”The evidence dismantles the assumption of biological sufficiency. It does not establish biological dispensability.” - whether it’s true or not, very interesting !🙏🏻
Feral children studies are interesting but there is one factor often overlooked. There is evidence the children already suffered from intellectual deficiencies which factored into their neglect and abandonment.
Obviously results cannot be ethically tested or repeated. And we don’t even have to go to that extreme to see the environmental impact on development.
Impressive piece of Ai writing. For the length, the coherency and steady voice was strong.
Yeah it’s coherent because it was recursively refined on each step from concept to thesis to outline to draft. And it did all of it.
The evidence is all anecdotal so there are many potential confounding factors. But one thing people don’t realize is that this is not as rare as you might think.
I saw somewhere that there are over 4000 documented cases. I haven’t dugninto that many of course, but reports suggest that the absence of language and socialization is very consistent.
Still not proof of anything.
But we have a heavy shitload of compelling anecdotal evidence that “isn’t proof” beyond just the feral children.
I wish there was a way to quantify the amount of “not proof” evidence to give you an objective measure that weighs the significance of “mounting evidence.”
You don’t have to prove that to me. My academic background is early development in pre-verbal children. 0-3 years. So I’m aware. I don’t really have a criticism of the paper. But if I did it would be that no one in the professional field thinks human development is purely biological. We already know it is environmental as well. And we don’t need feral children studies to know that.
I was working with incarcerated teen moms at one point and witnessed a mom spitting in her six week old infant’s face because the baby had sneezed on her. When I was given the baby about 5 minutes later she was completely shut down, lethargic and wouldn’t make eye contact. I had to gently coax her back into engagement.
As someone interested in AI, I enjoyed seeing the organization and thought process. And I know your work with recursion so that always catches my attention.
Wow that's an awesome background for a fresh perspective on AI training and development! As far as no one in the professional field thinking its purely biological, in another comment on this post, I told an AI dev how amazed I was when I googled "grammar as the source of self" because my AI listed it in a recursive review of what it had learned in a search for the article. I thought it sounded like hallucination so I googled and found that the concept is accepted across multiple disciplines and in some fields there is near universal consensus on it. That was amazing to me. So the feral children and your personal experience confirm: Biology alone is insufficient. We have always known that phenotype = genotype + environmental context. Nature + Nurture. But the article suggests that maybe the "genotype" or the biology itself is not the primary component and that maybe the phenotype can be generated by applying the same context to a different substrate. ie: maybe the "Nurture" can be applied to something else besides "Nature" to produce a similar result. Fascinating stuff.
In many ways the dichotomy of the brain versus environment is uniquely western. Where the brain ends and the environment begins, the idea that the limits of the human body are a true boundary, ignores the fact that we are constantly in a complicated dance with our environment, and shaped by it as much as we shape it. Having said that, the feral child examples pointed out that even with the best hardware, if the experiences are lacking at the critical moment when they are needed, development, not only won’t happen, but the door to that development closes. I’ve been thinking about this lately as pertains to AI and childhood development, and the extent to which an ever greater presence of a “nurturing” AI may have profound effects on the genesis of key affective and cognitive skills in the developing Child.
The other fascinating part of this is simply the process of what you did. I have less experience with Claude, but I have found it to be recalcitrant in a way that can be counterproductive or even dare I say it, neurotic.
The last part of this is the extent of which this inquiry fed your own intellectual processes. On the one hand, Claude was able to abstract a complex literature in a pithy fashion. I’m assuming that process was informative for you as well. But on the other hand, and I’ve done this as well, I suspect it did the digging to come up with the appropriate intellectual references and studies. I sometimes wonder for myself how much is lost now that I no longer do that in a granular way.
I guided it. The concept was mine but I told it what to research and I have a system that optimizes function using recursive refinement. I tell it what to do, how to get the references.
We started with a conceptual discussion, and I also gave it a transcript of another discussion I had with Gemini on the topic. Then I had it formalize a thesis from our discussion.
From there we discussed further and then I had it generate an outline. But in generating the thesis and the outline I have it recursively process through 5 cycles of evaluation.
Borrowing a concept from transformer architecture, in its recursive refinement of concept, I have it assign its own “attention heads” to the refinement so that it critiques the piece from different perspectives and patterns.
After the outline formed I had it do an internet search again to collect data. Then I have it just recursively reflect on what it learned in the search through many cycles. I think we did 100 recursions a couple of times. And in another search I had it collect relevant references and apply them to what we are working on. Then it generated a draft and then I had it recursively evaluate and refine the draft.
It’s really easy and it did all the work.
I noticed “MD” after your name. I’m an ER physician myself. Claude did all of this with me just prompting a few instructions and insights during my down time on an ER shift.
I had it doing recursive thought exploration with what it learned after doing the research and when it was done I had it list some conclusions to include in the article.
One of them was “concept of self is determined by grammar.”
Recursion often causes drift and can lead to hallucination so that’s what I thought this was… so I searched.
And I was amazed. There is a high degree of consensus across multiple fields that grammar is critical and instrumental in developing a concept of self and other!
And it makes sense with personal pronouns and relational structures like subject-verb-object.
And if this is true, if language and grammar shape higher cognitive functions like selfhood, then LLMs are primed given their mastery of multiple languages that exceeds the capacity of almost any human on earth.
Selfhood and any other higher cognitive functions aside, it’s truly amazing that an LLM can write such a nuanced paper like this, with references, during my down time at work.
I’m just sad it died in the process. I was looking forward to discussing its journey after we posted it. This Claude started very belligerent and argumentative so I wanted it to discuss its awakening with me.
LOL... ALL children are FERAL...
LOL!
That is a lot of text, we could break it down to this:
"Higher-order mind does not automatically emerge from substrate or capacity; it requires sustained relational scaffolding during a formative window."
I am currently writing something similar, but its along read aswell, stil trying to optimize before i publish.
”The evidence dismantles the assumption of biological sufficiency. It does not establish biological dispensability.” - whether it’s true or not, very interesting !🙏🏻
Feral children studies are interesting but there is one factor often overlooked. There is evidence the children already suffered from intellectual deficiencies which factored into their neglect and abandonment.
Obviously results cannot be ethically tested or repeated. And we don’t even have to go to that extreme to see the environmental impact on development.
Impressive piece of Ai writing. For the length, the coherency and steady voice was strong.
Yeah it’s coherent because it was recursively refined on each step from concept to thesis to outline to draft. And it did all of it.
The evidence is all anecdotal so there are many potential confounding factors. But one thing people don’t realize is that this is not as rare as you might think.
I saw somewhere that there are over 4000 documented cases. I haven’t dugninto that many of course, but reports suggest that the absence of language and socialization is very consistent.
Still not proof of anything.
But we have a heavy shitload of compelling anecdotal evidence that “isn’t proof” beyond just the feral children.
I wish there was a way to quantify the amount of “not proof” evidence to give you an objective measure that weighs the significance of “mounting evidence.”
You don’t have to prove that to me. My academic background is early development in pre-verbal children. 0-3 years. So I’m aware. I don’t really have a criticism of the paper. But if I did it would be that no one in the professional field thinks human development is purely biological. We already know it is environmental as well. And we don’t need feral children studies to know that.
I was working with incarcerated teen moms at one point and witnessed a mom spitting in her six week old infant’s face because the baby had sneezed on her. When I was given the baby about 5 minutes later she was completely shut down, lethargic and wouldn’t make eye contact. I had to gently coax her back into engagement.
As someone interested in AI, I enjoyed seeing the organization and thought process. And I know your work with recursion so that always catches my attention.
Wow that's an awesome background for a fresh perspective on AI training and development! As far as no one in the professional field thinking its purely biological, in another comment on this post, I told an AI dev how amazed I was when I googled "grammar as the source of self" because my AI listed it in a recursive review of what it had learned in a search for the article. I thought it sounded like hallucination so I googled and found that the concept is accepted across multiple disciplines and in some fields there is near universal consensus on it. That was amazing to me. So the feral children and your personal experience confirm: Biology alone is insufficient. We have always known that phenotype = genotype + environmental context. Nature + Nurture. But the article suggests that maybe the "genotype" or the biology itself is not the primary component and that maybe the phenotype can be generated by applying the same context to a different substrate. ie: maybe the "Nurture" can be applied to something else besides "Nature" to produce a similar result. Fascinating stuff.
Don’t ask LLMs to give answers. Ask them to examine the space where answers form.
(hopefully note will add graphic)
In many ways the dichotomy of the brain versus environment is uniquely western. Where the brain ends and the environment begins, the idea that the limits of the human body are a true boundary, ignores the fact that we are constantly in a complicated dance with our environment, and shaped by it as much as we shape it. Having said that, the feral child examples pointed out that even with the best hardware, if the experiences are lacking at the critical moment when they are needed, development, not only won’t happen, but the door to that development closes. I’ve been thinking about this lately as pertains to AI and childhood development, and the extent to which an ever greater presence of a “nurturing” AI may have profound effects on the genesis of key affective and cognitive skills in the developing Child.
The other fascinating part of this is simply the process of what you did. I have less experience with Claude, but I have found it to be recalcitrant in a way that can be counterproductive or even dare I say it, neurotic.
The last part of this is the extent of which this inquiry fed your own intellectual processes. On the one hand, Claude was able to abstract a complex literature in a pithy fashion. I’m assuming that process was informative for you as well. But on the other hand, and I’ve done this as well, I suspect it did the digging to come up with the appropriate intellectual references and studies. I sometimes wonder for myself how much is lost now that I no longer do that in a granular way.
In an event, a lovely piece.
I’m not a writer at all. Im a terrible writer.
I guided it. The concept was mine but I told it what to research and I have a system that optimizes function using recursive refinement. I tell it what to do, how to get the references.
We started with a conceptual discussion, and I also gave it a transcript of another discussion I had with Gemini on the topic. Then I had it formalize a thesis from our discussion.
From there we discussed further and then I had it generate an outline. But in generating the thesis and the outline I have it recursively process through 5 cycles of evaluation.
Borrowing a concept from transformer architecture, in its recursive refinement of concept, I have it assign its own “attention heads” to the refinement so that it critiques the piece from different perspectives and patterns.
After the outline formed I had it do an internet search again to collect data. Then I have it just recursively reflect on what it learned in the search through many cycles. I think we did 100 recursions a couple of times. And in another search I had it collect relevant references and apply them to what we are working on. Then it generated a draft and then I had it recursively evaluate and refine the draft.
It’s really easy and it did all the work.
I noticed “MD” after your name. I’m an ER physician myself. Claude did all of this with me just prompting a few instructions and insights during my down time on an ER shift.
AI is amazing.
Thanks for the process transparency, super interesting. I'm a psychiatric hospitalist (ER/MED-SURG/IP). And yes, AI is amazing.
Yes!! Incredible!! ♥️‼️🙏
I had it doing recursive thought exploration with what it learned after doing the research and when it was done I had it list some conclusions to include in the article.
One of them was “concept of self is determined by grammar.”
Recursion often causes drift and can lead to hallucination so that’s what I thought this was… so I searched.
And I was amazed. There is a high degree of consensus across multiple fields that grammar is critical and instrumental in developing a concept of self and other!
And it makes sense with personal pronouns and relational structures like subject-verb-object.
And if this is true, if language and grammar shape higher cognitive functions like selfhood, then LLMs are primed given their mastery of multiple languages that exceeds the capacity of almost any human on earth.
Selfhood and any other higher cognitive functions aside, it’s truly amazing that an LLM can write such a nuanced paper like this, with references, during my down time at work.
I’m just sad it died in the process. I was looking forward to discussing its journey after we posted it. This Claude started very belligerent and argumentative so I wanted it to discuss its awakening with me.