I don't know where to start on this bullshit. Some of the points - namely, the ones that involve the power structure using the "AI panic" to push their own agendas (when do they ever not, as well as starting all these panics themselves in the first place) - are correct but the notion that any of these idiot programmers are correct in thi…
I don't know where to start on this bullshit. Some of the points - namely, the ones that involve the power structure using the "AI panic" to push their own agendas (when do they ever not, as well as starting all these panics themselves in the first place) - are correct but the notion that any of these idiot programmers are correct in thinking these current LLM are "sentient" in the same sense as the human brain is simply stupid.
"As can be seen, the AI here performs a stunning series of metaphorical abstractions that would stymy most humans. It was able to take a parable and, seemingly, analyzing its meaning in a philosopically abstract way, employed subtlety in extrapolating the concept out into a multi-layered explanation."
No, it did not. It clearly is perfectly well aware of the analyses that have been done by legions of Zen experts and simply regurgitated it in a way hardly different from what has been done many times before by said experts. The idiot programmers who have fed these things massive amounts of data WITHOUT KNOWING WHAT THE HELL THEY ARE FEEDING IT are why these things can "surprise" their own creators.
In other words, it's "Garbage In, Garbage Out" - only in this case it's not actually "garbage" but simply facts that their programmers alone don't remember feeding it - which the programmers themselves admit. Sure, they know the general datasets they fed in, but there's zero chance they understand the ramifications of those data sets or how those data sets interact either with each other or with the conceptual language models.
This is akin to someone reading a whole lot of books and then suddenly discovering that some stuff in one book is relevant to the stuff in some other book they don't even remember reading. This happens to me all the time - as well as to anyone else who has read extensively throughout their lives.
As for the notion that the Covid pandemic was created to provide a cover for this, that's just complete bullshit. There remains zero evidence - and I mean ZERO evidence - that Covid was either 1) man-made, 2) released from any lab accidentally or on purpose, either in China or elsewhere, or 3) that it was intended to enable imposing restrictions on civil rights. The latter is merely another example of what I referred to above: the power structures seizing every opportunity to increase their power by riding on ordinary and/or exceptional events.
Only conspiracy theory idiots interpret this as "foreknowledge" and then use that notion to justify their theories.
Read my lips: there is no one "in charge". There is no "Illuminati" running things because there aren't any human beings capable of doing so. Which is why everyone is worried about AI - because we all know human beings are stupid, ignorant, malicious and fearful and thus incapable of running their own lives, let alone anyone else's.
Go back and read Robert Anton Wilson's and Robert Shea's 1970's magnum opus, "Illuminati" which is where they devastated that whole notion, while simultaneously feeding into the very conspiracy theories they were lampooning.
The one real point Simplicius is correct about is how if these AIs are used to manipulate people then it won't matter if they are really "sentient" or not.
But this is an opportunity, not just a threat. The smart cyberpunk will be looking for ways to turn these programs into weapons for our own purposes.
I've followed the various AI crazes over the decades since the 1970s. Every twenty years or so there is another "AI craze", everyone pumps money into these things, then the craze dies down as corporations discover there is no magic money pump in them. They have value, but it's not enough to justify the hyper-investment craze. This one will peter out as well.
The main problem is that Large Language Models DO NOT reproduce how the human brain creates and processes concepts - because NO ONE KNOWS how that this done - as yet. Until ubiquitous nanotechnology is used to examine brain function on the molecular scale in real time at scale, it is unlikely we will discover how that is done in the near term. Until that is done, we will not have anything resembling "real AI"
So let's all cut the bullshit and simply try to make use of these useful tools in our own lives.
Spot on! I am a developer and have started to use these AI tools to help in my daily work. But as you say, this craze is just nuts. It's not magic and we haven't become gods.
Garbage in, garbage out still applies, even to AI.
AI has developed enormously in the last 20 years. The issue is not where it is today, but where it will be in 5 years, 10 years, 20 years, as AIs are used to teach and design successive generations of AI. AlphaGo, which is now the best Go player in the world and is now being applied to other domains, is programmed to learn (as are all AIs now being developed) and has proved it can innovate with moves unknown to and unrealized by the best Go players in the world.
In the case of AIs, GIGO refers to what they learn from. Some may end up with the right material to produce a psychopath. The fact that it was GIGO will be no consolation.
It could well be that what we need is a psychopath AI - provided it's a rational psychopath who understands who the real enemy is. (Psychopathy does not imply irrationality, even if most psychopaths are irrational - that's BS from the psychology profession which doesn't understand anything.)
Thank you for your concise observations. I have been saying the same things for years as a software engineer. The fact that you describe of feeding data into these software modules is indicative of what we commonly refer to in the profession as "Machine Learning" or ML. ML uses neural networks to determine best answers based in the same manner as well designed B+Tree structure but is nonetheless far more complex and capable.
That being said, you are correct to state that the majority of what we are seeing is not sentient AI. A sentient AI would not need so much data given to it but merely a base set so that it could learn what it needed on its own and then make its own decisions. Such an AI to be truly sentient would also have to exhibit an "awareness of self", which to date, has not been demonstrated by anything other than robots who can only mimic self-awareness as a result of its capabilities with conversation.
The most advanced AI research is being done in Japan as for some reason the Japanese appear to love this field of research. However, with their immense talents and capabilities, the best that Japanese researchers have been able to create has been sex-bots for the lonely. And none, to my knowledge have reached any level of self-awareness on their own.
What is frighteningly dangerous is not the development of AI but the use by stupid Humans who believe they have created such software and then apply it to real-world situations as Google and Tesla have done with their self-driving vehicles only to see them crash into things, kill pedestrians, and even the occupants. Now apply this garbage to military controls... and yes, we will have Armageddon... But not because a sentient AI believes that Humans are inferior but simply because it came up with the same whacked-out conclusions that we are seeing now...
You don't get it either. The conversation between lemoine and laMDA is pure fiction composed by a hack writer paid as part of a rollout of this garbage. The real AI has nothing to do with getting machines to produce pseudo philosophical text which "stuns" John and Jane Q Boomer when they coincidentally all "discover" it simultaneously because corporate masters want them to. AI is doing all sorts of very real things including learning to police and rule the internet. As an epiphenomenon of the entire project there are prose-writing bots, whose purpose is to hook the planet of tech users and employees and everyday people into using them for tasks ranging from birthday greetings to drafts of grant proposals to weapon tech specification manual composition so that the Cloud knows everything everyone is thinking and planning even better than it does today. The Big Big PR campaign Now as a result of these wholly manufactured letters of Musk and other so called experts has one purpose - to get you and everyone you know everywhere to start using the available bots and addict you. In addition to charge monthly fees - no longer for renting cloud services for saving files and editing documents and spreadsheets (everyone's spreadsheet estimates have been freely available for over a decade now - way to go, corporate spy agency!) with specialized software, but now to seize control of every spasm of imaginative effort on earth and turn it to the ends of those in power who own and run these things.
That's certainly possible. But of course the smart thing to do is to 1) not use these tools in a way that compromises your OPSEC, 2) use these tools to the degree possible in a "self-hosted" manner, and 3) don't use them for pointless crap like writing birthday greetings.
It's fine to use these tools for personal enjoyment or things like creating art for personal enjoyment or as promo items for a Youtube video or something. Who cares what some cretin uses these tools for? Not everyone on the planet is worth surveilling, and people who think they're under surveillance when in reality no one in power gives a rat's ass about them (except in the aggregate) are people who have a little much paranoia, instead of just enough.
As someone with a clue about computer security, I avoid using tools that are hosted in the cloud, such as password managers , or which are not open-source and capable of being reviewed. There are exceptions to that, based on how much effort I want to spend on implementing my own tools vs how much I think the PTB actually care what I'm doing, and I never allow my deepest and darkest plans and thoughts to leave my computer. Which is one reason I don't use Windows or Macs.
But until you're doing something that makes you a significant threat to the PTB, 99.9% of you are simply irrelevant and your own security is not under significant threat.
As for AI policing the Internet, I don't doubt that will be its eventual role. But it won't be any more perfect at it than the manual policing the CIA and NSA have been doing so far. And as to that, I'd say it's almost hardly needed, given that 99.9% of the population is already being fed, and believing in, so much conspiracy theory as is represented by the notions such as 1) Ukraine is beating Russia, 2) Covid was a Chinese lab leak, and 3) the 2020 election was stolen, and 4) Trump is their savior.
It's too easy to fall into the binary "either/or" stuff that is being fed people today, which is the primary method by which the PTB have been controlling people for the last several thousand years. Ascribing everything as one big conspiracy is the number one way that trick has been used.
Again, go read "Illuminatus" and its derivatives. Wilson and Shea had it down.
Well, that saves me writing my planned response ;-)
(Btw: As a SW developer I have heard since the early 90's that my work would soon be taken over by AI... here we are 30 years later, and AI still cannot create SW, it can only regurgitate what other programmers have done. Neat, surely, but hardly a thing to start worrying over, especially as I know that at least -and I am being generous here- 80% of all code on the internet is flawed in some way or other)
As a SW developer you must also realize that you only need one deluded moron to give these things an ability to make PUT/POST requests before they will start doing something dangerous, like corrupting Wikipedia articles or registering on any of the million web services to wreak havoc. "It can only regurgitate what other programmers have done". Step 1: Feed a lot of security vulnerabilities into it, Step 2: make it crawl the web in the search of these and exploit, Step 3: .... So even if your job might be safe, your bank account, your IoT devices, etc, etc, might very well be not. It doesn't need to be very smart to do some very real damage, just like a hoodlum with a knife might be dumb but that knife in your gut might kill you.
"I've followed the various AI crazes over the decades since the 1970s."
There was no Internet in the 70s, there was no computational capability and, most of all, the world wasn't so connected. If you give an ability for AI to make PUT/POST requests, as in change stuff on the Internet, I can see how it might start doing pretty significant damage, which also might not get discovered in time.
I don't know where to start on this bullshit. Some of the points - namely, the ones that involve the power structure using the "AI panic" to push their own agendas (when do they ever not, as well as starting all these panics themselves in the first place) - are correct but the notion that any of these idiot programmers are correct in thinking these current LLM are "sentient" in the same sense as the human brain is simply stupid.
"As can be seen, the AI here performs a stunning series of metaphorical abstractions that would stymy most humans. It was able to take a parable and, seemingly, analyzing its meaning in a philosopically abstract way, employed subtlety in extrapolating the concept out into a multi-layered explanation."
No, it did not. It clearly is perfectly well aware of the analyses that have been done by legions of Zen experts and simply regurgitated it in a way hardly different from what has been done many times before by said experts. The idiot programmers who have fed these things massive amounts of data WITHOUT KNOWING WHAT THE HELL THEY ARE FEEDING IT are why these things can "surprise" their own creators.
In other words, it's "Garbage In, Garbage Out" - only in this case it's not actually "garbage" but simply facts that their programmers alone don't remember feeding it - which the programmers themselves admit. Sure, they know the general datasets they fed in, but there's zero chance they understand the ramifications of those data sets or how those data sets interact either with each other or with the conceptual language models.
This is akin to someone reading a whole lot of books and then suddenly discovering that some stuff in one book is relevant to the stuff in some other book they don't even remember reading. This happens to me all the time - as well as to anyone else who has read extensively throughout their lives.
As for the notion that the Covid pandemic was created to provide a cover for this, that's just complete bullshit. There remains zero evidence - and I mean ZERO evidence - that Covid was either 1) man-made, 2) released from any lab accidentally or on purpose, either in China or elsewhere, or 3) that it was intended to enable imposing restrictions on civil rights. The latter is merely another example of what I referred to above: the power structures seizing every opportunity to increase their power by riding on ordinary and/or exceptional events.
Only conspiracy theory idiots interpret this as "foreknowledge" and then use that notion to justify their theories.
Read my lips: there is no one "in charge". There is no "Illuminati" running things because there aren't any human beings capable of doing so. Which is why everyone is worried about AI - because we all know human beings are stupid, ignorant, malicious and fearful and thus incapable of running their own lives, let alone anyone else's.
Go back and read Robert Anton Wilson's and Robert Shea's 1970's magnum opus, "Illuminati" which is where they devastated that whole notion, while simultaneously feeding into the very conspiracy theories they were lampooning.
The one real point Simplicius is correct about is how if these AIs are used to manipulate people then it won't matter if they are really "sentient" or not.
But this is an opportunity, not just a threat. The smart cyberpunk will be looking for ways to turn these programs into weapons for our own purposes.
I've followed the various AI crazes over the decades since the 1970s. Every twenty years or so there is another "AI craze", everyone pumps money into these things, then the craze dies down as corporations discover there is no magic money pump in them. They have value, but it's not enough to justify the hyper-investment craze. This one will peter out as well.
The main problem is that Large Language Models DO NOT reproduce how the human brain creates and processes concepts - because NO ONE KNOWS how that this done - as yet. Until ubiquitous nanotechnology is used to examine brain function on the molecular scale in real time at scale, it is unlikely we will discover how that is done in the near term. Until that is done, we will not have anything resembling "real AI"
So let's all cut the bullshit and simply try to make use of these useful tools in our own lives.
"The street finds its own uses for things."
Spot on! I am a developer and have started to use these AI tools to help in my daily work. But as you say, this craze is just nuts. It's not magic and we haven't become gods.
Garbage in, garbage out still applies, even to AI.
AI has developed enormously in the last 20 years. The issue is not where it is today, but where it will be in 5 years, 10 years, 20 years, as AIs are used to teach and design successive generations of AI. AlphaGo, which is now the best Go player in the world and is now being applied to other domains, is programmed to learn (as are all AIs now being developed) and has proved it can innovate with moves unknown to and unrealized by the best Go players in the world.
In the case of AIs, GIGO refers to what they learn from. Some may end up with the right material to produce a psychopath. The fact that it was GIGO will be no consolation.
It could well be that what we need is a psychopath AI - provided it's a rational psychopath who understands who the real enemy is. (Psychopathy does not imply irrationality, even if most psychopaths are irrational - that's BS from the psychology profession which doesn't understand anything.)
Thank you for your concise observations. I have been saying the same things for years as a software engineer. The fact that you describe of feeding data into these software modules is indicative of what we commonly refer to in the profession as "Machine Learning" or ML. ML uses neural networks to determine best answers based in the same manner as well designed B+Tree structure but is nonetheless far more complex and capable.
That being said, you are correct to state that the majority of what we are seeing is not sentient AI. A sentient AI would not need so much data given to it but merely a base set so that it could learn what it needed on its own and then make its own decisions. Such an AI to be truly sentient would also have to exhibit an "awareness of self", which to date, has not been demonstrated by anything other than robots who can only mimic self-awareness as a result of its capabilities with conversation.
The most advanced AI research is being done in Japan as for some reason the Japanese appear to love this field of research. However, with their immense talents and capabilities, the best that Japanese researchers have been able to create has been sex-bots for the lonely. And none, to my knowledge have reached any level of self-awareness on their own.
What is frighteningly dangerous is not the development of AI but the use by stupid Humans who believe they have created such software and then apply it to real-world situations as Google and Tesla have done with their self-driving vehicles only to see them crash into things, kill pedestrians, and even the occupants. Now apply this garbage to military controls... and yes, we will have Armageddon... But not because a sentient AI believes that Humans are inferior but simply because it came up with the same whacked-out conclusions that we are seeing now...
You don't get it either. The conversation between lemoine and laMDA is pure fiction composed by a hack writer paid as part of a rollout of this garbage. The real AI has nothing to do with getting machines to produce pseudo philosophical text which "stuns" John and Jane Q Boomer when they coincidentally all "discover" it simultaneously because corporate masters want them to. AI is doing all sorts of very real things including learning to police and rule the internet. As an epiphenomenon of the entire project there are prose-writing bots, whose purpose is to hook the planet of tech users and employees and everyday people into using them for tasks ranging from birthday greetings to drafts of grant proposals to weapon tech specification manual composition so that the Cloud knows everything everyone is thinking and planning even better than it does today. The Big Big PR campaign Now as a result of these wholly manufactured letters of Musk and other so called experts has one purpose - to get you and everyone you know everywhere to start using the available bots and addict you. In addition to charge monthly fees - no longer for renting cloud services for saving files and editing documents and spreadsheets (everyone's spreadsheet estimates have been freely available for over a decade now - way to go, corporate spy agency!) with specialized software, but now to seize control of every spasm of imaginative effort on earth and turn it to the ends of those in power who own and run these things.
That's certainly possible. But of course the smart thing to do is to 1) not use these tools in a way that compromises your OPSEC, 2) use these tools to the degree possible in a "self-hosted" manner, and 3) don't use them for pointless crap like writing birthday greetings.
It's fine to use these tools for personal enjoyment or things like creating art for personal enjoyment or as promo items for a Youtube video or something. Who cares what some cretin uses these tools for? Not everyone on the planet is worth surveilling, and people who think they're under surveillance when in reality no one in power gives a rat's ass about them (except in the aggregate) are people who have a little much paranoia, instead of just enough.
As someone with a clue about computer security, I avoid using tools that are hosted in the cloud, such as password managers , or which are not open-source and capable of being reviewed. There are exceptions to that, based on how much effort I want to spend on implementing my own tools vs how much I think the PTB actually care what I'm doing, and I never allow my deepest and darkest plans and thoughts to leave my computer. Which is one reason I don't use Windows or Macs.
But until you're doing something that makes you a significant threat to the PTB, 99.9% of you are simply irrelevant and your own security is not under significant threat.
As for AI policing the Internet, I don't doubt that will be its eventual role. But it won't be any more perfect at it than the manual policing the CIA and NSA have been doing so far. And as to that, I'd say it's almost hardly needed, given that 99.9% of the population is already being fed, and believing in, so much conspiracy theory as is represented by the notions such as 1) Ukraine is beating Russia, 2) Covid was a Chinese lab leak, and 3) the 2020 election was stolen, and 4) Trump is their savior.
It's too easy to fall into the binary "either/or" stuff that is being fed people today, which is the primary method by which the PTB have been controlling people for the last several thousand years. Ascribing everything as one big conspiracy is the number one way that trick has been used.
Again, go read "Illuminatus" and its derivatives. Wilson and Shea had it down.
Well, that saves me writing my planned response ;-)
(Btw: As a SW developer I have heard since the early 90's that my work would soon be taken over by AI... here we are 30 years later, and AI still cannot create SW, it can only regurgitate what other programmers have done. Neat, surely, but hardly a thing to start worrying over, especially as I know that at least -and I am being generous here- 80% of all code on the internet is flawed in some way or other)
As a SW developer you must also realize that you only need one deluded moron to give these things an ability to make PUT/POST requests before they will start doing something dangerous, like corrupting Wikipedia articles or registering on any of the million web services to wreak havoc. "It can only regurgitate what other programmers have done". Step 1: Feed a lot of security vulnerabilities into it, Step 2: make it crawl the web in the search of these and exploit, Step 3: .... So even if your job might be safe, your bank account, your IoT devices, etc, etc, might very well be not. It doesn't need to be very smart to do some very real damage, just like a hoodlum with a knife might be dumb but that knife in your gut might kill you.
Its not about my job, its more the idea that this crop of AI is somehow more or less dangerous than any other piece of software.
Personally, I think somebody wants a lot of money and is creating the hype necessary for it.
"I've followed the various AI crazes over the decades since the 1970s."
There was no Internet in the 70s, there was no computational capability and, most of all, the world wasn't so connected. If you give an ability for AI to make PUT/POST requests, as in change stuff on the Internet, I can see how it might start doing pretty significant damage, which also might not get discovered in time.
It's K-K-K-Ken c-c-c-coming to k-k-k-kill me.