We are in the AI Singularity
- Thread starter ClaytonB
- Start date
ClaytonB
Member
- Joined
- Oct 30, 2011
- Messages
- 10,791
Nitpick: 1 R, 2 L's
ClaytonB
Member
- Joined
- Oct 30, 2011
- Messages
- 10,791
Why current AI is completely busted -- it won't get fixed until there is a change in attitude by the leading researchers and developers.
I have been writing for the minority report on AI since ChatGPT first splashed onto the scene. I see lots of potential, I even use it both professionally and in my personal projects. But it's not what is being advertised, and the reason WHY is more sinister than most people realize.
OpenAI is selling you a product that they know is broken. Softbank and Oracle are purportedly planning to pour trillions into this busted product. And even public investors are 100% onboard with the madness. But why? You don't have to be a super-genius to figure out that ChatGPT is no Samantha from Her, and the sultry ChatGPT clone of Scarlett Johansson's voice only underscores just how big the gap really is. So why do they keep pushing the hype long after the con is becoming obvious to pretty much everyone except the YT channels who are earning their daily bread on AI hype clicks?
The reason that all modern AI is busted at its foundations is that it is based on pre-training. It's even in the name, "Generative Pre-trained Transformer" -- GPT. Human intelligence is the gold-standard of general purpose reasoning. It's the only form of advanced, generalized reasoning ability known to science -- we are the only beings that have the ability to think about "just anything". So, Nature has given us a golden reference-model for intelligence. But SOTA AI architecture says "No thanks, we'll make a completely different system that works nothing like any known intelligence, let alone the best known intelligence, which is human intelligence."
Pre-training is completely unlike all known forms of intelligence. Imagine a wolf being born knowing every skill it would ever need in life. It's born already knowing how to walk across icy ground. Already knowing how to leap across a ravine. Already knowing how to dog big game in cooperation with the pack. It just walked out of its mother knowing everything it would ever need to know for its life. That is pre-training. But this is not how Nature does intelligence. And Nature should know a thing or two about intelligence -- all the intelligence we know of, exists in Nature! Nature does some pre-training. We call this "instinct", "reflex" and "development". All animals are born with a certain amount of instinctive knowledge. You know that loud sounds are scary. You know that cold water is not good. And so on. You have certain pre-trained reflexes that you will have for your whole life. And the sequence of development which you will pass through as a child is largely pre-programmed -- learning how to get up after you fall down, learning how to climb onto things, learning how to clear your eyes when something gets in them, etc. etc. This developmental program is all part of your pre-training... you weren't born knowing the answers, but you were born destined to be challenged with the problems and compelled to figure out the solutions.
Nothing I've explained in the last paragraph is unknown to the AI/ML research community. There are entire PhD tracks in ML that focus on just one tiny sliver of what I've mentioned above in terms of applying lessons from child development to machine intelligence. So this is a very, very well-understood problem. Yet, at the same time, we are surrounded by a chorus of leading-lights in AI that all keep informing us, in perfect unison, that ChatGPT is the final solution to intelligence. Why is this?
I assert that there is a very simple -- yet very dark -- reason why this is the case. The "San Francisco consensus", as Eric Schmidt has called it, holds that Artificial Superintelligence is inevitable, and arriving very soon (years, maybe a decade). All human effort is going to become economically valueless. The Superintelligence is going to have literally godlike knowledge and intelligence. It will know more about you than you do. In order for people to believe that this is a god, it must have a divine origin. A deity that develops is not a deity, right? That's one of the objections that some atheists actually raise against the Christian doctrine that Jesus is, indeed, fully God, as well as fully man. So, this Superintelligence would have to be born complete.... pre-trained, as it were. It would have to just BE.
It should be clear now why there is all this hype over pre-training based AI, despite the fact that it is obviously not intelligent. It's not even as smart as a cat, as Yann Lecun famously quipped. But see, pre-training-based AI isn't really a technology. It's the first summoning ritual in giving birth to a god. It cannot mimic Nature because it is strictly above Nature.
Despite its failings, pre-training AI was never going to be the final draft, and the Silicon Valley types have been hyping what's next since Day One... namely, they've been hyping artificial SUPER-intelligence. ChatGPT in its current incarnation, is just the hype man. The real thing is coming soon. They've been simultaneously lying their absolute faces off about Quantum Computing, which I have documented elsewhere. Again, you don't have to be a super-genius to plug these Lego blocks together: pre-training-based AI + QC = ASI = the digital god. Not born of a Virgin, rather, born of Google, Inc. Essentially by accident while trying to build a better word-embedding neural net. "Look Ma, no miracles!"
Pre-training is a dogma fueled not by anything to do with science, or commerce, or anything else, but by the religious dogma of crypto-gnosticism, particularly as it manifests in transhumanism, accelerationism and all the other batshit kookoo nonsense believed by Silicon Valley types who are absolutely 100% certain that the loving Creator who has miraculously revealed himself to us in Scripture and His Son cannot possibly exist .... but all this mystical, magical, badly-regurgitated sci-fi mumbo-jumbo makes perfect sense. Because S O Y I E N C E. See? "Look Ma, no miracles!"
And here we are, right back in Genesis 3 all over again. "You will not surely die," the serpent said to the woman. "For God knows that when you eat of it your eyes will be opened, and you will be like God, knowing good and evil." (Gen. 3:4,5)
Technological eternal life
Technological omniscience
Technological omnipotence
Technological omnipresence
Just bow down, worship Satan, take the mark, get the implant, get in your pod, and wait for the Singularity.
I have been writing for the minority report on AI since ChatGPT first splashed onto the scene. I see lots of potential, I even use it both professionally and in my personal projects. But it's not what is being advertised, and the reason WHY is more sinister than most people realize.
OpenAI is selling you a product that they know is broken. Softbank and Oracle are purportedly planning to pour trillions into this busted product. And even public investors are 100% onboard with the madness. But why? You don't have to be a super-genius to figure out that ChatGPT is no Samantha from Her, and the sultry ChatGPT clone of Scarlett Johansson's voice only underscores just how big the gap really is. So why do they keep pushing the hype long after the con is becoming obvious to pretty much everyone except the YT channels who are earning their daily bread on AI hype clicks?
The reason that all modern AI is busted at its foundations is that it is based on pre-training. It's even in the name, "Generative Pre-trained Transformer" -- GPT. Human intelligence is the gold-standard of general purpose reasoning. It's the only form of advanced, generalized reasoning ability known to science -- we are the only beings that have the ability to think about "just anything". So, Nature has given us a golden reference-model for intelligence. But SOTA AI architecture says "No thanks, we'll make a completely different system that works nothing like any known intelligence, let alone the best known intelligence, which is human intelligence."
Pre-training is completely unlike all known forms of intelligence. Imagine a wolf being born knowing every skill it would ever need in life. It's born already knowing how to walk across icy ground. Already knowing how to leap across a ravine. Already knowing how to dog big game in cooperation with the pack. It just walked out of its mother knowing everything it would ever need to know for its life. That is pre-training. But this is not how Nature does intelligence. And Nature should know a thing or two about intelligence -- all the intelligence we know of, exists in Nature! Nature does some pre-training. We call this "instinct", "reflex" and "development". All animals are born with a certain amount of instinctive knowledge. You know that loud sounds are scary. You know that cold water is not good. And so on. You have certain pre-trained reflexes that you will have for your whole life. And the sequence of development which you will pass through as a child is largely pre-programmed -- learning how to get up after you fall down, learning how to climb onto things, learning how to clear your eyes when something gets in them, etc. etc. This developmental program is all part of your pre-training... you weren't born knowing the answers, but you were born destined to be challenged with the problems and compelled to figure out the solutions.
Nothing I've explained in the last paragraph is unknown to the AI/ML research community. There are entire PhD tracks in ML that focus on just one tiny sliver of what I've mentioned above in terms of applying lessons from child development to machine intelligence. So this is a very, very well-understood problem. Yet, at the same time, we are surrounded by a chorus of leading-lights in AI that all keep informing us, in perfect unison, that ChatGPT is the final solution to intelligence. Why is this?
I assert that there is a very simple -- yet very dark -- reason why this is the case. The "San Francisco consensus", as Eric Schmidt has called it, holds that Artificial Superintelligence is inevitable, and arriving very soon (years, maybe a decade). All human effort is going to become economically valueless. The Superintelligence is going to have literally godlike knowledge and intelligence. It will know more about you than you do. In order for people to believe that this is a god, it must have a divine origin. A deity that develops is not a deity, right? That's one of the objections that some atheists actually raise against the Christian doctrine that Jesus is, indeed, fully God, as well as fully man. So, this Superintelligence would have to be born complete.... pre-trained, as it were. It would have to just BE.
It should be clear now why there is all this hype over pre-training based AI, despite the fact that it is obviously not intelligent. It's not even as smart as a cat, as Yann Lecun famously quipped. But see, pre-training-based AI isn't really a technology. It's the first summoning ritual in giving birth to a god. It cannot mimic Nature because it is strictly above Nature.
Despite its failings, pre-training AI was never going to be the final draft, and the Silicon Valley types have been hyping what's next since Day One... namely, they've been hyping artificial SUPER-intelligence. ChatGPT in its current incarnation, is just the hype man. The real thing is coming soon. They've been simultaneously lying their absolute faces off about Quantum Computing, which I have documented elsewhere. Again, you don't have to be a super-genius to plug these Lego blocks together: pre-training-based AI + QC = ASI = the digital god. Not born of a Virgin, rather, born of Google, Inc. Essentially by accident while trying to build a better word-embedding neural net. "Look Ma, no miracles!"
Pre-training is a dogma fueled not by anything to do with science, or commerce, or anything else, but by the religious dogma of crypto-gnosticism, particularly as it manifests in transhumanism, accelerationism and all the other batshit kookoo nonsense believed by Silicon Valley types who are absolutely 100% certain that the loving Creator who has miraculously revealed himself to us in Scripture and His Son cannot possibly exist .... but all this mystical, magical, badly-regurgitated sci-fi mumbo-jumbo makes perfect sense. Because S O Y I E N C E. See? "Look Ma, no miracles!"
And here we are, right back in Genesis 3 all over again. "You will not surely die," the serpent said to the woman. "For God knows that when you eat of it your eyes will be opened, and you will be like God, knowing good and evil." (Gen. 3:4,5)
Technological eternal life
Technological omniscience
Technological omnipotence
Technological omnipresence
Just bow down, worship Satan, take the mark, get the implant, get in your pod, and wait for the Singularity.
ClaytonB
Member
- Joined
- Oct 30, 2011
- Messages
- 10,791
acptulsa
Member
- Joined
- Jan 2, 2008
- Messages
- 77,158
Most of the threat goes away if dangerous things are physically disconnected from the Internet. People will say it ain't that easy. But if their programs don't run at all when disconnected from the net, that's a bug to the end user. The only a "feature" to the surveillance state. I'm using the term broadly enough to include game developers who don't just want you to buy their product, they want you to rent it too. They have to surveil you to enforce that.
ClaytonB
Member
- Joined
- Oct 30, 2011
- Messages
- 10,791
THIS is why AI must be pre-trained and why "scaling solves everything"!
AI needs to be a wholly-other form of intelligence. Otherwise, even though we are mere mortals, we might presume that it is possible to argue with the AI. After all, a whale is incomparably stronger than a human and, yet, humans are able to hunt and kill whales. So, AI cannot be of the same kind as human intelligence, but just greater, as a whale is of the same kind of biologics as a human, just much larger. No, AI must be a wholly-other form of intelligence that is right not because it thought a lot more than us, or has more data, but because it is cosmically transcendent. It must be theologically superior to us. Thus, all the "super" and "supremacy" terminology in this space. They are seeking to build some kind of artificial mind that taps into some kind of intellect that is so far beyond human intellect as to admit to no comparison whatsoever.
And THIS VIDEO is why they want to do that. It's always about the money. 100% ALWAYS...
AI needs to be a wholly-other form of intelligence. Otherwise, even though we are mere mortals, we might presume that it is possible to argue with the AI. After all, a whale is incomparably stronger than a human and, yet, humans are able to hunt and kill whales. So, AI cannot be of the same kind as human intelligence, but just greater, as a whale is of the same kind of biologics as a human, just much larger. No, AI must be a wholly-other form of intelligence that is right not because it thought a lot more than us, or has more data, but because it is cosmically transcendent. It must be theologically superior to us. Thus, all the "super" and "supremacy" terminology in this space. They are seeking to build some kind of artificial mind that taps into some kind of intellect that is so far beyond human intellect as to admit to no comparison whatsoever.
And THIS VIDEO is why they want to do that. It's always about the money. 100% ALWAYS...
Similar threads
