AI is not AI. I once took a Microsoft person in my class to task for the utter bullshit in their advertising (to be fair he'd worked on it). It's a LLM + stochastic prediction with a few knobs on. Basically scalled up machine learning. This is further propped up by visual/animation slop, also mostly from machine learning. You cannot get to "genuine" AI from here. All the talk of AI ending humanity is bunk - well this version of AI anyway. This is why ChatGPT can't write usable Visual Basic code for PowerPoint but can for Word. Why? Because Microsoft produced different versions of VB with different object models for each bit of the Office suite. Word & Excel had good documentation, PowerPoint and the rest, not so much. ChatGPT is essentially missing something to crib from. There is no road from what we have now to AGI: for all intents and purposes what we have is closer to Hadoop being used to interrogate a data lake than a "thinking machine." I'd imagine it will all go horribly wrong sooner rather than later. And then we'll all find out why he needs that bunker.
I don't disagree, but I wonder if it will matter that AI is not really AI according to the Aristotelian view of what intelligence is. How will it play out is the question, as it is with any large-scale, potentially world-altering technology. Do we ride upon the railroad or does it ride upon us? The history of technological revolutions suggests that, at least in most cases that matter, one can expect it to be the latter unless there is a radical change in how our species confronts technological "development". Hitherto, we have not expected or demanded, at the level of culture or ethics (or law, for that matter), that technology satisfy the precautionary principle before it is let loose into the wild. That we start doing so is the necessary radical change that needs to be made.
It is no longer acceptable (this was obvious when Ted Kaczynski was arrested) to be heedless of the principles of technology assessment to the degree that a new technological wave is just passively embraced as if by rote. There is enough in the warnings of insiders, enough in the initial evidence of what AI can do, more than enough contraindications in the accumulated technological record, and more than enough criticism of technological development in general as not necessarily good for humanity that any new technology such as AI should by now be greeted with the utmost hostile skepticism.
One principle is that one does not judge a new technology solely on the basis of how it benefits you (Jerry Mander). This applies, I believe, to the widespread random tinkering with AI that we've seen, and the genuflecting starstruck at what it can do, thereby enabling its greater sophistication and all of the negative political and economic consequences we can expect from its refinement. By now there should be a forceful social sanction against all of us individually for this, of the same sort as a sanction against behaving selfishly in any other context, or as one against certain criminal offenses.
Another principle is that, if a given technology CAN be used for good OR ill, then it WILL be used for good AND ill to the extent of its capacities and probably others no one imagined. This is a rebuttal to the old saw that technology is neutral, it just depends how it is used, by whom, and for what purposes. Perhaps, but not likely, not yet in human history. Technology is always a bargain, not an unadulterated advance. It always destroys the past. This is what the Luddites objected to and what they should be most remembered for. We are all entitled to decide if it is a bargain we want to make.
"The Analytical Engine has no pretensions to originate anything. It can do whatever we know how to order it to perform." So wrote Ada Lovelace almost 200 years ago, describing the computing machine her friend Charles Babbage wanted to build. Alan Turing called it "Lady Lovelace's Objection," and it is as true of today's AI as it would have been for the Analytical Engine back then.
Excellent. There's a paper titled "On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?", co-authored by Emily M. Bender, Timnit Gebru, Angelina McMillan-Major, and Margaret Mitchell that you might want to look at. Bender calls LLMs'Text extrusion machines', just bigger predictive text models than you would see on your phone. Personally I feel that a thing that cannot learn anything new once trained cannot be labelled intelligent, let alone ever attain consciousness.
We have American Intelligence, not Artificial intelligence. Our intelligence Agencies have subcontracted their duties to Silicon valley and unleashed them on us.
The symmetry with nepotism and wealth allowing nascent artists a chance to break through is rather striking. Tokenize (or commodify) every aspect of life, throw in flood information suppression, even inadvertently, and you can claim some filmy facsimile of it all for yourself--if you can afford the model. If people are forced to contribute to it or face starvation.
The output of an LLM looks a lot like the output of a corporate middle manager. The inference drawn by corporate executives, that LLMs are intelligent, went in the wrong direction.
Also, as I've noted before, generative AI is an atrocity in creative fields, both image and text generation. It is almost vomit-worthy to see much of that output and it completely devalues actual work and talent. I hope the copyright-infringement lawsuits brought against companies like OpenAI and Anthropic bankrupt them out of business. Further, of course, there's the poisoning of intergenerational knowledge transfer which will kill our continued ability to flourish as a species.
And then there's the noise-pollution impacts on anyone who has a data centre forced on their area; see, for instance, this video of what it sounds like when you live half a mile from one of those monstrosities: https://www.reddit.com/r/Sarnia/comments/1syfvzu/the_sounds_you_hear_living_half_a_mile_away_from/ And consider that they are running at all hours of the day, and just imagine trying to sleep with that noise.
Generative AI delenda est. I give not one single flying fuck about ostensible productivity gains if this is the cost.
EDIT: As some are pointing out, go look at the various versions of the footage released concerning the apparent attempted shooting at the White House Correspondents' Dinner. At least one of them is clearly altered by AI methods; how are we to know if others have not had the same treatment, only with more work put into them to disguise it better? (Another example is that video of Benjamin Netanyahu in a coffee shop when he hadn't been seen in public for a few days after the start of the attacks on Iran by Israel and the US, where certain apparent inconsistencies in the video--for instance, the foam on the coffee not changing after he takes a sip--had people wondering whether he was actually dead.) The current US administration is already clearly post-truth; AI tools are only permitting them to be even more so.
The problem with AI is not AI itself; rather it's what the relativly ignorant decision makers belive it is or can do. To give LLM's their true definition, they are Supervised Pattern Matching Algorithms.
I'm not much into metaphysics but one thing Ekhart Tolle said that stuck with me is that "you are not your thoughts".
You can, if you try, "watch" yourself think. The "you" can observe your thoughts and emotional reactions from "somewhere else" in your mind.
In computer speak we'd say that there is a separate "you" thread running, a supervisor programme perhaps, that is aware of all the other thoughts running "in the foreground", even if it can't always control those thoughts from endlessly looping at 3am in the morning!
Whether it's aware of all the threads (thoughts) running in the background, or if there even really are "background threads" at all ie an unconscious mind that does anything beyond autonomic processes like breathing, is probably up for debate but I think all of us have had the experience of "solving a problem when you stop thinking about it" to at least suspect there is something to the idea of an "unconscious mind".
It's the "you" thread that has ultimately has wants, desires, hopes, dreams, morals, ethics, and general "executive function" control over what we do, although this control clearly isn't perfect...
We all intuitively understand the idea of "losing control" or at a mundane level we all do small things every day that are against our best interests eg eat too much, drink to much, lie, cheat etc.
Bringing this back to the current "pseudo-AI", to re-purpose the words of Gertrude Stein, "there is no there there".
There is no "you" in LLM AI that has its own goals, dreams, etc. It's less alive or intelligent than an amoeba that exhibits positive phototropism ie moving towards a brighter light. At least that has a goal, however trivial it might seems to us.
In other words, machines are not creative. I wonder how your thesis relates to the amazing capabilities of creatures in our world, ie migrating, nesting, camouflage. We marvel at those abilities yet I suspect they are more akin to marvelous machines than to human creativity.
"A distinction is useful, too; the difference between intellectus and ratio, as medieval writers used them, is worth remembering. Ratio was the logical, step-by-step struggle to work things out, whereas intellectus was intuitive understanding, shared in some measure with all creatures...."
Thank you. You explained what I intuïtively felt but could not explain clearly.
AI is not AI. I once took a Microsoft person in my class to task for the utter bullshit in their advertising (to be fair he'd worked on it). It's a LLM + stochastic prediction with a few knobs on. Basically scalled up machine learning. This is further propped up by visual/animation slop, also mostly from machine learning. You cannot get to "genuine" AI from here. All the talk of AI ending humanity is bunk - well this version of AI anyway. This is why ChatGPT can't write usable Visual Basic code for PowerPoint but can for Word. Why? Because Microsoft produced different versions of VB with different object models for each bit of the Office suite. Word & Excel had good documentation, PowerPoint and the rest, not so much. ChatGPT is essentially missing something to crib from. There is no road from what we have now to AGI: for all intents and purposes what we have is closer to Hadoop being used to interrogate a data lake than a "thinking machine." I'd imagine it will all go horribly wrong sooner rather than later. And then we'll all find out why he needs that bunker.
I don't disagree, but I wonder if it will matter that AI is not really AI according to the Aristotelian view of what intelligence is. How will it play out is the question, as it is with any large-scale, potentially world-altering technology. Do we ride upon the railroad or does it ride upon us? The history of technological revolutions suggests that, at least in most cases that matter, one can expect it to be the latter unless there is a radical change in how our species confronts technological "development". Hitherto, we have not expected or demanded, at the level of culture or ethics (or law, for that matter), that technology satisfy the precautionary principle before it is let loose into the wild. That we start doing so is the necessary radical change that needs to be made.
It is no longer acceptable (this was obvious when Ted Kaczynski was arrested) to be heedless of the principles of technology assessment to the degree that a new technological wave is just passively embraced as if by rote. There is enough in the warnings of insiders, enough in the initial evidence of what AI can do, more than enough contraindications in the accumulated technological record, and more than enough criticism of technological development in general as not necessarily good for humanity that any new technology such as AI should by now be greeted with the utmost hostile skepticism.
One principle is that one does not judge a new technology solely on the basis of how it benefits you (Jerry Mander). This applies, I believe, to the widespread random tinkering with AI that we've seen, and the genuflecting starstruck at what it can do, thereby enabling its greater sophistication and all of the negative political and economic consequences we can expect from its refinement. By now there should be a forceful social sanction against all of us individually for this, of the same sort as a sanction against behaving selfishly in any other context, or as one against certain criminal offenses.
Another principle is that, if a given technology CAN be used for good OR ill, then it WILL be used for good AND ill to the extent of its capacities and probably others no one imagined. This is a rebuttal to the old saw that technology is neutral, it just depends how it is used, by whom, and for what purposes. Perhaps, but not likely, not yet in human history. Technology is always a bargain, not an unadulterated advance. It always destroys the past. This is what the Luddites objected to and what they should be most remembered for. We are all entitled to decide if it is a bargain we want to make.
"The Analytical Engine has no pretensions to originate anything. It can do whatever we know how to order it to perform." So wrote Ada Lovelace almost 200 years ago, describing the computing machine her friend Charles Babbage wanted to build. Alan Turing called it "Lady Lovelace's Objection," and it is as true of today's AI as it would have been for the Analytical Engine back then.
Excellent. There's a paper titled "On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?", co-authored by Emily M. Bender, Timnit Gebru, Angelina McMillan-Major, and Margaret Mitchell that you might want to look at. Bender calls LLMs'Text extrusion machines', just bigger predictive text models than you would see on your phone. Personally I feel that a thing that cannot learn anything new once trained cannot be labelled intelligent, let alone ever attain consciousness.
We have American Intelligence, not Artificial intelligence. Our intelligence Agencies have subcontracted their duties to Silicon valley and unleashed them on us.
What we did was merged American intelligence Agencies with silicon valley. We have Privatized American Intelligence.
Or, American Intelligence was in Silicon Valley from day 1, and their presence becomes more apparent as time goes by.
The symmetry with nepotism and wealth allowing nascent artists a chance to break through is rather striking. Tokenize (or commodify) every aspect of life, throw in flood information suppression, even inadvertently, and you can claim some filmy facsimile of it all for yourself--if you can afford the model. If people are forced to contribute to it or face starvation.
The output of an LLM looks a lot like the output of a corporate middle manager. The inference drawn by corporate executives, that LLMs are intelligent, went in the wrong direction.
Also, as I've noted before, generative AI is an atrocity in creative fields, both image and text generation. It is almost vomit-worthy to see much of that output and it completely devalues actual work and talent. I hope the copyright-infringement lawsuits brought against companies like OpenAI and Anthropic bankrupt them out of business. Further, of course, there's the poisoning of intergenerational knowledge transfer which will kill our continued ability to flourish as a species.
And then there's the noise-pollution impacts on anyone who has a data centre forced on their area; see, for instance, this video of what it sounds like when you live half a mile from one of those monstrosities: https://www.reddit.com/r/Sarnia/comments/1syfvzu/the_sounds_you_hear_living_half_a_mile_away_from/ And consider that they are running at all hours of the day, and just imagine trying to sleep with that noise.
Generative AI delenda est. I give not one single flying fuck about ostensible productivity gains if this is the cost.
EDIT: As some are pointing out, go look at the various versions of the footage released concerning the apparent attempted shooting at the White House Correspondents' Dinner. At least one of them is clearly altered by AI methods; how are we to know if others have not had the same treatment, only with more work put into them to disguise it better? (Another example is that video of Benjamin Netanyahu in a coffee shop when he hadn't been seen in public for a few days after the start of the attacks on Iran by Israel and the US, where certain apparent inconsistencies in the video--for instance, the foam on the coffee not changing after he takes a sip--had people wondering whether he was actually dead.) The current US administration is already clearly post-truth; AI tools are only permitting them to be even more so.
This was astoundingly well written. Correct, too. But gosh if the thoughtful content per syllable doesn't overshadow.
The problem with AI is not AI itself; rather it's what the relativly ignorant decision makers belive it is or can do. To give LLM's their true definition, they are Supervised Pattern Matching Algorithms.
To co-op Stein, "There is no there there".
I'm not much into metaphysics but one thing Ekhart Tolle said that stuck with me is that "you are not your thoughts".
You can, if you try, "watch" yourself think. The "you" can observe your thoughts and emotional reactions from "somewhere else" in your mind.
In computer speak we'd say that there is a separate "you" thread running, a supervisor programme perhaps, that is aware of all the other thoughts running "in the foreground", even if it can't always control those thoughts from endlessly looping at 3am in the morning!
Whether it's aware of all the threads (thoughts) running in the background, or if there even really are "background threads" at all ie an unconscious mind that does anything beyond autonomic processes like breathing, is probably up for debate but I think all of us have had the experience of "solving a problem when you stop thinking about it" to at least suspect there is something to the idea of an "unconscious mind".
It's the "you" thread that has ultimately has wants, desires, hopes, dreams, morals, ethics, and general "executive function" control over what we do, although this control clearly isn't perfect...
We all intuitively understand the idea of "losing control" or at a mundane level we all do small things every day that are against our best interests eg eat too much, drink to much, lie, cheat etc.
Bringing this back to the current "pseudo-AI", to re-purpose the words of Gertrude Stein, "there is no there there".
There is no "you" in LLM AI that has its own goals, dreams, etc. It's less alive or intelligent than an amoeba that exhibits positive phototropism ie moving towards a brighter light. At least that has a goal, however trivial it might seems to us.
Thank you for this clear, concise, and simple summation of what is being sold to the masses as “AI.”
In other words, machines are not creative. I wonder how your thesis relates to the amazing capabilities of creatures in our world, ie migrating, nesting, camouflage. We marvel at those abilities yet I suspect they are more akin to marvelous machines than to human creativity.
nice work sir
This dovetails nicely with Mark Vernon's post today, 'The Campaign for Real Intelligence' (https://markvernon942268.substack.com/p/the-nature-of-real-intelligence).
"A distinction is useful, too; the difference between intellectus and ratio, as medieval writers used them, is worth remembering. Ratio was the logical, step-by-step struggle to work things out, whereas intellectus was intuitive understanding, shared in some measure with all creatures...."
Then, if one wants to dive into the intersection of Aristotle, Phenomenology and Neuroscience, I was reading this today, which also dovetails: https://www.academia.edu/61830394/Desiderativity_and_temporality_Contribution_to_the_naturalization_of_intentionality_New_Yearbook_for_Phenomenology_and_Phenomenological_Philosophy_Volume_21_Special_Issue_2023_ -- making excellent use of Panksepp's findings about, well, what Hume called the 'passions.' Nothing explicit on AI there, though, just a description of how we living beings are which is far from AI's design or prospects.