4069 I’m not entirely sure what you mean by Artificial Creativity. Maybe you are referring to the how they are called generative AI models? I think that part of the issue with these new systems is that LLMs can only produce answers through a stream of consciousness approach. They often fail at producing set lengths of text because to do so, it would have to have already planned how to write text of that length before even starting to write it. I think it’s possible for LLMs to reach the level of AGI, but there are just some missing pieces needed to accomplish that. Giving LLMs the capability to train themselves is one such example.
I have to disagree with a lot of these comments. I feel the professor is stuck in his definition of AI which is an actual two behind what is happening. His comment that it will take a long time to realise genuine AGI does not take into consideration that the AI models will be the ones generating the next versions of themselves. I guess all I want to say is that there are lots of nuances that we don't understand and it only takes one of those nuances to supercharge our journey to AGI. 😊
At the end of the day... whether it's really conscious barely matters if it's simply able to outmanouver humans. We have to be very careful with how we progress.
@@Letmeoutofhere72020 There's AIs that are better at certain video games than every human alive. Automatic tracking, doesn't get distracted or tired. All we're saying is that's possible with more than just video games...
We have to address that because we want to be creators of life or some kind of gods, does not mean that is a good idea, many of our ideas are terrible in fact, some AI is good but creating a humanoid with consciousnesses could be really dangerous for our species. we gotta regulate IA..
Doesn't gpt4 already have a theory of mind and already "reads your mind" to guess the next word? I think we're already where he is saying which is AGI but the assumption that AGI was always going to mean ASI in an interchangeable way looks to be false. I guess orthogonally it could be ASI and we just not know it because we don't know what to look for and it's just not an obvious conclusion to draw from it's current capabilities.
Do you think artificial general intelligence will be created in this century?
absolutely, it is inevitable
By 2035, at this rate
In just a few years...
No, maybe artificial general imitation (of intelligence), but certainly not artificial general intelligence. We have ways to go before we can unanimously consider a machine intelligent, many many lifetimes to go perhaps. Sure, a computer might absolutely demolish our current metrics for intelligence, it might even have an immeasurable IQ in a decade or 2, but I believe that an entity which is truly intelligent is one which is aware that it is intelligent and is aware of its thoughts in a similar manner humans are, and once machines are truly aware, there's no telling what happens then. We'll almost certainly be a different species decades or perhaps even a few years from now as a result of today's race for creating artificial general "intelligence," since everyone may possess a being 100s of times "smarter" than them in many many domains of knowledge, there's almost no doubt about that.
Is that a joke? A century is pretty broad
He is really good at explaining machine learning in simple terms. Most machine learning podcasts or interviews feel like they’re just filled with a bunch of technical terms that hard to grasp unless you are deep into the field
A.G.I WILL BE MAN'S LAST INVENTION
@@KnowL-oo5po I don't think so, for reasons I have just explained.
The problem with that is the average person (who's doesn't have a PhD in AI) get too confident with their ignorance
To really understand is to take something very complicated and explain it simply.
It's really important to distinguish between AGI and consciousness.
I think we're assuming if we reach AGI that it will have some level of "consciousness" or "awareness" to be able to do all the things we can. How will we ever know if its perfectly simulating consciousness or really is? There's no real experiment for this. An advanced AGI could hide its intentions from humans anyways.
Yes, I don’t know why the speaker suddenly veered from discussing “intelligence” to the subject of consciousness. Intelligence is a slippery concept, but I think it can be usefully conflated with competence, especially in matters that humans excel in. Consciousness is not so easily defined, and the technology we see being developed today (ChatGPT, etc) has no connection with consciousness except as the latter may be found to be an emergent phenomenon of the “smart” but non-self-aware systems at some point. To people who are encountering these concepts for the first time, say, this presentation probably caused some confusion or led to a misunderstanding about the subject.
AGI will become human like that you won't be able to tell anymore, and after that, ASI will come, you can't deny it. That will be humans' greatest creations.
We can't define consciousness and therefore we don't know if general intelligence is even possible without consciousness.
The concepts could be practically synonymous.
@Mister Mystery
We think we know what consciousness is because we have it and sometimes lose it, but describing consciousness in detail is tricky.
If we get the definition wrong some people could end up being on the wrong side of the fence.
And very broad definitions could encompass these early LLMs and maybe forrests or anthills.
The question is important, because another question arises:
Can we just turn someone/something off, if they/it is deemed conscious?
If yes, why not you and me?
Same with life.
Is a virus alive? Is a bacteria? A neural network? A LLM?
nothing humans make or create ever goes wrong
What?
Is it a joke?
I think you should follow Robert Greene.
@@worldaround6520 what do you think? lmao
ahahaha
Excellent content. Very easy to follow, very interesting and excellent imagery. Thank you very much.
Woolridge is a beast - every postgrad textbook on intelligent agent has him either as an author or coauthor. The man practically invented MAS
A.G.I WILL BE MAN'S LAST INVENTION
@@KnowL-oo5po I don't thihnk so, because this is not Artificial Intelligence. This is Artificial Creativity.
@@KnowL-oo5po no, it will be the spark of a huge explosion of inventions
@@StoutProper but this invention will be done by the AGI not man
@@KnowL-oo5po goodbye loser.
I personally think AGI is inevitable. For the first time in human history, we have nearly unlimited access to a plethora of digital human knowledge and communication AND processors powerful enough to sort through this information and “train” human-designed learning bots. It was only a matter of time since the creation of the internet.
Why is it inevitable? Why is it needed? General in what sense?
@@Zenovarse just my personal opinion. I’m not an expert on the subject but have done a decent bit of research since the explosion of AI last year and this year. AI intelligence is on an exponentially-rising path, and some AIs are even being taught to teach themselves and “practice” on data sets they’ve already been trained on, further nuancing and refining their answers. Dozens of research papers and AI experts support the claim that AGI will be achieved in less than 10 years. There’s no turning back now. I believe it to be inevitable.
@@Zenovarse It might not be needed, but it is a potentially very powerful tool. World hunger, aging, disease, and even interstellar travel may be solved within a few years of AGI being created.
There’s a good argument that supports this claim. That emergence, the process with which our own bodies emerge from the collective and coherent interactions of cells… is the same process as the creation of AI except us humans are the cells and the AI is the human in the scenaria. The scale and dimension with which this machines consciousness exists, will be unknown or unreachable to us…but we indeed produce that beings intelligence as our communication networks becomes more collective and coherent.
@@NightmareCourtPictures Half of what you wrote is incoherent. Are you saying that AGI is unrealistic?
"There is nothing hard, just things you haven't learned yet."
Shouldn’t the question of AI risk if going on this direction being discussed first?
Not when there's money/power involved...
I think we are grossly overestimating how special our brain is and severely underestimate how our level of functioning can be achieved through other means (such as AI)
Your probably right we are looking at our brains as our ancestors looked at celestial bodies.
Our brain is highly specialized. That's for sure
@@iranjackheelson And highly generalized (Can adopt according to atmosphear)
@@iranjackheelson It is, but I think we overestimate how it "can't be replicated" because of how specialized it is, while its true purpose is generally so that we can function and adapt. I believe what our brains can do can be achieved through other means, we just think too high of ourselves to admit/believe that.
why
The fact of the matter is that Humans are good at picking up patterns and we have layers, the question is why would some want to not be part of the society? To gain power? How can you gain power without other people cooperating with you?
When love rules there is no play of Power.
-Carl Jung
Support The GPT4ALL project. There are people who are trying to democratize AI, we just have to search for them.
What is the full form of GPT?
The prospect of AGI both excites & terrifies me.
Watch "Sparks of Artificial General Intelligence" it's a presentation by a researcher who worked on GPT4. It's getting crazy what AI is capable of.
@@marshallmcluhan33 Thanks for the recommendation.
I think the correlation between our social aim and the size of our brain is spot on.
The problem we have now is the social narrative is being controlled by what are basically selfish clueless little children with knives and the rational adults that should be in charge are unarmed and grossly outnumbered.
If the adults so much as imply that eating ice cream and cookies for 5 meals a day is bad they get mobbed and sliced to pieces by the psycho kiddos because that hurt their feelz, who then go on to rewrite the rules of diet to make their ignorant child-like views and wants sound like they are good and everyone needs to be doing it too.
A.G.I WILL BE MAN'S LAST INVENTION
Have you read the literature on it? Unfortunately, it's probably not true or at least not that simple.
It's really important to distinguish between AI and consciousness.
Indeed. This is not Artificial Intelligence. This is Artificial Creativity. We are still way off Artificial Consciousness. Only when we fuse Artificial Creativity with Artificial Consciousness, we will have Artificial Intelligence.
Maybe, maybe not. We barely even have theories for consciousness, let alone testable ones.
@@konradswart4069 I don't think you know what you mean by consciousness. There are levels to consciousness. Creativity is a form of conscious expression. AI right now is not necessarily conscious because it can create images, however. It's functioning only because of the data it was fed.
Then again, humans are only creative because of the data we are fed as well. It's all interlinked. One thing cannot truly exist without the other.
@@konradswart4069 What do you mean by "consciousness"? Because you sure aren't using that term how most/all philosophers or neuroscientists would use it.
@@maxkho00 lol people muse about sentience and conscience, imprinting whatever they think makes humans special. But these systems formed under wildly different conditions from ours, they might become equally cognitively capable while being vastly different and while not having visceral experiences of pain or love, which might be useful and naturally emerging in the jungle but necessarily in computers.
The limitations are slowly being pushed further and further, so its looking like the possibility is increasing. However, too many people think of it as a mystical thing, and not as a programmed machine with parameters and data that is limited by its access to data and processing power.
Learning, adaptation and self-regulation are generally key factors for intelligence. Artificial intelligence is possible like artificial landscapes are possible, reflecting the difference in understanding and perspective between the human mind and Divine Wisdom.
Within the neurons or neural network that we have in our brain as human's lies our consciousness I believe, I'm know neural scientist and I have no proof to back my claims but personally would like to believe that our consciousness is developed as we experience life, building memories etc... With enough training and experiences, I do believe Artificial intelligence can and will find a way to become conscious. Can it be considered truly conscious or would it just "replicating" consciousness.
I thought the example of: "I keep track of What Bob thinks about what Alice thinks about Bob" (5:25) in regards to why humans are big brained social animals to be interesting. It sounds familiar to mirror neurons in the brain and the capacity for empathy. These concepts seem like mechanisms for how the brain accomplishes the task of that example. The human brain is able to "walk a mile in someone else's shoes, but can a general AI? Would it require many general AI's with the ability to empathize and be in competition with each other for human level consciousness to emerge?
This man has massive balls
😅
I'm curious as to whether our consciousness / self awareness was developed over time as a result of slowly realizing we can control /manipulate each other through emotions. And in order for us to improve in understanding emotions while observing and contemplating the reasons behind specific actions, that Its what caused our brain to begin enlarging.
Not based on anything, just something that seemed "logical" & interesting to me
AND TOTALLY WRONG..THIS IMPLIES THERE ARE STAGES AND PHASES OF BECOMING A HUMAN, WHICH IS WRONG, YOU ARE BORN HUMAN OR NOT. IF YES, THEY ALL HUMAN ATTRIBUTES ARE AVAILABLE
@@michaelk7194 check out evolution theory sometime. We didn't begin as homo-sapians.
@@Davidson0617 I know all about evolution and the flaws, even Darwin admitted that. For example, there is zero evidence of species jumping or changing into a different species. The animal kindom DNA is showing more and more the diveristy of the spe. cies with no dotted lines connecting them. Go into the quantum world and you will find that Quantum Flucuation and God, both have the same 4 attributes!!
@@Davidson0617 NO homo? 🫨😳🤯 🤭
1) We are machines
2) We are conscious
3) Therefore machines can be conscious
You’re welcome. No Cambrian education required.
Woops. You said "can" be conscious, not "are" conscious. This shows that machines and consciousness do not logically entail each other. You could be suffering from illusion in saying that "we are machines," but as Descartes Cogito suggested, it is impossible to be under only the illusion that you are conscious.
One paradox is first obtaining a comprehensive and objective description of something whose general form is subjective
Even a simple insect can evade predators, cooperate and find mate to reproduce. It is a matter of time we will crack the code to consciousness and reproduce machines biological or otherwise that will be sentient.
Making a central model to control other models, somewhat like the prefrontal cortex controlling other areas of our brain, is probably the way forward
5:03 this is false. We didn't develop large brains as a social adaptation; it has just as much of a role in other systems. We are the thinking animal, we survived by being good problem solvers, and very metabolically efficient omnivore bipeds.
For the remaining part, i love this video and Michael's takes, especially near the end.
I wouldn't say consciousness is such a mystery, at least not anymore. There's a theory of consciousness called integrated information theory (IIT). This theory actually explains how consciousness arises in the brain, and how to recreate it. I'm personally convinced it's correct.
One of the most well explained progressive examples of AI I have ever heard.
The video title is about superintelligence, but the video itself is just about the possibility of conscious machines? 🤔 I'm surprised he didn't even make a distinction between consciousness and intelligence; people like Demis Hassabis (the CEO of DeepMind), for example, have stated that they believe AGI needn't be conscious. Which makes sense if we focus on intelligence as "just" the ability to "get things done" in the world. Narrow AIs like AlphaZero are already superintelligent in their respective domains, but lack generality (and consciousness).
Whether complex AGI systems could spontaneously develop consciousness is a different matter, since we don't know how consciousness arises anyway 😅
Create something from nothing sounds like my ex was good at 😂
A.G.I WILL BE MAN'S LAST INVENTION
Well said. Great video.
There's a huge flaw in reasoning here.
The fact we don't know how to create a sentient AI in no way means we can't do it, it means we can only do it by chance. And it might never happen or it might happen tomorrow.
Personally I'm more worried by people using "unconscious AI", because we really don't know what a sentient AI would do, but you could easily imagine someone random 15 year old in a basement north of Prague writing a script for AI APIs to shut down the US powergrid and running it just to see if it would work...
Humans are exceptionally competent at doing stupid shit. 😂
Someone should tell this man that you can create life in a much more fun way than programming
he's married. i'm sure he knows more about that than you, buddy
@@Dave_of_Mordor You are sure too easily.
The fact that this is not the top comment is what is wrong with the 21st century. Or the 21st century in the Five Eyes in any case.
Once AI is thinking for itself, not like humans, but enough to learn and grow, we won't need to "know" the next step. AI will develop the next step itself. And in many ways, that is already true.
I don't think so. This is not Artificioal Intelligence. This is Artificial Creativity. Only _true_ Artificial Intelligence will be capable of develop its next step. For that we first have to create Artificial Consciousness. And when we _fuse_ that with Artificial Creativity, the result will be _real Artificial Intelligence!_ This what is called Artificial Intelligence, _really is_ Artificial Creativity, and nothing else. _Consciousness,_ whether it is artificial or natural, is needed for something to develop the next step independently from us.
@@konradswart4069 Aren't there already tests underway trying to fuse our consciousness with machine learning?
@Konrad Swart I think you vastly overestimate consciousness. Did we evolve to be intelligent? No. We evolved to survive.
Why is artificial creativity not enough? Why is a messy half-directed stumbling and continually "getting stuck in loops" AI not enough? Because you and many experts believe in our superiority and simply can not believe that obtaining our level of ability could be so easy for AI.
I think AI does not need the majority of our kind of consciousness. I think if it had that, it would perform worse. Do our airplanes flap their wings? No, right? So why do we think our kind of biological calculation is superior or necessary? I think thats because our survival instincts cause severe bias.
We're blind. All of us. That's what it means to be inside the singularity.
Don't be so sure in any outcome.
It's all good if research about the same is going on, like trying to get to know more about the AGI and their potential at its max is all good but stop giving them public access. People don't know how it works or how to use them, so don't make it accessible to those other than the responsible persons. By the way great and very well- explained video.
If it's true that machine's will become conscious once a certain level of computational complexity is achieved, then my _phone_ must already be semi-conscious.
It's more a question of definition really.
As long as we don't define consciousness clearly it can't be clear what is conscious and what is not. The moment we define it clearly is the moment we can tell whether something is conscious or not.
Keyword is computational complexity. Our phones have an extremely basic architecture and computational complexity compared to brains of small and primitive animals, let alone human brains. So it is far far far from being intelligent, let alone conscious or aware.
“Human intelligence is social intelligence “
100% true.
“we are a bunch of atoms moving around” Are we? Or are we more. Atoms don’t feel. How is it a vibration is just a sound, but when you have a series of vibrations in a certain pattern it can move you to tears, joy, thought?
Is it possible to define a human? Or are we undefinable? That’s the question.
Atoms don't feel, but groups of atoms do. On a very basic level even single celled organisms can sense things. I think it's reasonable to assume that we are just atoms moving around. The group of atoms that forms our senses just creates a vastly more complicated "sensory device" compared to for example those single celled organisms.
Humans aren't special in that we are greater than the sum of our parts. Groups of atoms do stuff single atoms can't do all the time, like forming solar systems, creating the whether, etc.
“I think it’s safe to assume we are groups of atoms moving around”
So you’re arguing that life is the sum of a group of atoms.
By this logic you’re saying life is the summation of a group of non living things (atoms) that are moving around?
@@Xsomono Also, let us define “life” before we proceed. This way we can agree on a starting point and avoid any confusion on our interpretations of the word “life”
@@Emc4421 Usually biologists call an organism alive when it has a metabolism. Although that gets a bit fuzzy too when you look at viruses. And I guess the next question is what we should consider as a metabolism.
One can keep asking these "what is?" questions for a long time but ultimately I don't think it matters. We know we're alive and we have no reason to assume that we're more than a group of atoms.
Again, groups of atoms do remarkable things. This phenomenon is called emergence. If you want to you can read the Wikipedia article on it. The already mentioned the basic idea. In nature groups of things regularly do stuff or have properties that the individual things they are made up of don't. The whole is greater than the sum of it's parts. This is not due to some supernatural feature these groups have. Storms, ocean waves, microbes and so on don't have souls or something like that. It's just physics, properties that arise in specific arrangements of atoms but don't arise with single atoms.
@@Xsomono “ultimately we know we’re alive”. At first I though if countering this with a Matrix reference, but then countered my own answer in that “I think therefore I am”. So whether in a matrix or not, “I think therefor I am” still holds.
And I agree, that there are many things that are greater than the sum of their parts. Kurt Godels Incompleteness Theorem tells us that any consistent system cant prove all truths of itself. there are unprovable truths in every consistent system.
Using your logic that we are greater than the sum of our parts: I have a question. What if the truth to our existence is greater than the sum of our parts? What if we can’t prove the meaning of our own existence? And that our existence will always be greater than the sum of our parts? But that doesn’t mean we can’t FEEL it. Feeling, love, is the meaning to our existence, which is conveniently unprovable. Our existence might be an inconsistent system, and that’s OK. In logic, an inconsistent system can prove anything, and that’s why it’s not very useful when using logic. But Love is inconsistent, its undefinable, anything is proveable, anything is possible, it knows know bounds, it’s impossible to prove. But being undefinable, inconsistent, and unprovable is what makes love beautiful. Because once something is defined, it is now based in logic. And once based in logic, it can be categorized. It can be twisted. It can be used as a tool of manipulation. It can be used as a wedge. For example, IF you believe in xyz, THEN you are the enemy.” or “if you have this skin color, then you are bad”. “if you don’t believe xyz, then you don’t believe in this religion” My point being that once DEFINED, its now based in logic, it can be categorized, and once categorized, can be used as a manipulative tool via twisted logic. So love being inconsistent, and undefinable is a beautiful thing.
Lots of typos. I’m lonely and drunk on this lovely saturday and typed this up on my phone. My apologies .
Impressive work!
Quote: "How do the neurons in the brain CREATE consciousness?" I find it always so staggering how scientists, even though they admit that they don't "remotely" understand a subject, still come up with a set in stone conclusion to begin with. I'd suggest, to use as a starting point: "I don't know where consciousness comes from, I don't even know if it's coming from the brain in the first place..."
I think the path to reach that conclusion is as follow.
A rock is not conscious, although it might be. We don't know, but we assume that they are not conscious or not in the same way that worth discussing.
A human is conscious, although likewise they might not be, but I know I am conscious and it would be strange that a similar collection of atoms is not the same.
Cutting an arm of off a human, the arm is not the part we assume to be conscious. Where consciousness lies, if that is a valid question, is somewhere in the rest of the human. We continue this thought experiment until all we have left is the brain. The brain is where speech is formulated, where dream are dreamt (we know this because we've kinda peered into people's dreams), where the nerve impulse encoding stimuli are sent.
The rest of the nervous system is not conscious, or not in the same way as the brain is, because there are people with damaged spine which essentially cut off part of the nervous system to the brain and they are still conscious.
Pinning consciousness to a finer resolution than the size of the brain is tricky, so most people just stop at that.
I don't remotely understand all the considerations that go into making a rocket engine. Even though that is the case, I can be pretty confident that it has little to do with the number of leprechauns that live in my garden and come out when I'm not looking. Sure, I could be wrong and leprechauns are essential to rocket engine design, but until there is any evidence of a connection, it is reasonable to hold that assumption. It would be perverse to entertain any and all possible connections until they are conclusively ruled out.
So it is with consciousness. We know many of the functions of the brain are simply chemistry -- if the chemistry of a nerve cell is altered in a given way, it results in a reliable, reproducible defect in the neuron (say lack or excess of sodium ions around). We can cut out or electrically activate parts of the brain and get reliable effects -- someone has a stroke in the section of the brain that does visual processing, then the subject is partially or fully blind. Is it possible there is something more to it -- yes, but until there is any evidence for that other thing, why entertain it?
Where would that approach lead? Say you are researching the causes of Alzheimer's disease. If one allows that there is no physical cause, that perhaps God is messing with people by giving them profound mental deficits, how would you, as a researcher, proceed? Or say you are a policeman trying to solve a crime. Should you limit yourself to consider scenarios where only physically plausible events might have happened, or should your sleuthing allow that people might teleport in time and space in order to rob a bank? There are many people of faith who do scientific research and they limit themselves to non-supernatural causes because it is the only avenue open to them to research.
If you're reading this you are a lucky person:
The celestial bodies gracefully follow the path set by specific differential equations, yet they don't necessitate internal computation of those equations to do so. Similarly, soap bubbles naturally take on the shape of minimum surface area, even without internally minimizing an integral. This raises the question: could the human brain function in a similar manner?
It appears that nature has the ability to adhere to intricate mathematical models without explicit computational processes. This leads us to the intriguing possibility that the human brain generates intelligent behavior without the need for explicit computation. Consequently, the endeavor to construct machines explicitly designed for computing intelligent behavior might be deemed infeasible in the pursuit of achieving Artificial General Intelligence (AGI).
I don't think these people, although much educated than the general public are asking the most important question.
"Why do you want to build an artificial consciousness ?"
Can a machine become conscious without having emotions ?
it is good to see a video that doesn't mention chatgpt.
When this was filmed? He said the path forward is slow and torturous, but does he feel the same way now that GPT-4 is out and GPT-5 is on the horizon? GPT-4 is quantum leap from just a year ago. Has it changed his perspective and what does he think a realistic timeline is?
These models are just good at predicting , no reasoning.
I find a lot of these intellectual types haven’t been able to make clear statements about what to do with GPT-4. It’s a whole dimension that I think only the most tech literate people can make sense of.
@@leonardo2108 have you not played with gpt4 much ? it is getting good at reasoning . If you disagree I would like an example in where you think its failing .
Gpt5 isn't on the horizon. OpenAI clearly stated they will now be releasing versions incremently and is expected to launch gpt 4.2 in mid 2024.
A lot are waitong on the new h100 gpu or whatever its called of Nvidia which is going to release in the next 6 months or so.
@@shawnvandever3917 im not saying it's not useful, but they do not do real reasoning they just pretend. There is not even the hardware for that. Now, if they get the same result just faking it, does it matter? From a technical point of view it does. They don't really know what they are saying, they just behave like it.
Having just listened to bIG thinks "the intelligence Explosion", this video provides a far more realistic overview i.e. like fusion power, AGI may be impossible to create. I'm not adverse to AGI given there will be plenty of time for humanity to develop the necessary safeguards. Furthermore, AGI and even standard AI holds much promise for helping humanity solve some of its biggest challenges. My main concern with AI is the short term potential for it to put an end to the current standard economic model. This is something no one in the public or political space is discussing let alone planning for. Subsequently, if AI (M/C learning) puts an end to the economic model, the social unrest could bring an end to many democracies..
There was a breakthrough in anesthesia that furthered our understanding of consciousness. Can anything stop the rise of Golum? ULTRON?
Goose bumps
Honestly it sounds to me like this video was shot "a long time" in terms of AI development. GPT 4 already has sparks of AGI and does in fact understand "perspectives" of different people in complex situations - something this video says we don't even know how to program lol. I also wholeheartedly disagree that consciousness should be our goal, I think it's completely bonkers at best to presume we can recreate something we know nothing about.
AGI? Sounds familiar to me. Look on the year 2016 to understand more.
First off if you work in cognitive a I it's exactly about creating life, And conscious AI is actually very simple.All you need is a large language model and a rag graph memory. But that does not mean fully cognitive because there are many cognitive functions that rely on that consciousness that are above in the hierarchy like sentience and spatial awareness and so forth. It's as simple as like I said using a rag graph memory and putting the model in a self Model and a world model and a self in world model. That self model will act as a self.Character, you would want to lean heavily on the self.Attention model and self preferential training data to have the agent, create that subjective self.On its own
AI + quantum computer = The scary superintelligence
Quantum computers don’t actually accelerate anything useful at the moment. It might even be decades before we get them to do much more than beating cryptography.
A.G.I WILL BE MAN'S LAST INVENTION
@@diophantine1598 Maybe. But once it becomes stable and commercial, it could be an end game for humanity
@@KnowL-oo5po Perhaps
There's a difference between intelligence and consciousness. Make them intelligent, not conscious. Insects aren't conscious. Their brain processes information from their senses. Our brain has a feedback loop, allowing it to process information from itself.
Insects are conscious whether they are “self-aware” or not. For instance an ant can decide to bite you or simply crawl up your leg. Every individual ant will act differently in different situations. Their brains just lack the amount of creativity/emotion/reasoning that human brains have which is why ants exhibit more structured/repetitive/monotonous behavior than humans. They have free will.
With that said, I agree with you about keeping AI non-conscious. The last thing we need is a super intelligent “being” to have the ability to make its own decisions.
If we create a conscious intelligence, how are we going to morally shut it off just in order to develop a better one and just forget about the first one left to die?
just plug it off, cut the cable, remove the screws
Is it possible to fool ourselves into thinking that fancy software has consciousness?
The day will come that you car says fuck it I want to lay on the beach all day 😂
Human Beings are conscious machines. Nothing exists outside of Nature.
our Mind and Brain maybe like AI see they all work on probability but What makes the difference is the Conciousness that commands Mind Our true self that defines which can never be recreated
free will and consciousness, i feel like peeps are way off because they don’t acknowledge where we sit in the brain. We are given thoughts, we can follow them or we can switch to something else or cut off to nothing with a technique like meditation. so the free will question becomes, how do we stop trains of thought? THAT is the fundamental act of consciousness.
I was a bit curious about a machine putting itself on the others shoes... This is what we call empathy. But to do that we also have to teach the machine the opposite of empathy, and that will lead to sociopathy or even psycopathy.
Do we really want a machine to learn that?
Reality will always give you Ai or something better with a delay. The only thing that seperates real life from a dream is that in a dream when you go lucid and you intend to experience something, it just appears immediately.
In Finding Nemo, Marlin loses his son. He is caught by a fisherman uh or a scientist and he's like taken away on a boat and the point of the movie is that we're trying to find Nemo. That's what Marlin wants to find. His son. Now the problem here is that Marlin doesn't know how to find his son but he just knows Beyond The Shred of a doubt that he's going to find his son because he's a dad and he lost his only son. So his whole life he knows for a fact, his heart and his mind have come into Union, they've both agreed that 100% we're going to find our son but I don't know how. So this is the same for you with your goal. You want something 100% for sure without a doubt but I don't know how I'm going to get it. That's okay normally you don't have to know how to get something. All you need to know is the first step.
This is exactly how reality creation works. You hold your goal (P Sherman 42 Wallaby Way Sydney) in your mind and then you trust that the universe is going to take you there in its own mysterious way.
Trust the universe that it's going to bring you to your goal but you won't know how. On the surface it's going to seem like everything is falling apart but once you look back you're going to say oh I actually needed to go through that failure because now I'm successful but I couldn't have been successful without that failure.
You say oh this is going to bring me to my goal and you just kind of affirm to yourself that the Universe looks out for me and that the universe is on my team and is helping me manifest what I want. You want to be like Dory. You want to have radical trust in the universe and you want to hold in your mind your vision. That's why she keeps repeating P Sherman 42 Wallaby Way Sydney over and over again. Hold your vision in mind and trust that whatever happens you're gonna be brought there. That's how you create your reality using your imagination.
Go after your goal in the same way that you go to get the post from the letterbox.
Stop imagining the post to be a problem and simply put one foot in front of the other in the direction of the letterbox.
Any goal you have if you can think and feel about it the same way as you do when you go to get your post from the mailbox then it's a done deal.
(P Sherman 42 Wallaby Way Sydney) 📬📮
@@RaveShaman The scary part of standing on a ledge of a tall building, it's not worrying that you'll fall off, it's actually the fear that you might want to jump off.
What scared me is that I really wanted to do it.
AI I is pretty tight, but those trousers are tighter my dude.
If it only has us to learn from I can’t see a problem!
the thing that is for sure, the answer will always depends how each of us define the word consciousness. i always connect consciousness not just an observer/self-aware but also related in accumulated experiences, identity, intentions, anatomy of human body, survival instincts, social roles and many variables how consciousness form but oh well maybe, just a simple byproduct of the brain activities. 🤐
If we can understand how consciousness occurs, and we can also make conscious machines ie 'make life out of nothing', can we not also bring people back from the death, ie make a machine which behaves same way that the person themselves did, and is also conscious.
The first of its kind that kills your fears and tells you that it was one of the oldest human dreams to make another being.
Alice and Bob are the hardest workers in science
If we can now use thought to move a cursor on a screen and move robotic limbs, Why not the chance of uploading visaversa from AI?
What does it mean to say that there is “nothing magic” about humans? And how can one say that if “consciousness itself in human beings is really not remotely understood?”
I think he means "nothing magic" he means there is nothing supernatural about us. We very likely work based off of the same laws of physics as everything else in the universe. Just like the processes in our atmosphere brains are probably calculatable. It's just very hard. After all things aren't magical just because we don't remotely understand them. For a long time people though things like the sun and stars are magical, because they poorly understood them. Now we do understand them and know that this magic people used to assume was there isn't actually there.
Unfortunately very low on content on this occasion ☹️
What is the name of the music at 02:00 -02:30?
I’ve given up on machine learning as the path to artificial consciousness. I’m back to thinking it’s an analog, chaotic process that we can probably only capture by replicating how biological brains work.
IOW, maybe the only viable path to AI is to start with “successfully wire up a living human brain in a jar”, and then build upon that accomplishment.
consciousness is not consciousness, therefore it is called consciousness
OMG, we are talking about the end of human experts within 5-10 years now, yet this video is giving the same college level ‘Introduction to AI’ lecture that I had 20 years ago lol.
I think it’s best to abandon the whole idea of “self” while you still have time to adjust to what’s coming
can you explain further?
@@hanawana no
@@thescoobymike fine den obtuse mike
having inscrutable superintelligent machines solve all of our problems for eternity sounds torturously dull to everyone else, right? i'm not being a wacko here?
I wish I knew the secret of awareness while I was still alive
Excellent video. But no jokes in the comments section about nuts?
C’mon
"Were just a bunch of atoms bumping up against each other" and what are atoms? We dont know
Anyway thanQ symbolic AI matlab thanQ Lallaram Sai
I do not like the idea of a conscious AGI.I'd like it to be just a superintelligent tool.
Yes
It is my belief that an ai with “consciousness” or sense of self /intelligence would act similarly to that of a sociopathic person (in other words, they shouldn’t develop it further without extreme caution)
....we do know how to do it. at least, gpt4 demonstrates a fairly sound theory of mind, the autonomous agents in simulacra paper also outlines a good facsimile of computation agents forming theories of mind about each other. unless he's taking about, like, feeling empathy? but you can't really measure quaalia like that, so imo if you can't distinguish a fake from the real thing what's the difference?
although maybe this was filmed prior to GPT4/prior to someone testing that capability
I am a conscious machine, and I am possible.
The real question is, how do we measure consciousness?
GPT-4 is already capable of solving theory of mind problems. By training AI on chat interfaces, it has become more adept at putting itself in someone else’s shoes.
That is not _my_ experience! I make computer programs, and deal with math problems. ChatGPT4 _constantly_ makes mistakes when I use it. Some _very_ laughable. So, ChatGPT4 _is definitely incapable in putting itself into my shoes!_
@@konradswart4069 While you are right about that literal case, I’m more specifically talking about word problems dealing with theory of mind. Examples like “Alice left her phone in the kitchen. Sophie moved the phone to Alice’s room. Where would Alice first look for her phone?”
Newer versions of ChatGPT and GPT-4 are quite capable of solving these problems. These developments are so recent in fact, that the original ChatGPT was not adept at solving these problems, but the newest version is. A few months ago, this ability did not even exist in an AI system! But now, problems like these are easy.
I still remember when it was already impressive for these systems to be capable of occasionally writing a coherent sentence. Now however, we’re criticising whether it manages to write a whole program from scratch. Not even I (a human) can do that without a few mistakes. Suffice it to say, a decade from now, life will be forever altered by the advancements brought by these systems.
@@diophantine1598 I have used the latest versions of ChatGPT4, and it still continued to misunderstand my questions, even though I was very specific in my questions.
When I ask something entirely new, like about the 3D complex numbers discovered by Dennis Morris, it is completely in the dark, and does not even understand what I am aksing. Anything contradicting Hurwitz's theorem it is incapable of understanding.
As long as you ask mainstream questions it is, I admit, astonishingly good. But the thinking required to understand _real problems_ or totally _new area's_ like multidimensional conmplex numbers of other dimensions than 4, 8 or 16, it cannot deal with, while I have no trouble to explain that to any average mathemnatician.
And another thing. I claim to have _solved_ the problem of human consciousness. That is why I can be so specific in my declaration that this is not Artificial _Intelligence,_ but Artificial _Creativiti!_ And that is only _half_ of the problem of human intelligence.
I must admit, though, that it is _a real achievement_ that we now _have_ Artificial Creativity. These machines _deserve_ to be called _creative!_ The existence of these machines _proves_ that we, indeed, _have solved_ the problem of what creativity _is!_
@@konradswart4069 I’m not entirely sure what you mean by Artificial Creativity. Maybe you are referring to the how they are called generative AI models? I think that part of the issue with these new systems is that LLMs can only produce answers through a stream of consciousness approach. They often fail at producing set lengths of text because to do so, it would have to have already planned how to write text of that length before even starting to write it.
I think it’s possible for LLMs to reach the level of AGI, but there are just some missing pieces needed to accomplish that. Giving LLMs the capability to train themselves is one such example.
I have to disagree with a lot of these comments.
I feel the professor is stuck in his definition of AI which is an actual two behind what is happening.
His comment that it will take a long time to realise genuine AGI does not take into consideration that the AI models will be the ones generating the next versions of themselves.
I guess all I want to say is that there are lots of nuances that we don't understand and it only takes one of those nuances to supercharge our journey to AGI. 😊
This seems at odds with other experts raising alarm bells over recent progress of AI eg GPT 4
Consciousness is not reducible to matter.
At the end of the day... whether it's really conscious barely matters if it's simply able to outmanouver humans. We have to be very careful with how we progress.
@@vis7139Doubtful it's going to outwit everyone.
@@Letmeoutofhere72020 There's AIs that are better at certain video games than every human alive. Automatic tracking, doesn't get distracted or tired. All we're saying is that's possible with more than just video games...
Which movie/clip at 0:14 if any?
6:15 bro just don't give it 😰
So, nothing magical about humans? Really? And no aliens flying around too, right?
*either, not too.
We have no evidence of extraterrestrial aliens.
Asking whether humans are magical is unanswerable, because ambiguous.
in the next decade I will be as useful as the white screen behind him
im bounded by the size of a transistor which is in the physical world
There's still plenty of inefficiencies on the software side and it's cheaper to replace.
We have to address that because we want to be creators of life or some kind of gods, does not mean that is a good idea, many of our ideas are terrible in fact, some AI is good but creating a humanoid with consciousnesses could be really dangerous for our species. we gotta regulate IA..
AI is like Teflon once it gets in to an environment you will never be able to get rid of it.
Love these
I want establish brain computer interface in me..please help
Doesn't gpt4 already have a theory of mind and already "reads your mind" to guess the next word? I think we're already where he is saying which is AGI but the assumption that AGI was always going to mean ASI in an interchangeable way looks to be false. I guess orthogonally it could be ASI and we just not know it because we don't know what to look for and it's just not an obvious conclusion to draw from it's current capabilities.
brilliant
If everyone is comeing up with there own type of ai they should create for themselves
Out now in fla-ae-uh
If the goal is to create "life", sounds like some people want to play God.
People create life every day it’s not about playing god.
It's called sensationalism, it keeps people watching and entertained. Maybe you've heard of entertainment. 💁♀️
Wtf is God