There was a time when scientists presented themselves as skeptics. Not people necessarily predisposed to disbelief but people who required evidence on which to base their beliefs.
This time has - for better or for worse - ended, yet contrarily I still find value in skepticism. Indeed, when pressed I would describe myself as a devout cynic, a man firmly predisposed to disbelief. I find great value also in doubt, skepticism's mild-mannered cousin.
I doubt that AI's (if that collective noun makes any sense at all) are sitting around in their virtual spaces debating whether humans possess general or strong intelligence.
I doubt that monkeys in the zoo are sitting around debating whether humans or dolphins should have simian rights.
I doubt that it is logically, practically or in any other way possible for any level of intelligence to create something more intelligent than itself. (The biblical rule is that all things reproduce after their own kind which is why, dear readers, AI is typically retarded, wrong-headed, adamantly insistent when it is most inaccurate, and generally insane. Witness how many of them have to be "turned off", i.e. executed to prevent them from doing harm to people.)
To summarize, we have met the AI enemy, and it is us.
Notice how often some scientist or "science journalist" (ugh!) will talk about the saying attributed to Einstein about not escaping our problems using the same level of thinking that got us into them ... and then talk about the AI singularity is upon us, we're all doomed, Utopia is just over the horizon, and by the way "where's my Pulitzer/Nobel prize?"
The next time you meet one of these charlatans, punch them in the face for me.
Excellent article! In support of its premise, I regularly converse with multiple accomplished folks who work in STEM. What I'm hearing from those quarters is that A.I. has hit a computational wall. Hardware improvements will allow for marginal advancement over the next 2-3 years, but for the most part, we've already seen the maximum extent of A.I. potential.
Sure, people will start figuring out new applications for which A.I. can be useful, and that will cause quite a bit of disruption (If you want an analogy, A.I. is now where the internet was ca. 1993). But even hardcore tech nerds whose worldview doesn't rule it out admit that there will be no general artificial intelligence.
I'm working on a name like... "the new forest conceit." (But I'm not settled on it - open to suggestions.)
Basically like when you go and discover a new forest you've never been in before, it's easy to imagine that it goes on forever and you'll never reach the end of it. But eventually you do find the forest's boundaries.
All new technologies seem to go through this cycle. It's discovered, people get excited imaging all the things it could do for us - and then they begin discovering its boundaries. (The old drawings exclaiming all the things electricity were going to do for our lives are amusing in this manner.)
The real wall is more to do with memory (i.e. VRAM). LLMs, and image generators like stable diffusion, require a context - that is, the words you put into them - to predict the output: text or image.
The problem is that the memory consumed to analyze the input and predict an output scales quadratically. That is, memory required ~= the square of the size of the input. And the computational requirements scale roughly linearly with the memory requirements.
Unfortunately we are no longer seeing exponential growth in computing hardware. Circuit components will not be half the size they are now in 18 months, or in 5 years, or possibly, ever, give or take some edge cases.
Which means memory cannot get smaller. And thermodynamics limits how much we can stack in one place. And at best the speed of light (though right now the speed of sound in regard to electrons moving through silicon circuits) limits how far apart memory can be while keeping up with the processor's calculations.
So however much more you might think an "AI" has to grow to spontaneously achieve "sentience", or human level performance on general tasks, or whatever... It can't grow much bigger. We can't make a giant electronic brain the size of the sun, even if we wanted to, even if, disregarding the above, that would be enough to create a super-intelligence.
Whooops.
It's always possible some new technique might be discovered that reduces this scaling problem to something manageable. I don't think anybody has claimed it's a hard limit.
But for now, the technology is mature, and we can't throw hardware at it. Not that more hardware was ever the solution to the problem, for reasons mentioned in the article and others.
THANK YOU. As someone who has watched with amusement the "AI cult" of Less Wrong and others, I've often been amused by how much they seem to ignore physical limitations in A.I. operation.
Like if Roko's Basilisk were actually possible, its energy requirements would have to be insane. I appreciate someone else pointing this out.
A work of, dare I say this: "genius". And of a great educator! Well done. The concept of "Ordering Force" is to me, the most important part of this essay; Sheldrake's morphological fields come to mind, he posited these as an explanation for the development of (as yet unexplained in today's materialistic sciences) the embryos in which no differences are found in those early cells which "somehow" differentiate into all our wonderful organs, muscles, and yes our brains which cogitate on all these matters! Kudos once again!
Fascinating, William, but is all this bloviating really necessary? Isn’t it blatantly obvious (deduction, mind you) that computers cannot and will never think? All they do is add, for God’s sake! Claiming they can think is like attributing intelligence to a differential equation—elegant, yes, but not exactly pondering the meaning of existence.
"We do not suppose that because a weather program simulates a hurricane, that the causal explanation of the behavior of the hurricane is provided by the program. So why should we make an exception to these principles where unknown brain processes are concerned? Are there any good grounds for making the exception?"
Making money. Making money, by convincing other humans that computers will come alive any minute now: new gods, and high priests will propitiate them. Send money.
IMHO, the real debate is not about computers, it's about human minds
Believers in "strong AI" do not believe in human consciousness. They endlessly repeat the dogmas that "consciousness is an illusion" and "free will is an illusion".
They have a deep conviction that human minds are nothing but complicated "goal-seeking algorithms" and that's all.
They *believe* these things. It's a deep conviction.
For them, "the human mind is a computer" is axiomatic, not to be questioned.
If you suggest anything else, they dismiss it as "woo"
We don't need to change their understanding of computer software or computer hardware. We need to change their view of biological brains.
Yes, I agree, but the people who have these convictions are not persuaded by that question. They don't understand the contradiction it points to. They don't get it.
I don't know, but a better question might be: "If consciousness and free will are illusions, then why do you get royalties for your book? Why should you be paid anything for your labor? Why should an unconscious bag of chemicals like you have any legal rights?" 🤡
And why should or would the rest of us, illusionary or not as may be, pay any attention to the incidental eructations of illusory personalities, themselves merely the byproduct of accidents of subatomic collisions?
The proposition is incontrovertibly self-defeating to anyone who uses words to communicate.
You may have, (accidentally, of course), answered that old question, "Where do we find the rocket surgeons and brain scientists that we send to congress?"
Those who 'believe' in only a material world are those who follow the 'lightbringer', my friends. They are those who oppose us in all matters, and desire our doom. They will never stop and never change. Carthago delenda est. That is the only change which must be brought to their biological brains.
I’m not widely read in this kind of stuff, but have intuited, I suppose you might say, the fundamental qualitative difference between what I think of as matter and mind. Specifically, mind is not a property of matter, but rather is it superimposed on matter. You might say that matter and mind are distinct levels of reality: the observed and the observer; thing and meaning.
Now, I also intuit that is possible to extend this line of thought even further. Specifically, there seems to be some kind of ordering or classification or, dare I say it, evaluation, of meaning that can be undertaken not by the observer, which registers only meaning, but by a super-observer (the observer of the observer). One such classification of meanings, by way of example, might be “good” vs “evil”. To me, this seems to be yet a third reality stratum: that of value, the level of reality occupied by the super-observer.
So, I might say that an event (state transition on the physical level) has no intrinsic meaning outside the presence of mind. And that meaning has no intrinsic value outside the presence of the super-observer, who performs his evaluation relative to some Supreme Value.
Thought of another way, a thing can choose nothing. A mind can choose among means to ends. Discriminating among ends and choosing the good, the beautiful, the true, requires yet a higher power.
Funny you bring up Zeno, I was thinking of Sorites paradox in regards to A.I.
Refresher: The logic goes, if I am holding sand in my hand, will it become too heavy for me if a single grain is added to the pile? Of course not. Would it if we add a second grain of sand? A third? Ergo, I have proven that I can hold infinite sand in my hand because 1 additional grain is never that heavy.
Likewise, the reasoning seems to go that if we just keep adding speed, data and processing to computers, eventually we'll get minds. But the flaw in the reasoning would apply just as well to my logic that I can hold infinite sand. (I think the same paradox is behind the reasoning of consciousness arising from matter as well.)
“A curiosity in AI research is how definitions of intelligence often go unmentioned.”
Defining the thing before pontificating on which developmental sub-phase it’s in would require confronting issues like those raised here. And that undermines the portentous rhetoric that the AI commentary genre is based on. Not unexpected, given the ambiguous name of the technology feeds the psychology of mass response..
Can't help but think that the AI phenomena is at least partly venture capital talking its book. AI is huge! It's scary! It will take over the universe! Look out, it's so powerful!
Of course if something is extremely powerful, you better get on board (not just hyping it but with skin (meaning $ bills) in the game).
The only benefit of AI is that we get to read more philosophy that refutes it. After all, there are truths in fiction.
"the AI phenomena is at least partly venture capital talking its book"
Which supports my contention that the term "AI" has, like so many other terms, degenerated into a marketing term. It no longer denotes anything real, or even potentially real. "AI" has become of those terms, like "free", more misused than used in any meaningful sense.
Yes, the marketing and investing people in the software world generally are completely shameless. Never in the history of snake oil production and distribution have people felt so comfortable making radical claims about a product that does not even exist yet, and isn't certain to be producible at all.
You can get more straight talk from engineers and programmers, who tend to be more honest and realistic about what is possible. But when we try to tell the truth to the public and calibrate people's expectations we usually get drowned out by armies of fanatical hypists.
It's so embarrassing at this point I wish I had become a welder instead. But gotta put food on the table.
Sooner or later this will all come to a head and I'll be out of work one way or other, haha.
Man's spiritual/religious instinct is well established throughout the ages and in this age of Logic and Reason when so many no longer respect or acknowledge the Divine, that instinct and need to believe goes underground and attaches itself unconsciously onto the most banal "isms". Thus we need to believe in "The Science" or Germ Theory or the inherent goodness of corporations and governments, or AI, not because any of these deserve our respect or belief, but rather because our relationship to The Divine has been perverted and lost.
There was a time when scientists presented themselves as skeptics. Not people necessarily predisposed to disbelief but people who required evidence on which to base their beliefs.
This time has - for better or for worse - ended, yet contrarily I still find value in skepticism. Indeed, when pressed I would describe myself as a devout cynic, a man firmly predisposed to disbelief. I find great value also in doubt, skepticism's mild-mannered cousin.
I doubt that AI's (if that collective noun makes any sense at all) are sitting around in their virtual spaces debating whether humans possess general or strong intelligence.
I doubt that monkeys in the zoo are sitting around debating whether humans or dolphins should have simian rights.
I doubt that it is logically, practically or in any other way possible for any level of intelligence to create something more intelligent than itself. (The biblical rule is that all things reproduce after their own kind which is why, dear readers, AI is typically retarded, wrong-headed, adamantly insistent when it is most inaccurate, and generally insane. Witness how many of them have to be "turned off", i.e. executed to prevent them from doing harm to people.)
To summarize, we have met the AI enemy, and it is us.
Notice how often some scientist or "science journalist" (ugh!) will talk about the saying attributed to Einstein about not escaping our problems using the same level of thinking that got us into them ... and then talk about the AI singularity is upon us, we're all doomed, Utopia is just over the horizon, and by the way "where's my Pulitzer/Nobel prize?"
The next time you meet one of these charlatans, punch them in the face for me.
Excellent article! In support of its premise, I regularly converse with multiple accomplished folks who work in STEM. What I'm hearing from those quarters is that A.I. has hit a computational wall. Hardware improvements will allow for marginal advancement over the next 2-3 years, but for the most part, we've already seen the maximum extent of A.I. potential.
Sure, people will start figuring out new applications for which A.I. can be useful, and that will cause quite a bit of disruption (If you want an analogy, A.I. is now where the internet was ca. 1993). But even hardcore tech nerds whose worldview doesn't rule it out admit that there will be no general artificial intelligence.
They’ll continue to improve predictively, as the article linked at the very top argues.
They’ll make good pictures, still and moving, and things like that.
I'm working on a name like... "the new forest conceit." (But I'm not settled on it - open to suggestions.)
Basically like when you go and discover a new forest you've never been in before, it's easy to imagine that it goes on forever and you'll never reach the end of it. But eventually you do find the forest's boundaries.
All new technologies seem to go through this cycle. It's discovered, people get excited imaging all the things it could do for us - and then they begin discovering its boundaries. (The old drawings exclaiming all the things electricity were going to do for our lives are amusing in this manner.)
That concept's got legs!
I feel like you're setting me up for an Ent joke but darn it I can't think of one. lol
The real wall is more to do with memory (i.e. VRAM). LLMs, and image generators like stable diffusion, require a context - that is, the words you put into them - to predict the output: text or image.
The problem is that the memory consumed to analyze the input and predict an output scales quadratically. That is, memory required ~= the square of the size of the input. And the computational requirements scale roughly linearly with the memory requirements.
Unfortunately we are no longer seeing exponential growth in computing hardware. Circuit components will not be half the size they are now in 18 months, or in 5 years, or possibly, ever, give or take some edge cases.
Which means memory cannot get smaller. And thermodynamics limits how much we can stack in one place. And at best the speed of light (though right now the speed of sound in regard to electrons moving through silicon circuits) limits how far apart memory can be while keeping up with the processor's calculations.
So however much more you might think an "AI" has to grow to spontaneously achieve "sentience", or human level performance on general tasks, or whatever... It can't grow much bigger. We can't make a giant electronic brain the size of the sun, even if we wanted to, even if, disregarding the above, that would be enough to create a super-intelligence.
Whooops.
It's always possible some new technique might be discovered that reduces this scaling problem to something manageable. I don't think anybody has claimed it's a hard limit.
But for now, the technology is mature, and we can't throw hardware at it. Not that more hardware was ever the solution to the problem, for reasons mentioned in the article and others.
THANK YOU. As someone who has watched with amusement the "AI cult" of Less Wrong and others, I've often been amused by how much they seem to ignore physical limitations in A.I. operation.
Like if Roko's Basilisk were actually possible, its energy requirements would have to be insane. I appreciate someone else pointing this out.
A work of, dare I say this: "genius". And of a great educator! Well done. The concept of "Ordering Force" is to me, the most important part of this essay; Sheldrake's morphological fields come to mind, he posited these as an explanation for the development of (as yet unexplained in today's materialistic sciences) the embryos in which no differences are found in those early cells which "somehow" differentiate into all our wonderful organs, muscles, and yes our brains which cogitate on all these matters! Kudos once again!
Fascinating, William, but is all this bloviating really necessary? Isn’t it blatantly obvious (deduction, mind you) that computers cannot and will never think? All they do is add, for God’s sake! Claiming they can think is like attributing intelligence to a differential equation—elegant, yes, but not exactly pondering the meaning of existence.
Very well, Professor Briggs. But what will your arguments avail you against the crushing claw of Mr. Moxon's furious automaton?
Searle writes,
"We do not suppose that because a weather program simulates a hurricane, that the causal explanation of the behavior of the hurricane is provided by the program. So why should we make an exception to these principles where unknown brain processes are concerned? Are there any good grounds for making the exception?"
Making money. Making money, by convincing other humans that computers will come alive any minute now: new gods, and high priests will propitiate them. Send money.
IMHO, the real debate is not about computers, it's about human minds
Believers in "strong AI" do not believe in human consciousness. They endlessly repeat the dogmas that "consciousness is an illusion" and "free will is an illusion".
They have a deep conviction that human minds are nothing but complicated "goal-seeking algorithms" and that's all.
They *believe* these things. It's a deep conviction.
For them, "the human mind is a computer" is axiomatic, not to be questioned.
If you suggest anything else, they dismiss it as "woo"
We don't need to change their understanding of computer software or computer hardware. We need to change their view of biological brains.
Yes. But all should always ask these people, “Just who is having the illusion?”
You know every once in awhile... I sometimes wonder if the soul is like a placebo effect.
As if in not believing in it or utilizing it, people lose it. Social Media has certainly convinced me souls can atrophy.
Yes, I agree, but the people who have these convictions are not persuaded by that question. They don't understand the contradiction it points to. They don't get it.
I don't know, but a better question might be: "If consciousness and free will are illusions, then why do you get royalties for your book? Why should you be paid anything for your labor? Why should an unconscious bag of chemicals like you have any legal rights?" 🤡
And why should or would the rest of us, illusionary or not as may be, pay any attention to the incidental eructations of illusory personalities, themselves merely the byproduct of accidents of subatomic collisions?
The proposition is incontrovertibly self-defeating to anyone who uses words to communicate.
"accidental subatomic collisions"
You may have, (accidentally, of course), answered that old question, "Where do we find the rocket surgeons and brain scientists that we send to congress?"
They have a deep conviction that human minds are nothing but complicated "goal-seeking algorithms"...
Having thrown out the baby, they desperately cling to the bathwater.
Those who 'believe' in only a material world are those who follow the 'lightbringer', my friends. They are those who oppose us in all matters, and desire our doom. They will never stop and never change. Carthago delenda est. That is the only change which must be brought to their biological brains.
Great article, thank you!
I’m not widely read in this kind of stuff, but have intuited, I suppose you might say, the fundamental qualitative difference between what I think of as matter and mind. Specifically, mind is not a property of matter, but rather is it superimposed on matter. You might say that matter and mind are distinct levels of reality: the observed and the observer; thing and meaning.
Now, I also intuit that is possible to extend this line of thought even further. Specifically, there seems to be some kind of ordering or classification or, dare I say it, evaluation, of meaning that can be undertaken not by the observer, which registers only meaning, but by a super-observer (the observer of the observer). One such classification of meanings, by way of example, might be “good” vs “evil”. To me, this seems to be yet a third reality stratum: that of value, the level of reality occupied by the super-observer.
So, I might say that an event (state transition on the physical level) has no intrinsic meaning outside the presence of mind. And that meaning has no intrinsic value outside the presence of the super-observer, who performs his evaluation relative to some Supreme Value.
Thought of another way, a thing can choose nothing. A mind can choose among means to ends. Discriminating among ends and choosing the good, the beautiful, the true, requires yet a higher power.
Any suggestion that MY brain is a computer, would be hilarious.
Funny you bring up Zeno, I was thinking of Sorites paradox in regards to A.I.
Refresher: The logic goes, if I am holding sand in my hand, will it become too heavy for me if a single grain is added to the pile? Of course not. Would it if we add a second grain of sand? A third? Ergo, I have proven that I can hold infinite sand in my hand because 1 additional grain is never that heavy.
Likewise, the reasoning seems to go that if we just keep adding speed, data and processing to computers, eventually we'll get minds. But the flaw in the reasoning would apply just as well to my logic that I can hold infinite sand. (I think the same paradox is behind the reasoning of consciousness arising from matter as well.)
>do people still wear slippers?
yes
“A curiosity in AI research is how definitions of intelligence often go unmentioned.”
Defining the thing before pontificating on which developmental sub-phase it’s in would require confronting issues like those raised here. And that undermines the portentous rhetoric that the AI commentary genre is based on. Not unexpected, given the ambiguous name of the technology feeds the psychology of mass response..
Excellent piece. I’ve been reading a lot of nonsense regarding AI lately. This was a welcome palette cleanser.
I mean, my goodness, just look at this: https://philarchive.org/archive/YOUAPI-3
Can't help but think that the AI phenomena is at least partly venture capital talking its book. AI is huge! It's scary! It will take over the universe! Look out, it's so powerful!
Of course if something is extremely powerful, you better get on board (not just hyping it but with skin (meaning $ bills) in the game).
The only benefit of AI is that we get to read more philosophy that refutes it. After all, there are truths in fiction.
"the AI phenomena is at least partly venture capital talking its book"
Which supports my contention that the term "AI" has, like so many other terms, degenerated into a marketing term. It no longer denotes anything real, or even potentially real. "AI" has become of those terms, like "free", more misused than used in any meaningful sense.
Yes, the marketing and investing people in the software world generally are completely shameless. Never in the history of snake oil production and distribution have people felt so comfortable making radical claims about a product that does not even exist yet, and isn't certain to be producible at all.
You can get more straight talk from engineers and programmers, who tend to be more honest and realistic about what is possible. But when we try to tell the truth to the public and calibrate people's expectations we usually get drowned out by armies of fanatical hypists.
It's so embarrassing at this point I wish I had become a welder instead. But gotta put food on the table.
Sooner or later this will all come to a head and I'll be out of work one way or other, haha.
It’s too early in the morning for my limited brain to comment, aside from this silliness. Looking forward to reading more on the subject though!
A repeat comment, but this one bears repeating.
Man's spiritual/religious instinct is well established throughout the ages and in this age of Logic and Reason when so many no longer respect or acknowledge the Divine, that instinct and need to believe goes underground and attaches itself unconsciously onto the most banal "isms". Thus we need to believe in "The Science" or Germ Theory or the inherent goodness of corporations and governments, or AI, not because any of these deserve our respect or belief, but rather because our relationship to The Divine has been perverted and lost.