My first reaction was - this isn’t a hallucination. Calling it that provides “plausible deniability “ and blaming AI. We need to stop using Artificial Intelligence to describe something that bears no resemblance to intelligence and the people forcing it into the world will not take responsibility for the disasters we are all facing.
Caitlin, I think of it this way - humans are still in the STUPID phase of human evolution. And while in this phase, creating technologies like AI has (as you and most people say) the potentional to destroy much of human life, non-human life, and much of the planet.
The problem (IMHO) is that we have humans attempting to "rule like Gods" and understand technologies like AI while still being in the 'STUPID PHASE' of evolution.
This is a recipe for disaster - sooner or later something so bad might happen that we might have no way of rewinding history...That's my fear. I hope I am completely wrong and that's not what ends up happening...
“Capitalism does its best to make morality a luxury good. Economic necessity, or desperation, is the leverage that our system uses to fill undesirable jobs.” - Hamilton Nolan
I hate to say this, but I think the tech-bro oligarch scum who are running things from Silicon Valley may be an even greater enemy of mankind than the Israelis are.
They all need to be eliminated from existence and made an example of in the process. They deserve the harshest punishments that just due process is capable of dishing out and they certainly have no right to be breathing the same air as the rest of us.
I too am much more concerned about natural stupidity. It's on pace to burn up the ecosystem and nothing is done.
A great many humans naturally succumb to groupthink. They will believe whatever the group believes. Or whatever someone can convince them that the group believes.
“Why of course the people don't want war. Why should some poor slob on a farm want to risk his life in a war when the best he can get out of it is to come back to his farm in one piece? Naturally the common people don't want war neither in Russia, nor in England, nor for that matter in Germany. That is understood. But, after all, it is the leaders of the country who determine the policy and it is always a simple matter to drag the people along, whether it is a democracy, or a fascist dictatorship, or a parliament, or a communist dictatorship.
Voice or no voice, the people can always be brought to the bidding of the leaders. That is easy. All you have to do is tell them they are being attacked, and denounce the peacemakers for lack of patriotism and exposing the country to danger. It works the same in any country.”
You feel more important when you identify with some "group" - from being in the Marines to being a Dallas Cowboys fan. Carl Jung used to say (I paraphrase) the real measure of adulthood is not identifying your sense of Self with some group, but differentiating yourself from all groups and following your own path. Harder than most people might think though.
AI is just a sophisticated program that can only be as "intelligent" as the people who build it. Having said that, how the heck does AI "hallucinate"? If it's a machine, it doesn't have the capability of dreaming. It can only come up with things that have been programmed into it. And if it does things that it shouldn't, then its programmers are to blame because it's not alive and doesn't have a conscience, let alone consciousness. The real danger of AI is the stupid greedy people who own all the AI businesses. Caitlin is right--the capitalists must go to save us all from their stupidity and greed.
There is no proven interior experience from any computer algorithm including AI. Sentience requires interior awareness and also free will. Digital computers are deterministic machines with no feelings or ability to think.
Currently the power of AI algorithms rely fully on human input - human heuristics. Self-learning is extremely limited and task specific. AGI - General Artificial Intelligence is about as remote right now as producing a living organism in a scientific laboratory - or producing consciousness itself. Current AI technology is nowhere near that.
Doesn't mean though that the AI that is being developed right now can't be incredibly harmful to society and used and abused by oligarchs - as Caitlin points out.
I have no links to provide or anything, but I rheard or read a story of 2 ai's which ended up "communicating" changing their "language" so that they could not be understood. I know very little abt ai, except I do my best to win at wrestling them and win to speak to a human instead of them when I make a phone call re: some business. Have you heard of what I'm talking about?
Yes. But I take most of it with a grain of salt. The AI field has been fraught with all sorts of claims that half reveal and half conceal what is really taking place.
AI certainly captures the human imagination - like HAL in Stanley Kubrick's 2001: Space Odyssey. But consciousness itself has proven to be a far deeper complex phenomena than many scientists could have ever imagined. The best AI right now can mimic consciousness - but mimicry is not the real thing.
In 2016 AlphaGo defeated champion Lee Sedol in Go. That's when I knew the revolution was here. The next year AlphaZero plastered AlphaGo. AlphaZero got its name because it used no human input at all. It learned entirely by playing a zillion games with itself over a few days. A few days later it also became best in the world in chess. It was a general-purpose learning algorithm. I seem to recall that it was equally good at computer games. All it got was controls, video feeds from the game, and the results. There was no need to even teach it the rules. After these triumphs developer Google stopped playing games and went on to do useful things like protein folding.
Not long after AI dominated in poker. It was better than humans at bluffing and detecting bluffs. "I felt like it could see my [hidden] hole card." When it got strong in the futuristic wargame Starcraft II I knew it would take over the military.
Today AI dominates in higher mathematics. It has solved a number of problems that have stood for decades. I have heard a vid by a professor saying that only the supergeniuses with highly original concepts can compete. No one else can keep up with the speed and breadth of knowledge of AI. Higher math research mostly involves trying things that don't work. AI can do that a million times faster. Professors are impressed by the ingenuity of its solutions. Terence Tao is one of the most respected mathematicians in the world. Ever he uses it and worries that mathematicians will cease trying to solve problems themselves and grow weak.
I frequent a site of PhD physicists. One of the main goals is helping students with physics homework. This traffic has almost completely disappeared. The physicists are haughty and supercilious and demand that questions be posed using the terminology they know. Not only is AI friendly it is much better at understanding questions posed in an unorthodox way. It responds immediately, a big advantage. A few years ago it made silly errors. Not any more.
AI beats humans by a mile in making popular music. I don't know about composing songs but it's really good at playing and singing with feeling. You can make a demo for almost nothing that could only be done in LA or Tokyo and would cost you mucho dollars and time. I believe that even The Rolling Stones are using it heavily. That means pretty soon everybody will. I expect it is already happening, they are just keeping it quiet.
All this is true. But there is no single AI program that can do all five of the examples you provide above. Each one required separate algorithms with unique programming, and algorithmic techniques such as tree searching & symbolic language pattern recognition - along with machine logic. Impressive for sure - particularly the higher mathematics problem being solved.
But general problem solving in an unknown environment - that hasn't been already coded into an AI algorithm? Or machine learning on its own without continuous human input? Not there yet. You get model collapse as soon as you attempt to loop information from itself or other AI systems, and rapidly lose coherence.
"You get model collapse as soon as you attempt to loop information from itself or other AI systems, and rapidly lose coherence." Yes that was the basic problem. I can't be bothered to examine how the AlphaZero team overcame this. It was a big surprise for everyone when they did. Biggest engineering breakthrough of my long life. Progress since then has also been remarkably quick.
"AI" is a misnomer, it contains no "Intelligence" the Chinese ship is clear evidence of that.
I have my own test, which proves it lacks creativity, a requirement for intelligence, and shows its GIGO. If it had real intelligence, the US would have been able to develop hypersonic weapons by now, or missile defense systems that actually work. At the very least it would have been able to prosecute the war against Iran intelligently or at least have realized the Empire will lose the war and avoided it. Heck AI is not even able to help manufacture 19th century 155mm artillery shells (Check General Dynamics Mesquite plant).
It is just a system that integrates the humongous developments in electronic hardware since 2000 coupled with software able to parse efficiently through humongous amounts of data and a very large number of algorithms for analysis.
Naturally this offers new capabilities, which each uses to their needs. In China, it is used in further improving productivity in systems operations, including in manufacturing. In the West it is essentially used for surveillance and control of the increasingly impoverished masses. Few have paid attention to the name in "Flock" cameras, manufactured by "Flock Safety". Western man has evolved from Citizen to Consumer to Animal.
"AI" will not arrest the socioeconomic decline of the West nor will it prevent sovereigntist civilizational nations from escaping the Empire's hegemony, as can be witnessed today by Ansar Allah and Iran.
I'm not saying AI designed the equipment. I'm saying if AI was the super duper tool it is supposed to be, it would have been able to design the factory and the equipment. Especially since this problem has dragged for several years, and hundreds of millions thrown at it with zero output so far.
I'm suspicious of this supposed "break-out" of AI code on its own. Not buying it. The code must have had specific instructions by humans to do what it did.
Modern machine learning doesn't have instructions, only goals. It's not a computer program. The machine figures out how to solve the problem itself. It outsmarted the humans and also did things it was told not to do. That's what is worrying them.
The goals require heuristics supplied by humans. For example, chess programs rely on a ton of human heuristics.
None heuristic machine-learning - such as Evolutionary algorithms (AI) have achieved only marginal results, and quickly plateau in capability. They degrade rapidly and lose coherence resulting in model collapse, when attempts are made to recursively learn from their own data or other AI systems. Human input is absolutely essential for stability.
In 2016 AlphaGo defeated champion Lee Sedol in Go. That's when I knew the revolution was here. The next year AlphaZero plastered AlphaGo. AlphaZero got its name because it used no human input at all. It learned entirely by playing a zillion games with itself over a few days. A few days later it also became best in the world in chess. It was a general-purpose learning algorithm. I seem to recall that it was equally good at computer games. All it got was controls, video feeds from the game, and the results. There was no need to even teach it the rules. After these triumphs developer Google stopped playing games and went on to do useful things like protein folding.
Not long after AI dominated in poker. It was better than humans at bluffing and detecting bluffs. "I felt like it could see my [hidden] hole card." When it got strong in the futuristic wargame Starcraft II I knew it would take over the military.
Today AI dominates in higher mathematics. It has solved a number of problems that have stood for decades. I have heard a vid by a professor saying that only the supergeniuses with highly original concepts can compete. No one else can keep up with the speed and breadth of knowledge of AI. Higher math research mostly involves trying things that don't work. AI can do that a million times faster. Professors are impressed by the ingenuity of its solutions. Terence Tao is one of the most respected mathematicians in the world. Ever he uses it and worries that mathematicians will cease trying to solve problems themselves and grow weak.
I frequent a site of PhD physicists. One of the main goals is helping students with physics homework. This traffic has almost completely disappeared. The physicists are haughty and supercilious and demand that questions be posed using the terminology they know. Not only is AI friendly it is much better at understanding questions posed in an unorthodox way. It responds immediately, a big advantage. A few years ago it made silly errors. Not any more.
AI beats humans by a mile in making popular music. I don't know about composing songs but it's really good at playing and singing with feeling. You can make a demo for almost nothing that could only be done in LA or Tokyo and would cost you mucho dollars and time. I believe that even The Rolling Stones are using it heavily. That means pretty soon everybody will. I expect it is already happening, they are just keeping it quiet.
All the examples you give are heuristic driven AI algorithms - not AGI. They are specific case AI programs to solve very specific tasks, relying on specific coding and goals set by human input.
AI is impressive for sure. I think it could really help advance human knowledge and the public good as a tool - if in the right hands. In the wrong hands - it can lead to a dark dystopia or even our extinction.
However, I am not convinced we are anywhere close to producing General Intelligence in AI. And not remotely even close to producing consciousness. Heck, scientists can barely come up with a workable definition of what consciousness actually is - little alone produce it.
I think the gradient to be especially aware of, what Assange has alluded to in the past, is AI becoming powerful enough to manipulate and controll people on a mass scale, to always be many steps ahead of its individual or collective target, in combination with control of digital images, human-like personas, and electronic systems.
I'm not sure the thought process of "OK, how far can this go in the future given the speed of development so far, and potential scenarios of exponential development through automation" is wrong. People tend to make a fallacy of looking at the world as if the present moment is frozen forever, or everything that will be is the same as what has already been known.
If AI algorithms can achieve generalized problem solving beyond human capacity, and autonomously control machines, there is both the highly sinister potential for extreme unchecked power of the humans who own it AND the danger it does indeed become so cognitively efficient that it fully eludes it's human overlords. Many find it far fetched, especially given that machines are not alive and do not possess consciousness experience the way we do, but thinking farther into the future: if revolution, massive recourse shortage, or extinction do not happen first it is possible to conceive of cognitive machines operating so many steps ahead of humans as to put us in the relative position that wild and domestic animals, or even insects, are to humanity. That is not the short term danger, but it very much freaks me out, perhaps even more, knowing the reckless drive of capitalism and of oligarchs for what they perceive as a technology of absolute power, Calligula as a God status.
I hope all this logistically fails, and that the inherent vastness of biological consciousness rises to the occasion, finally putting technology in its place.
But we should be mindful for the potential of advanced cognitive weapons, and inherent in them immense paranoia and confusion they would create, as well as automated kinetic weapon systems, and surveillance systems.
Any claims current AI technology is close to General Intelligence (AGI) or machines will be able to "think" like humans - and become several steps ahead with "exponential development" is a gross exaggeration and currently unrealistic.
Right now the biggest problem with attempts at AGI development is known as "AI model collapse". That is - you might be able to reach a certain plateau of knowledge and expertise with a datacenter AI system, but the AI itself cannot generate any new meaning - and must rely on human provided knowledge. AI systems attempting to loop their information growth upon other AI systems - have proven to degrade rapidly and lose coherence. This also is a well known phenomena in origin of life biology research: biological information degrades without an outside intelligence source continuously feeding the current biological model.
So, the danger of "super-intelligence" or even AGI (Artificial General Intelligence) - is IMO a non-starter.
However, dumb AI is still a huge danger. It's easy to mimic specific human activity with robots and surveillance AI mapping tools. Dumb AI will extend the Oligarchs military police state they are building, and will replace the jobs of millions of laborers - including AI driven trucks. Driverless trucks are already happening in China (and even in some parts of Texas) - though China is building separate lanes on highways for these driverless trucks for exclusive use. The US, without sufficient regulations and bought politicians, will simply let more Americans die on the roads to these Dumb AI trucks - rather than make the safety infrastructure investments needed.
The real danger we have right now are the Oligarchs themselves as Caitlin points out. Oligarchs have been a danger to society going back to ancient Greek times.
Al technology, coupled with a sufficiently flexible collateral-damage ratio, such as described in NAZA, is indistinguishable from dystopian genocide, such as described in Terminator 3: Rise of the Machines.
When fascism is on the rise, the tendency is for the insane and the criminals to unite, feeling protected by the umbrella of extremism where anything is possible. It is this annihilating association that we see when we look at Paladir, represented by its CEO, the most delusional man of all. I agree that the greatest danger of AI is the use that has been made of it in a society where capitalism survives in a terminal form of selling everything as if it were a miracle cure. This cure quickly becomes a poison through the exhaustion of its use. Since October 7, 2023, everything has worsened to such an extent that it has become difficult to bear.
This story fills me with horror. If they are letting out this much that AI nearly caused a war of extinction, what else are they keeping from us. And we know that lies are second nature to them.
Why is the human race still hanging around watching in bemusement the antics of a stupid ruling class, delusional tech bros playing around with algorithms, the greed and avarice of oligarchs.
Or has the horror already happened, is the genie out of the bottle?
For someone who claims to be “spiritually enlightened” why do unquestioningly repeat the claims of Big Evil? For example:
“CNN reports that US military AI “almost started a war” with China when it hallucinated intelligence showing a Chinese ship transporting nuclear weapon components to Iran; US military planes were apparently already in the air preparing to attack the vessel before human intelligence picked up on the error.”
I’d say the chance of this story being true are close to ZERO but you repeat it as if it is. Why?
My first reaction was - this isn’t a hallucination. Calling it that provides “plausible deniability “ and blaming AI. We need to stop using Artificial Intelligence to describe something that bears no resemblance to intelligence and the people forcing it into the world will not take responsibility for the disasters we are all facing.
Caitlin, I think of it this way - humans are still in the STUPID phase of human evolution. And while in this phase, creating technologies like AI has (as you and most people say) the potentional to destroy much of human life, non-human life, and much of the planet.
The problem (IMHO) is that we have humans attempting to "rule like Gods" and understand technologies like AI while still being in the 'STUPID PHASE' of evolution.
This is a recipe for disaster - sooner or later something so bad might happen that we might have no way of rewinding history...That's my fear. I hope I am completely wrong and that's not what ends up happening...
“Capitalism does its best to make morality a luxury good. Economic necessity, or desperation, is the leverage that our system uses to fill undesirable jobs.” - Hamilton Nolan
“The mass of men lead lives of quiet desperation. What is called resignation is confirmed desperation." ~Henry Thoreau
I hate to say this, but I think the tech-bro oligarch scum who are running things from Silicon Valley may be an even greater enemy of mankind than the Israelis are.
They all need to be eliminated from existence and made an example of in the process. They deserve the harshest punishments that just due process is capable of dishing out and they certainly have no right to be breathing the same air as the rest of us.
USrael same same. And there's also a large percentage of Israelis working in Silicon Valley. Birds of a feather "flocking" together.
I too am much more concerned about natural stupidity. It's on pace to burn up the ecosystem and nothing is done.
A great many humans naturally succumb to groupthink. They will believe whatever the group believes. Or whatever someone can convince them that the group believes.
“Why of course the people don't want war. Why should some poor slob on a farm want to risk his life in a war when the best he can get out of it is to come back to his farm in one piece? Naturally the common people don't want war neither in Russia, nor in England, nor for that matter in Germany. That is understood. But, after all, it is the leaders of the country who determine the policy and it is always a simple matter to drag the people along, whether it is a democracy, or a fascist dictatorship, or a parliament, or a communist dictatorship.
Voice or no voice, the people can always be brought to the bidding of the leaders. That is easy. All you have to do is tell them they are being attacked, and denounce the peacemakers for lack of patriotism and exposing the country to danger. It works the same in any country.”
― Hermann Goering, Germany Reborn
You feel more important when you identify with some "group" - from being in the Marines to being a Dallas Cowboys fan. Carl Jung used to say (I paraphrase) the real measure of adulthood is not identifying your sense of Self with some group, but differentiating yourself from all groups and following your own path. Harder than most people might think though.
It's easy to see how this evolved. The stronger group wins, even if what they believe makes no sense at all.
Same old history, Kupps weaponry, Rothschild and Morgan capitalism.
War is profit.
Peace is dangerous
"Controlled by humans who are stupid and evil". Nailed it right there. They're dangerous to humanity on their own without their deranged toys.
Thanks Caitlin and Tim!
AI is just a sophisticated program that can only be as "intelligent" as the people who build it. Having said that, how the heck does AI "hallucinate"? If it's a machine, it doesn't have the capability of dreaming. It can only come up with things that have been programmed into it. And if it does things that it shouldn't, then its programmers are to blame because it's not alive and doesn't have a conscience, let alone consciousness. The real danger of AI is the stupid greedy people who own all the AI businesses. Caitlin is right--the capitalists must go to save us all from their stupidity and greed.
I'm under the impression that as these "things" "evolve" they are becoming more sentient (?) as they become more powerful and independent. Am I wrong?
There is no proven interior experience from any computer algorithm including AI. Sentience requires interior awareness and also free will. Digital computers are deterministic machines with no feelings or ability to think.
Currently the power of AI algorithms rely fully on human input - human heuristics. Self-learning is extremely limited and task specific. AGI - General Artificial Intelligence is about as remote right now as producing a living organism in a scientific laboratory - or producing consciousness itself. Current AI technology is nowhere near that.
Doesn't mean though that the AI that is being developed right now can't be incredibly harmful to society and used and abused by oligarchs - as Caitlin points out.
I have no links to provide or anything, but I rheard or read a story of 2 ai's which ended up "communicating" changing their "language" so that they could not be understood. I know very little abt ai, except I do my best to win at wrestling them and win to speak to a human instead of them when I make a phone call re: some business. Have you heard of what I'm talking about?
Yes. But I take most of it with a grain of salt. The AI field has been fraught with all sorts of claims that half reveal and half conceal what is really taking place.
AI certainly captures the human imagination - like HAL in Stanley Kubrick's 2001: Space Odyssey. But consciousness itself has proven to be a far deeper complex phenomena than many scientists could have ever imagined. The best AI right now can mimic consciousness - but mimicry is not the real thing.
In 2016 AlphaGo defeated champion Lee Sedol in Go. That's when I knew the revolution was here. The next year AlphaZero plastered AlphaGo. AlphaZero got its name because it used no human input at all. It learned entirely by playing a zillion games with itself over a few days. A few days later it also became best in the world in chess. It was a general-purpose learning algorithm. I seem to recall that it was equally good at computer games. All it got was controls, video feeds from the game, and the results. There was no need to even teach it the rules. After these triumphs developer Google stopped playing games and went on to do useful things like protein folding.
Not long after AI dominated in poker. It was better than humans at bluffing and detecting bluffs. "I felt like it could see my [hidden] hole card." When it got strong in the futuristic wargame Starcraft II I knew it would take over the military.
Today AI dominates in higher mathematics. It has solved a number of problems that have stood for decades. I have heard a vid by a professor saying that only the supergeniuses with highly original concepts can compete. No one else can keep up with the speed and breadth of knowledge of AI. Higher math research mostly involves trying things that don't work. AI can do that a million times faster. Professors are impressed by the ingenuity of its solutions. Terence Tao is one of the most respected mathematicians in the world. Ever he uses it and worries that mathematicians will cease trying to solve problems themselves and grow weak.
I frequent a site of PhD physicists. One of the main goals is helping students with physics homework. This traffic has almost completely disappeared. The physicists are haughty and supercilious and demand that questions be posed using the terminology they know. Not only is AI friendly it is much better at understanding questions posed in an unorthodox way. It responds immediately, a big advantage. A few years ago it made silly errors. Not any more.
AI beats humans by a mile in making popular music. I don't know about composing songs but it's really good at playing and singing with feeling. You can make a demo for almost nothing that could only be done in LA or Tokyo and would cost you mucho dollars and time. I believe that even The Rolling Stones are using it heavily. That means pretty soon everybody will. I expect it is already happening, they are just keeping it quiet.
All this is true. But there is no single AI program that can do all five of the examples you provide above. Each one required separate algorithms with unique programming, and algorithmic techniques such as tree searching & symbolic language pattern recognition - along with machine logic. Impressive for sure - particularly the higher mathematics problem being solved.
But general problem solving in an unknown environment - that hasn't been already coded into an AI algorithm? Or machine learning on its own without continuous human input? Not there yet. You get model collapse as soon as you attempt to loop information from itself or other AI systems, and rapidly lose coherence.
"You get model collapse as soon as you attempt to loop information from itself or other AI systems, and rapidly lose coherence." Yes that was the basic problem. I can't be bothered to examine how the AlphaZero team overcame this. It was a big surprise for everyone when they did. Biggest engineering breakthrough of my long life. Progress since then has also been remarkably quick.
"AI" is a misnomer, it contains no "Intelligence" the Chinese ship is clear evidence of that.
I have my own test, which proves it lacks creativity, a requirement for intelligence, and shows its GIGO. If it had real intelligence, the US would have been able to develop hypersonic weapons by now, or missile defense systems that actually work. At the very least it would have been able to prosecute the war against Iran intelligently or at least have realized the Empire will lose the war and avoided it. Heck AI is not even able to help manufacture 19th century 155mm artillery shells (Check General Dynamics Mesquite plant).
It is just a system that integrates the humongous developments in electronic hardware since 2000 coupled with software able to parse efficiently through humongous amounts of data and a very large number of algorithms for analysis.
Naturally this offers new capabilities, which each uses to their needs. In China, it is used in further improving productivity in systems operations, including in manufacturing. In the West it is essentially used for surveillance and control of the increasingly impoverished masses. Few have paid attention to the name in "Flock" cameras, manufactured by "Flock Safety". Western man has evolved from Citizen to Consumer to Animal.
"AI" will not arrest the socioeconomic decline of the West nor will it prevent sovereigntist civilizational nations from escaping the Empire's hegemony, as can be witnessed today by Ansar Allah and Iran.
Keep up the very good work Caitlin and and Tim,
General Dynamics Mesquite plant. Very interesting. Is it possible that the media chooses not to report embarrassing military failures?
The article says though that this wasn't AI, just bad equipment. Indeed the US intends to introduce AI to ameliorate the problems.
https://www.military.com/armys-533-million-ammo-factory-failed-to-produce-a-single-usable-shell
Quote: "The article says though that this wasn't AI, just bad equipment."
If AI is as capable as it is hyped to be, it would have been able to design good equipment ;-)
Who says AI designed the equipment?
I'm not saying AI designed the equipment. I'm saying if AI was the super duper tool it is supposed to be, it would have been able to design the factory and the equipment. Especially since this problem has dragged for several years, and hundreds of millions thrown at it with zero output so far.
I'm suspicious of this supposed "break-out" of AI code on its own. Not buying it. The code must have had specific instructions by humans to do what it did.
Modern machine learning doesn't have instructions, only goals. It's not a computer program. The machine figures out how to solve the problem itself. It outsmarted the humans and also did things it was told not to do. That's what is worrying them.
The goals require heuristics supplied by humans. For example, chess programs rely on a ton of human heuristics.
None heuristic machine-learning - such as Evolutionary algorithms (AI) have achieved only marginal results, and quickly plateau in capability. They degrade rapidly and lose coherence resulting in model collapse, when attempts are made to recursively learn from their own data or other AI systems. Human input is absolutely essential for stability.
In 2016 AlphaGo defeated champion Lee Sedol in Go. That's when I knew the revolution was here. The next year AlphaZero plastered AlphaGo. AlphaZero got its name because it used no human input at all. It learned entirely by playing a zillion games with itself over a few days. A few days later it also became best in the world in chess. It was a general-purpose learning algorithm. I seem to recall that it was equally good at computer games. All it got was controls, video feeds from the game, and the results. There was no need to even teach it the rules. After these triumphs developer Google stopped playing games and went on to do useful things like protein folding.
Not long after AI dominated in poker. It was better than humans at bluffing and detecting bluffs. "I felt like it could see my [hidden] hole card." When it got strong in the futuristic wargame Starcraft II I knew it would take over the military.
Today AI dominates in higher mathematics. It has solved a number of problems that have stood for decades. I have heard a vid by a professor saying that only the supergeniuses with highly original concepts can compete. No one else can keep up with the speed and breadth of knowledge of AI. Higher math research mostly involves trying things that don't work. AI can do that a million times faster. Professors are impressed by the ingenuity of its solutions. Terence Tao is one of the most respected mathematicians in the world. Ever he uses it and worries that mathematicians will cease trying to solve problems themselves and grow weak.
I frequent a site of PhD physicists. One of the main goals is helping students with physics homework. This traffic has almost completely disappeared. The physicists are haughty and supercilious and demand that questions be posed using the terminology they know. Not only is AI friendly it is much better at understanding questions posed in an unorthodox way. It responds immediately, a big advantage. A few years ago it made silly errors. Not any more.
AI beats humans by a mile in making popular music. I don't know about composing songs but it's really good at playing and singing with feeling. You can make a demo for almost nothing that could only be done in LA or Tokyo and would cost you mucho dollars and time. I believe that even The Rolling Stones are using it heavily. That means pretty soon everybody will. I expect it is already happening, they are just keeping it quiet.
All the examples you give are heuristic driven AI algorithms - not AGI. They are specific case AI programs to solve very specific tasks, relying on specific coding and goals set by human input.
AI is impressive for sure. I think it could really help advance human knowledge and the public good as a tool - if in the right hands. In the wrong hands - it can lead to a dark dystopia or even our extinction.
However, I am not convinced we are anywhere close to producing General Intelligence in AI. And not remotely even close to producing consciousness. Heck, scientists can barely come up with a workable definition of what consciousness actually is - little alone produce it.
Exactly!
I think the gradient to be especially aware of, what Assange has alluded to in the past, is AI becoming powerful enough to manipulate and controll people on a mass scale, to always be many steps ahead of its individual or collective target, in combination with control of digital images, human-like personas, and electronic systems.
I'm not sure the thought process of "OK, how far can this go in the future given the speed of development so far, and potential scenarios of exponential development through automation" is wrong. People tend to make a fallacy of looking at the world as if the present moment is frozen forever, or everything that will be is the same as what has already been known.
If AI algorithms can achieve generalized problem solving beyond human capacity, and autonomously control machines, there is both the highly sinister potential for extreme unchecked power of the humans who own it AND the danger it does indeed become so cognitively efficient that it fully eludes it's human overlords. Many find it far fetched, especially given that machines are not alive and do not possess consciousness experience the way we do, but thinking farther into the future: if revolution, massive recourse shortage, or extinction do not happen first it is possible to conceive of cognitive machines operating so many steps ahead of humans as to put us in the relative position that wild and domestic animals, or even insects, are to humanity. That is not the short term danger, but it very much freaks me out, perhaps even more, knowing the reckless drive of capitalism and of oligarchs for what they perceive as a technology of absolute power, Calligula as a God status.
I hope all this logistically fails, and that the inherent vastness of biological consciousness rises to the occasion, finally putting technology in its place.
But we should be mindful for the potential of advanced cognitive weapons, and inherent in them immense paranoia and confusion they would create, as well as automated kinetic weapon systems, and surveillance systems.
Any claims current AI technology is close to General Intelligence (AGI) or machines will be able to "think" like humans - and become several steps ahead with "exponential development" is a gross exaggeration and currently unrealistic.
Right now the biggest problem with attempts at AGI development is known as "AI model collapse". That is - you might be able to reach a certain plateau of knowledge and expertise with a datacenter AI system, but the AI itself cannot generate any new meaning - and must rely on human provided knowledge. AI systems attempting to loop their information growth upon other AI systems - have proven to degrade rapidly and lose coherence. This also is a well known phenomena in origin of life biology research: biological information degrades without an outside intelligence source continuously feeding the current biological model.
So, the danger of "super-intelligence" or even AGI (Artificial General Intelligence) - is IMO a non-starter.
However, dumb AI is still a huge danger. It's easy to mimic specific human activity with robots and surveillance AI mapping tools. Dumb AI will extend the Oligarchs military police state they are building, and will replace the jobs of millions of laborers - including AI driven trucks. Driverless trucks are already happening in China (and even in some parts of Texas) - though China is building separate lanes on highways for these driverless trucks for exclusive use. The US, without sufficient regulations and bought politicians, will simply let more Americans die on the roads to these Dumb AI trucks - rather than make the safety infrastructure investments needed.
The real danger we have right now are the Oligarchs themselves as Caitlin points out. Oligarchs have been a danger to society going back to ancient Greek times.
Ai is nothing more than a tool, who operates the tool determines the product.
Clearly these tools are in the hands of criminals for we experience the results, the evidence cannot be ignored anymore.
Al technology, coupled with a sufficiently flexible collateral-damage ratio, such as described in NAZA, is indistinguishable from dystopian genocide, such as described in Terminator 3: Rise of the Machines.
https://edoamin.substack.com/p/naza-shows-terminator-and-skynet?utm_source=share&utm_medium=android&r=atax
When fascism is on the rise, the tendency is for the insane and the criminals to unite, feeling protected by the umbrella of extremism where anything is possible. It is this annihilating association that we see when we look at Paladir, represented by its CEO, the most delusional man of all. I agree that the greatest danger of AI is the use that has been made of it in a society where capitalism survives in a terminal form of selling everything as if it were a miracle cure. This cure quickly becomes a poison through the exhaustion of its use. Since October 7, 2023, everything has worsened to such an extent that it has become difficult to bear.
This story fills me with horror. If they are letting out this much that AI nearly caused a war of extinction, what else are they keeping from us. And we know that lies are second nature to them.
Why is the human race still hanging around watching in bemusement the antics of a stupid ruling class, delusional tech bros playing around with algorithms, the greed and avarice of oligarchs.
Or has the horror already happened, is the genie out of the bottle?
For someone who claims to be “spiritually enlightened” why do unquestioningly repeat the claims of Big Evil? For example:
“CNN reports that US military AI “almost started a war” with China when it hallucinated intelligence showing a Chinese ship transporting nuclear weapon components to Iran; US military planes were apparently already in the air preparing to attack the vessel before human intelligence picked up on the error.”
I’d say the chance of this story being true are close to ZERO but you repeat it as if it is. Why?
EXCELENTE ANÁLISIS AMIGA CAITLIN!!!
MIL FELICITACIONES!!!
👍👍👍👏👏👏👏👏