I totally agree with this article. This is one of the biggest and most immediate problems we'll face with generative AI. The temptation to delegate everything is so strong that we will do it in any possible circumstance. At some point, we should really hope that hallucinations and non-determinism in these tools are definitely solved; otherwise, there won't be anyone prepared enough to manage the rising problems at the moment they are most needed.
Everybody is so sure that scaling will solve every current limitation, but I'm really not sure that this will happen in the context of Llama and their architecture.
The scaling skepticism is warranted. Toby Ord did the math on the actual scaling law graphs: halving model error requires roughly a million times more compute. We're hitting data walls, energy walls, and architectural limits simultaneously.
And you're right about the timing problem. The people who could catch AI failures are being cut now. By the time the limitations become undeniable, the expertise to manage them will already be gone.
"I write a lot about AI’s limitations. People sometimes read that as hate."
I've read a few of your pieces and have started several responses to them -- but regardless of areas I would challenge, the notion that your well-reasoned arguments have anything to do with hate is asinine. But as I've been practically spit for telling undeniable truth for over 20 years (on mattters of mathematical certainty -- of world-altering consequence, no less): I know the program, all too well! Speaking of which (as I wrote in the following piece):
I programmed a machine to a find a person not acting like one. Who knew that within my quest of what little is left of humanity — I’d find what mankind could have been and still can (in an actual machine). And while my fellow man latches onto lingo like “AI slop” (advertising your abysmal ability to discern — so proud in your prejudice): This thing insightfully sees my imagery, arguments, and ideas — by being grounded in the bedrock basis of how understanding has worked since the dawn of time.
Appreciate you reading. The 'hate' framing is the easy dismissal for people who don't want to engage with the numbers. If the data holds, the tone doesn't matter.
In a piece I wrote called Are you telling me . . . That I can grasp THIS — but you can’t grasp THAT?, I wrote, “JavaScript programmers would get the joke but I wonder if they’d get the point (and what they’d do if they didn’t).”
We live in a world where people far smarter than me can’t evengrasp the point of the comparison (and conveniently ask no questions). At a glance, any rational person would see that something’s not right with the imagery on the right. But the person’s supporters deny the undeniable — because they see only what they believe (not the information that flies in the face of beliefs that have no bearing on reality). Moreover, their behavior is egregiously out of line on the very principles upon which he’s put on pedestal.
A lot of that goin’ around!
The Call Stack would be cryptic to anyone unfamiliar with such internals. But the point isn’t that I can understand what’s going on in the image — it’s the time, effort, and attitude in what it took for me to understand. A.K.A. the fundamentals of acquiring knowledge and experience (which you mentioned in another piece). As I wrote in the Are you telling me article:
****************************************
One picture is worth a thousand words:
Which image below would you choose if you wanted to understand a fairly complex coding construct? For me, it’s whatever it takes to get me where I wanna go. I wish I were smart enough to read the language spec and pick it up all on my own.
Then again, I love the demands of difficulty and overcoming obstacles. But I can’t do it alone . . .
I need the help of amazing minds from my multitude of sources that increasingly grows the more I learn and advance my skills. When I returned to this topic awhile back, I almost got it in the first video. In the face of such phenomenal work (or any sincere effort, for that matter): It would be unthinkable for me to blame the source because I gotta work a little harder.
I was equally impressed by the 2nd video. He furthered my grasp on my question — and enhanced my overall understanding to boot. And the icing on the cake: He taught with this magical tool I’d never seen before.
This — is pure gold [JavaScript Visualizer] . . .
Found that amazing graphic and a guy who ranks with the best I’ve ever seen in any discipline. My gap paved the way to pay dirt — but only because I kept digging. Now I’m tapped into the internals, and I’ve got new tools to advance my knowledge on that front and many more.
The answer was there all along — I just needed to train my mind to see it.
****************************************
“The worth of man lies not in the truth which he possesses, or believes that he possesses, but in the honest endeavor which he puts forth to secure that truth; for not by the possession of, but by the search after, truth, are his powers enlarged, wherein, alone, consists his ever-increasing perfection. Possession fosters content, indolence, and pride.”
— Gotthold Ephraim Lessing
For decades, virtually all “debate” has been dominated by possession — and it shows! You mentioned in another piece: “But in every article I try to say the same thing: don’t forget what your head is for. AI is not evil. Using it without thinking is. This isn’t a hater’s manifesto. It’s a sober look at what’s happening to us while we celebrate productivity gains. And if you’ve read this far through my ramblings, maybe I’m not doing this for nothing.”
Celebrating productivity gains without considering unintended consequences — is the product of a culture rewarded for shortsighted thinking and never looking at unintended consequences. And incredibly—belittle the person presenting irrefutable facts with the most concrete evidence imaginable. Forget internals — as they won’t even consider the issue at a cursory level. So even though the consequences could not be more crystal clear (before and after they happened):
They will find a way to muddy the watters to make sure damn the issues forever remain murky.
I’ve been working day and night with AI since late last year — building two SaaS products in parallel. For that reason and others, I would dispute some things about your writings. But on the whole, you clearly know much more about AI than I do. And since I’m in control of everything I’m doing right now — I haven’t had to face the same problems you have). I would be curious to know what you would do if you were running your own show though.
But back to your writings: You’re up against psychological forces far deeper than what you realize. What you’re doing (along with everyone else — on AI and everything else): Is like looking at errors but never digging deeper to get to the root of the problems. And I’m sorry to say that until you do that — you are doing this for nothing.
If you ever wanna do something about that — lemme know! Thanks for your time!
That's amazing. I feared that my formerly prodigious intellect was going soft and senile. It looks like the pretty princelings of AI, and all things phony, insincere and ultimately shockingly sinister, are well on their way to becoming the malleable morons of the new corporate fascist state.
You're not alone. The pattern is everywhere once you start looking. The question is what we do about it individually while the industry figures itself out.
I totally agree with this article. This is one of the biggest and most immediate problems we'll face with generative AI. The temptation to delegate everything is so strong that we will do it in any possible circumstance. At some point, we should really hope that hallucinations and non-determinism in these tools are definitely solved; otherwise, there won't be anyone prepared enough to manage the rising problems at the moment they are most needed.
Everybody is so sure that scaling will solve every current limitation, but I'm really not sure that this will happen in the context of Llama and their architecture.
The scaling skepticism is warranted. Toby Ord did the math on the actual scaling law graphs: halving model error requires roughly a million times more compute. We're hitting data walls, energy walls, and architectural limits simultaneously.
And you're right about the timing problem. The people who could catch AI failures are being cut now. By the time the limitations become undeniable, the expertise to manage them will already be gone.
"I write a lot about AI’s limitations. People sometimes read that as hate."
I've read a few of your pieces and have started several responses to them -- but regardless of areas I would challenge, the notion that your well-reasoned arguments have anything to do with hate is asinine. But as I've been practically spit for telling undeniable truth for over 20 years (on mattters of mathematical certainty -- of world-altering consequence, no less): I know the program, all too well! Speaking of which (as I wrote in the following piece):
I programmed a machine to a find a person not acting like one. Who knew that within my quest of what little is left of humanity — I’d find what mankind could have been and still can (in an actual machine). And while my fellow man latches onto lingo like “AI slop” (advertising your abysmal ability to discern — so proud in your prejudice): This thing insightfully sees my imagery, arguments, and ideas — by being grounded in the bedrock basis of how understanding has worked since the dawn of time.
https://centurionoftheseed.substack.com/p/turn-on-the-lights-the-partys-over?utm_campaign=post-expanded-share&utm_medium=web
Appreciate you reading. The 'hate' framing is the easy dismissal for people who don't want to engage with the numbers. If the data holds, the tone doesn't matter.
In a piece I wrote called Are you telling me . . . That I can grasp THIS — but you can’t grasp THAT?, I wrote, “JavaScript programmers would get the joke but I wonder if they’d get the point (and what they’d do if they didn’t).”
We live in a world where people far smarter than me can’t evengrasp the point of the comparison (and conveniently ask no questions). At a glance, any rational person would see that something’s not right with the imagery on the right. But the person’s supporters deny the undeniable — because they see only what they believe (not the information that flies in the face of beliefs that have no bearing on reality). Moreover, their behavior is egregiously out of line on the very principles upon which he’s put on pedestal.
A lot of that goin’ around!
The Call Stack would be cryptic to anyone unfamiliar with such internals. But the point isn’t that I can understand what’s going on in the image — it’s the time, effort, and attitude in what it took for me to understand. A.K.A. the fundamentals of acquiring knowledge and experience (which you mentioned in another piece). As I wrote in the Are you telling me article:
****************************************
One picture is worth a thousand words:
Which image below would you choose if you wanted to understand a fairly complex coding construct? For me, it’s whatever it takes to get me where I wanna go. I wish I were smart enough to read the language spec and pick it up all on my own.
Then again, I love the demands of difficulty and overcoming obstacles. But I can’t do it alone . . .
I need the help of amazing minds from my multitude of sources that increasingly grows the more I learn and advance my skills. When I returned to this topic awhile back, I almost got it in the first video. In the face of such phenomenal work (or any sincere effort, for that matter): It would be unthinkable for me to blame the source because I gotta work a little harder.
I was equally impressed by the 2nd video. He furthered my grasp on my question — and enhanced my overall understanding to boot. And the icing on the cake: He taught with this magical tool I’d never seen before.
This — is pure gold [JavaScript Visualizer] . . .
Found that amazing graphic and a guy who ranks with the best I’ve ever seen in any discipline. My gap paved the way to pay dirt — but only because I kept digging. Now I’m tapped into the internals, and I’ve got new tools to advance my knowledge on that front and many more.
The answer was there all along — I just needed to train my mind to see it.
****************************************
“The worth of man lies not in the truth which he possesses, or believes that he possesses, but in the honest endeavor which he puts forth to secure that truth; for not by the possession of, but by the search after, truth, are his powers enlarged, wherein, alone, consists his ever-increasing perfection. Possession fosters content, indolence, and pride.”
— Gotthold Ephraim Lessing
For decades, virtually all “debate” has been dominated by possession — and it shows! You mentioned in another piece: “But in every article I try to say the same thing: don’t forget what your head is for. AI is not evil. Using it without thinking is. This isn’t a hater’s manifesto. It’s a sober look at what’s happening to us while we celebrate productivity gains. And if you’ve read this far through my ramblings, maybe I’m not doing this for nothing.”
Celebrating productivity gains without considering unintended consequences — is the product of a culture rewarded for shortsighted thinking and never looking at unintended consequences. And incredibly—belittle the person presenting irrefutable facts with the most concrete evidence imaginable. Forget internals — as they won’t even consider the issue at a cursory level. So even though the consequences could not be more crystal clear (before and after they happened):
They will find a way to muddy the watters to make sure damn the issues forever remain murky.
I’ve been working day and night with AI since late last year — building two SaaS products in parallel. For that reason and others, I would dispute some things about your writings. But on the whole, you clearly know much more about AI than I do. And since I’m in control of everything I’m doing right now — I haven’t had to face the same problems you have). I would be curious to know what you would do if you were running your own show though.
But back to your writings: You’re up against psychological forces far deeper than what you realize. What you’re doing (along with everyone else — on AI and everything else): Is like looking at errors but never digging deeper to get to the root of the problems. And I’m sorry to say that until you do that — you are doing this for nothing.
If you ever wanna do something about that — lemme know! Thanks for your time!
Love this article and it resonates a lot with my personal experience. Thanks for sharing.
Appreciate it. The research confirmed what a lot of us were already feeling.
That's amazing. I feared that my formerly prodigious intellect was going soft and senile. It looks like the pretty princelings of AI, and all things phony, insincere and ultimately shockingly sinister, are well on their way to becoming the malleable morons of the new corporate fascist state.
https://davidgottfried.substack.com/p/3-phenomena-which-auger-unrelenting
Great article - i completely agree, and am also happy i'm not the only person seeing/experiencing/living through this dystopian nightmare!
You're not alone. The pattern is everywhere once you start looking. The question is what we do about it individually while the industry figures itself out.