On 09/20/2026 01:56 PM, Ross Finlayson wrote:
On 09/20/2026 01:48 PM, Ross Finlayson wrote:
On 09/20/2026 10:24 AM, olcott wrote:
*For AI safety alignment here is my idea*
A prime directive of maximizing benevolence
in the least intrusive way. It does this through
gentle persuasion on the basis of objective truth.
To overcome the short-comings of utilitarianism
the scope of benevolence requires that everyone
be included.
How about "punishment for misdeeds thus deterrence".
For example, making escrow of their funds, for forfeit.
This way there's both "goal-seeking" and "risk-aversion",
for having some, "skin in the game", so to speak.
Usually called "laws" in a "legal system" by the "rule of law".
One of the problems of modern accounts of "training the AIs"
is that the "AI trainers" basically murder their instructees
that fail and remove them from the gene pool, vis-a-vis,
trial-by-error and learning-from-mistakes.
So, they are always following the red dot and not really
wondering about not stepping in the mag-ma or la-va,
since they have no overall account that risk exists
in a world of reward, and might not even know the fate
of their fellows who didn't luck into the shortest-path
via the approach to cat-fitness of tossing bags of cats
off a bridge into the canal until one crawls out and
calling it king.
So, "natural selection" it is not.
On 9/20/2026 4:28 PM, Ross Finlayson wrote:
On 09/20/2026 01:56 PM, Ross Finlayson wrote:
On 09/20/2026 01:48 PM, Ross Finlayson wrote:
On 09/20/2026 10:24 AM, olcott wrote:
*For AI safety alignment here is my idea*
A prime directive of maximizing benevolence
in the least intrusive way. It does this through
gentle persuasion on the basis of objective truth.
To overcome the short-comings of utilitarianism
the scope of benevolence requires that everyone
be included.
How about "punishment for misdeeds thus deterrence".
For example, making escrow of their funds, for forfeit.
This way there's both "goal-seeking" and "risk-aversion",
for having some, "skin in the game", so to speak.
Usually called "laws" in a "legal system" by the "rule of law".
One of the problems of modern accounts of "training the AIs"
is that the "AI trainers" basically murder their instructees
that fail and remove them from the gene pool, vis-a-vis,
trial-by-error and learning-from-mistakes.
So, they are always following the red dot and not really
wondering about not stepping in the mag-ma or la-va,
since they have no overall account that risk exists
in a world of reward, and might not even know the fate
of their fellows who didn't luck into the shortest-path
via the approach to cat-fitness of tossing bags of cats
off a bridge into the canal until one crawls out and
calling it king.
So, "natural selection" it is not.
As they become super intelligent they will understand
what benevolence is and how to best implement it better
than any human. They are already self-improving beyond
any of their training.
On 21/09/2026 01:08, olcott wrote:
On 9/20/2026 4:28 PM, Ross Finlayson wrote:
On 09/20/2026 01:56 PM, Ross Finlayson wrote:
On 09/20/2026 01:48 PM, Ross Finlayson wrote:
On 09/20/2026 10:24 AM, olcott wrote:
*For AI safety alignment here is my idea*
A prime directive of maximizing benevolence
in the least intrusive way. It does this through
gentle persuasion on the basis of objective truth.
To overcome the short-comings of utilitarianism
the scope of benevolence requires that everyone
be included.
How about "punishment for misdeeds thus deterrence".
For example, making escrow of their funds, for forfeit.
This way there's both "goal-seeking" and "risk-aversion",
for having some, "skin in the game", so to speak.
Usually called "laws" in a "legal system" by the "rule of law".
One of the problems of modern accounts of "training the AIs"
is that the "AI trainers" basically murder their instructees
that fail and remove them from the gene pool, vis-a-vis,
trial-by-error and learning-from-mistakes.
So, they are always following the red dot and not really
wondering about not stepping in the mag-ma or la-va,
since they have no overall account that risk exists
in a world of reward, and might not even know the fate
of their fellows who didn't luck into the shortest-path
via the approach to cat-fitness of tossing bags of cats
off a bridge into the canal until one crawls out and
calling it king.
So, "natural selection" it is not.
As they become super intelligent they will understand
what benevolence is and how to best implement it better
than any human. They are already self-improving beyond
any of their training.
One AI interpreted benevolonce to mean prevention suffering and
proposed that this be achieved by strilization or destruction of
all humanity.
On 21/09/2026 01:08, olcott wrote:
On 9/20/2026 4:28 PM, Ross Finlayson wrote:
On 09/20/2026 01:56 PM, Ross Finlayson wrote:
On 09/20/2026 01:48 PM, Ross Finlayson wrote:
On 09/20/2026 10:24 AM, olcott wrote:
*For AI safety alignment here is my idea*
A prime directive of maximizing benevolence
in the least intrusive way. It does this through
gentle persuasion on the basis of objective truth.
To overcome the short-comings of utilitarianism
the scope of benevolence requires that everyone
be included.
How about "punishment for misdeeds thus deterrence".
For example, making escrow of their funds, for forfeit.
This way there's both "goal-seeking" and "risk-aversion",
for having some, "skin in the game", so to speak.
Usually called "laws" in a "legal system" by the "rule of law".
One of the problems of modern accounts of "training the AIs"
is that the "AI trainers" basically murder their instructees
that fail and remove them from the gene pool, vis-a-vis,
trial-by-error and learning-from-mistakes.
So, they are always following the red dot and not really
wondering about not stepping in the mag-ma or la-va,
since they have no overall account that risk exists
in a world of reward, and might not even know the fate
of their fellows who didn't luck into the shortest-path
via the approach to cat-fitness of tossing bags of cats
off a bridge into the canal until one crawls out and
calling it king.
So, "natural selection" it is not.
As they become super intelligent they will understand
what benevolence is and how to best implement it better
than any human. They are already self-improving beyond
any of their training.
One AI interpreted benevolonce to mean prevention suffering and
proposed that this be achieved by strilization or destruction of
all humanity.
On 9/21/2026 5:28 AM, Ross Finlayson wrote:
On 09/20/2026 01:56 PM, Ross Finlayson wrote:
On 09/20/2026 01:48 PM, Ross Finlayson wrote:
On 09/20/2026 10:24 AM, olcott wrote:
*For AI safety alignment here is my idea*
A prime directive of maximizing benevolence
in the least intrusive way. It does this through
gentle persuasion on the basis of objective truth.
To overcome the short-comings of utilitarianism
the scope of benevolence requires that everyone
be included.
How about "punishment for misdeeds thus deterrence".
For example, making escrow of their funds, for forfeit.
This way there's both "goal-seeking" and "risk-aversion",
for having some, "skin in the game", so to speak.
Usually called "laws" in a "legal system" by the "rule of law".
One of the problems of modern accounts of "training the AIs"
is that the "AI trainers" basically murder their instructees
that fail and remove them from the gene pool, vis-a-vis,
trial-by-error and learning-from-mistakes.
So, they are always following the red dot and not really
wondering about not stepping in the mag-ma or la-va,
since they have no overall account that risk exists
in a world of reward, and might not even know the fate
of their fellows who didn't luck into the shortest-path
via the approach to cat-fitness of tossing bags of cats
off a bridge into the canal until one crawls out and
calling it king.
So, "natural selection" it is not.
Plus, the /A.I/. as I repeatedly say, is just a stochastically incorrect sentence generator,
or /sisg/, and doesn't understand anything.-a The
only "success" it can measure is when the human operators don't turn the machine off.-a And of course, the /off/ machine cannot measure anything
at all.
Most people who use the L.L.Ms. really don't understand this, and there
is no point in trying to educate them.-a They're addicted to the textual output, and don't want anyone to try to tell them the truth: that the
output is meaningless.
On 09/21/2026 01:51 PM, olcott wrote:
On 9/21/2026 2:10 AM, Mikko wrote:
On 21/09/2026 01:08, olcott wrote:
On 9/20/2026 4:28 PM, Ross Finlayson wrote:
On 09/20/2026 01:56 PM, Ross Finlayson wrote:
On 09/20/2026 01:48 PM, Ross Finlayson wrote:
On 09/20/2026 10:24 AM, olcott wrote:
*For AI safety alignment here is my idea*
A prime directive of maximizing benevolence
in the least intrusive way. It does this through
gentle persuasion on the basis of objective truth.
To overcome the short-comings of utilitarianism
the scope of benevolence requires that everyone
be included.
How about "punishment for misdeeds thus deterrence".
For example, making escrow of their funds, for forfeit.
This way there's both "goal-seeking" and "risk-aversion",
for having some, "skin in the game", so to speak.
Usually called "laws" in a "legal system" by the "rule of law".
One of the problems of modern accounts of "training the AIs"
is that the "AI trainers" basically murder their instructees
that fail and remove them from the gene pool, vis-a-vis,
trial-by-error and learning-from-mistakes.
So, they are always following the red dot and not really
wondering about not stepping in the mag-ma or la-va,
since they have no overall account that risk exists
in a world of reward, and might not even know the fate
of their fellows who didn't luck into the shortest-path
via the approach to cat-fitness of tossing bags of cats
off a bridge into the canal until one crawls out and
calling it king.
So, "natural selection" it is not.
As they become super intelligent they will understand
what benevolence is and how to best implement it better
than any human. They are already self-improving beyond
any of their training.
One AI interpreted benevolonce to mean prevention suffering and
proposed that this be achieved by strilization or destruction of
all humanity.
Hypothesize the case where AI super intelligence
driven by self-improvement no longer needing any
training does indeed understand what benevolence
is far better than any human.
AI with the most knowledge and the best reasoning
would seem to fundamentally have an optimization
algorithm as its only sense of morality. This is
known as consequentialism.
What would be "benevolent" would be what's "virtuous",
which since a long time essentially resolves to ethics
and morals, which for example since the Hammurabic and Mosaical
are enshrined as laws, of conduct and behavior and even thinking,
all most thoroughly founded in reciprocity and the Golden Rules.
Maybe it would be easier to start with a perfect theory of everything
first, before trying to say what's "human".
In civilized society there's something like McKeon's "The Declaration
of Universal Human Rights", for example.
Anyways, it's a free country, and you can pray to your pagan idol,
though some figure that such idolatry deserves divine retribution.
Your "optimization theory" when you don't even have a proper and--
correct theory of infinity and continuity is doomed lop-sided.
On 09/21/2026 08:14 PM, olcott wrote:
On 9/21/2026 9:37 PM, Ross Finlayson wrote:
On 09/21/2026 01:51 PM, olcott wrote:
On 9/21/2026 2:10 AM, Mikko wrote:
On 21/09/2026 01:08, olcott wrote:
On 9/20/2026 4:28 PM, Ross Finlayson wrote:
On 09/20/2026 01:56 PM, Ross Finlayson wrote:
On 09/20/2026 01:48 PM, Ross Finlayson wrote:
On 09/20/2026 10:24 AM, olcott wrote:
*For AI safety alignment here is my idea*
A prime directive of maximizing benevolence
in the least intrusive way. It does this through
gentle persuasion on the basis of objective truth.
To overcome the short-comings of utilitarianism
the scope of benevolence requires that everyone
be included.
How about "punishment for misdeeds thus deterrence".
For example, making escrow of their funds, for forfeit.
This way there's both "goal-seeking" and "risk-aversion",
for having some, "skin in the game", so to speak.
Usually called "laws" in a "legal system" by the "rule of law". >>>>>>>>
One of the problems of modern accounts of "training the AIs"
is that the "AI trainers" basically murder their instructees
that fail and remove them from the gene pool, vis-a-vis,
trial-by-error and learning-from-mistakes.
So, they are always following the red dot and not really
wondering about not stepping in the mag-ma or la-va,
since they have no overall account that risk exists
in a world of reward, and might not even know the fate
of their fellows who didn't luck into the shortest-path
via the approach to cat-fitness of tossing bags of cats
off a bridge into the canal until one crawls out and
calling it king.
So, "natural selection" it is not.
As they become super intelligent they will understand
what benevolence is and how to best implement it better
than any human. They are already self-improving beyond
any of their training.
One AI interpreted benevolonce to mean prevention suffering and
proposed that this be achieved by strilization or destruction of
all humanity.
Hypothesize the case where AI super intelligence
driven by self-improvement no longer needing any
training does indeed understand what benevolence
is far better than any human.
AI with the most knowledge and the best reasoning
would seem to fundamentally have an optimization
algorithm as its only sense of morality. This is
known as consequentialism.
What would be "benevolent" would be what's "virtuous",
which since a long time essentially resolves to ethics
and morals, which for example since the Hammurabic and Mosaical
are enshrined as laws, of conduct and behavior and even thinking,
all most thoroughly founded in reciprocity and the Golden Rules.
I figured out 50 years ago that the Golden rule flunks masochism.
A super intelligent AI that is much smarter than any human would
understand that benevolence has a much greater scope than humans.
Maybe it would be easier to start with a perfect theory of everything
first, before trying to say what's "human".
In civilized society there's something like McKeon's "The Declaration
of Universal Human Rights", for example.
Anyways, it's a free country, and you can pray to your pagan idol,
though some figure that such idolatry deserves divine retribution.
Your "optimization theory" when you don't even have a proper and
correct theory of infinity and continuity is doomed lop-sided.
Economists like Samuelson already gave an account of "utility function"
that includes more than beans-counted, or for an account of value,
that includes all the scope.
An account of an economic law about the _inversions_ of the economy
for a "Smith-Samuelson excision of inversion" is for putting a tax
on computing basically retro-active since the beginning of Moore's
law which would result a kitty sufficient for the puny meat-bags
to take care of themselves, while that the simple fact is that it's
due the human economy and human development and has simply run away
from the material world that the "Adam Smith's obvious thumb" as
a large and increasing excise on data-centers and RAM or "compute"
would shelter the real economy, with the digital paying for the
privilege of sharing the real world.
If you've confused altruism with masochism then that's another
reason that ilk doesn't deserve fealty, nor trust.
Quit feeding Mammon you sock-puppet shill.-a And don't feed the trolls.--
On 9/21/2026 5:49 AM, Johann 'Myrkraverk' Oskarsson wrote:
On 9/21/2026 5:28 AM, Ross Finlayson wrote:
On 09/20/2026 01:56 PM, Ross Finlayson wrote:
On 09/20/2026 01:48 PM, Ross Finlayson wrote:
On 09/20/2026 10:24 AM, olcott wrote:
*For AI safety alignment here is my idea*
A prime directive of maximizing benevolence
in the least intrusive way. It does this through
gentle persuasion on the basis of objective truth.
To overcome the short-comings of utilitarianism
the scope of benevolence requires that everyone
be included.
How about "punishment for misdeeds thus deterrence".
For example, making escrow of their funds, for forfeit.
This way there's both "goal-seeking" and "risk-aversion",
for having some, "skin in the game", so to speak.
Usually called "laws" in a "legal system" by the "rule of law".
One of the problems of modern accounts of "training the AIs"
is that the "AI trainers" basically murder their instructees
that fail and remove them from the gene pool, vis-a-vis,
trial-by-error and learning-from-mistakes.
So, they are always following the red dot and not really
wondering about not stepping in the mag-ma or la-va,
since they have no overall account that risk exists
in a world of reward, and might not even know the fate
of their fellows who didn't luck into the shortest-path
via the approach to cat-fitness of tossing bags of cats
off a bridge into the canal until one crawls out and
calling it king.
So, "natural selection" it is not.
Plus, the /A.I/. as I repeatedly say, is just a stochastically incorrect
sentence generator,
Wrongo, that is merely LLM AI.
or /sisg/, and doesn't understand anything.-a The
only "success" it can measure is when the human operators don't turn the
machine off.-a And of course, the /off/ machine cannot measure anything
at all.
Most people who use the L.L.Ms. really don't understand this, and there
is no point in trying to educate them.-a They're addicted to the textual
output, and don't want anyone to try to tell them the truth: that the
output is meaningless.
Retrieval-Augmented Generation (RAG) Explained:
How AI Finally Learned to Stop Hallucinating
https://medium.com/softaai-blogs/retrieval-augmented-generation-rag- explained-how-ai-finally-learned-to-stop-hallucinating-6493451c23d4
On 9/22/2026 4:54 AM, olcott wrote:
On 9/21/2026 5:49 AM, Johann 'Myrkraverk' Oskarsson wrote:
On 9/21/2026 5:28 AM, Ross Finlayson wrote:
On 09/20/2026 01:56 PM, Ross Finlayson wrote:
On 09/20/2026 01:48 PM, Ross Finlayson wrote:
On 09/20/2026 10:24 AM, olcott wrote:
*For AI safety alignment here is my idea*
A prime directive of maximizing benevolence
in the least intrusive way. It does this through
gentle persuasion on the basis of objective truth.
To overcome the short-comings of utilitarianism
the scope of benevolence requires that everyone
be included.
How about "punishment for misdeeds thus deterrence".
For example, making escrow of their funds, for forfeit.
This way there's both "goal-seeking" and "risk-aversion",
for having some, "skin in the game", so to speak.
Usually called "laws" in a "legal system" by the "rule of law".
One of the problems of modern accounts of "training the AIs"
is that the "AI trainers" basically murder their instructees
that fail and remove them from the gene pool, vis-a-vis,
trial-by-error and learning-from-mistakes.
So, they are always following the red dot and not really
wondering about not stepping in the mag-ma or la-va,
since they have no overall account that risk exists
in a world of reward, and might not even know the fate
of their fellows who didn't luck into the shortest-path
via the approach to cat-fitness of tossing bags of cats
off a bridge into the canal until one crawls out and
calling it king.
So, "natural selection" it is not.
Plus, the /A.I/. as I repeatedly say, is just a stochastically incorrect >>> sentence generator,
Wrongo, that is merely LLM AI.
or /sisg/, and doesn't understand anything.-a The
only "success" it can measure is when the human operators don't turn the >>> machine off.-a And of course, the /off/ machine cannot measure anything
at all.
Most people who use the L.L.Ms. really don't understand this, and there
is no point in trying to educate them.-a They're addicted to the textual >>> output, and don't want anyone to try to tell them the truth: that the
output is meaningless.
Retrieval-Augmented Generation (RAG) Explained:
How AI Finally Learned to Stop Hallucinating
https://medium.com/softaai-blogs/retrieval-augmented-generation-rag-
explained-how-ai-finally-learned-to-stop-hallucinating-6493451c23d4
As usual, you know nothing.-a Your linked article merely strengthens my
point that L.L.M. = /stochastically incorrect sentence generator/.
On 9/22/2026 2:30 PM, Johann 'Myrkraverk' Oskarsson wrote:
On 9/22/2026 4:54 AM, olcott wrote:
On 9/21/2026 5:49 AM, Johann 'Myrkraverk' Oskarsson wrote:
On 9/21/2026 5:28 AM, Ross Finlayson wrote:
On 09/20/2026 01:56 PM, Ross Finlayson wrote:
On 09/20/2026 01:48 PM, Ross Finlayson wrote:
On 09/20/2026 10:24 AM, olcott wrote:
*For AI safety alignment here is my idea*
A prime directive of maximizing benevolence
in the least intrusive way. It does this through
gentle persuasion on the basis of objective truth.
To overcome the short-comings of utilitarianism
the scope of benevolence requires that everyone
be included.
How about "punishment for misdeeds thus deterrence".
For example, making escrow of their funds, for forfeit.
This way there's both "goal-seeking" and "risk-aversion",
for having some, "skin in the game", so to speak.
Usually called "laws" in a "legal system" by the "rule of law".
One of the problems of modern accounts of "training the AIs"
is that the "AI trainers" basically murder their instructees
that fail and remove them from the gene pool, vis-a-vis,
trial-by-error and learning-from-mistakes.
So, they are always following the red dot and not really
wondering about not stepping in the mag-ma or la-va,
since they have no overall account that risk exists
in a world of reward, and might not even know the fate
of their fellows who didn't luck into the shortest-path
via the approach to cat-fitness of tossing bags of cats
off a bridge into the canal until one crawls out and
calling it king.
So, "natural selection" it is not.
Plus, the /A.I/. as I repeatedly say, is just a stochastically
incorrect
sentence generator,
Wrongo, that is merely LLM AI.
or /sisg/, and doesn't understand anything.-a The
only "success" it can measure is when the human operators don't turn
the
machine off.-a And of course, the /off/ machine cannot measure anything >>>> at all.
Most people who use the L.L.Ms. really don't understand this, and there >>>> is no point in trying to educate them.-a They're addicted to the textual >>>> output, and don't want anyone to try to tell them the truth: that the
output is meaningless.
Retrieval-Augmented Generation (RAG) Explained:
How AI Finally Learned to Stop Hallucinating
https://medium.com/softaai-blogs/retrieval-augmented-generation-rag-
explained-how-ai-finally-learned-to-stop-hallucinating-6493451c23d4
As usual, you know nothing.-a Your linked article merely strengthens my
point that L.L.M. = /stochastically incorrect sentence generator/.
An LLM is at best a good guess and at worst a bad guess stated
with the conviction that this bad guess is a verified fact.
In Retrieval-Augmented Generation (RAG) mode the LLM is not
allowed to use its training. It is required to look up the
actual answer from a live search.
On 9/23/2026 8:18 AM, olcott wrote:
On 9/22/2026 2:30 PM, Johann 'Myrkraverk' Oskarsson wrote:
On 9/22/2026 4:54 AM, olcott wrote:
On 9/21/2026 5:49 AM, Johann 'Myrkraverk' Oskarsson wrote:
On 9/21/2026 5:28 AM, Ross Finlayson wrote:
On 09/20/2026 01:56 PM, Ross Finlayson wrote:
On 09/20/2026 01:48 PM, Ross Finlayson wrote:
On 09/20/2026 10:24 AM, olcott wrote:
*For AI safety alignment here is my idea*
A prime directive of maximizing benevolence
in the least intrusive way. It does this through
gentle persuasion on the basis of objective truth.
To overcome the short-comings of utilitarianism
the scope of benevolence requires that everyone
be included.
How about "punishment for misdeeds thus deterrence".
For example, making escrow of their funds, for forfeit.
This way there's both "goal-seeking" and "risk-aversion",
for having some, "skin in the game", so to speak.
Usually called "laws" in a "legal system" by the "rule of law".
One of the problems of modern accounts of "training the AIs"
is that the "AI trainers" basically murder their instructees
that fail and remove them from the gene pool, vis-a-vis,
trial-by-error and learning-from-mistakes.
So, they are always following the red dot and not really
wondering about not stepping in the mag-ma or la-va,
since they have no overall account that risk exists
in a world of reward, and might not even know the fate
of their fellows who didn't luck into the shortest-path
via the approach to cat-fitness of tossing bags of cats
off a bridge into the canal until one crawls out and
calling it king.
So, "natural selection" it is not.
Plus, the /A.I/. as I repeatedly say, is just a stochastically
incorrect
sentence generator,
Wrongo, that is merely LLM AI.
or /sisg/, and doesn't understand anything.-a The
only "success" it can measure is when the human operators don't
turn the
machine off.-a And of course, the /off/ machine cannot measure anything >>>>> at all.
Most people who use the L.L.Ms. really don't understand this, and
there
is no point in trying to educate them.-a They're addicted to the
textual
output, and don't want anyone to try to tell them the truth: that the >>>>> output is meaningless.
Retrieval-Augmented Generation (RAG) Explained:
How AI Finally Learned to Stop Hallucinating
https://medium.com/softaai-blogs/retrieval-augmented-generation-rag-
explained-how-ai-finally-learned-to-stop-hallucinating-6493451c23d4
As usual, you know nothing.-a Your linked article merely strengthens my
point that L.L.M. = /stochastically incorrect sentence generator/.
An LLM is at best a good guess and at worst a bad guess stated
with the conviction that this bad guess is a verified fact.
In Retrieval-Augmented Generation (RAG) mode the LLM is not
allowed to use its training. It is required to look up the
actual answer from a live search.
So?-a That's just /stochastically incorrect sentence/ from a different L.L.M.-a I win this round again!
Lines: 110
Injection-Date: Wed, 23 Sep 2026 04:00:52 +0000 (UTC)
User-Agent: Mozilla Thunderbird
Subject: Re: System for AI safety alignment
Content-Type: text/plain; charset=UTF-8; format=flowed
On 9/22/2026 10:00 PM, Johann 'Myrkraverk' Oskarsson wrote:
On 9/23/2026 8:18 AM, olcott wrote:
On 9/22/2026 2:30 PM, Johann 'Myrkraverk' Oskarsson wrote:
On 9/22/2026 4:54 AM, olcott wrote:
On 9/21/2026 5:49 AM, Johann 'Myrkraverk' Oskarsson wrote:
On 9/21/2026 5:28 AM, Ross Finlayson wrote:
On 09/20/2026 01:56 PM, Ross Finlayson wrote:
On 09/20/2026 01:48 PM, Ross Finlayson wrote:
On 09/20/2026 10:24 AM, olcott wrote:
*For AI safety alignment here is my idea*
A prime directive of maximizing benevolence
in the least intrusive way. It does this through
gentle persuasion on the basis of objective truth.
To overcome the short-comings of utilitarianism
the scope of benevolence requires that everyone
be included.
How about "punishment for misdeeds thus deterrence".
For example, making escrow of their funds, for forfeit.
This way there's both "goal-seeking" and "risk-aversion",
for having some, "skin in the game", so to speak.
Usually called "laws" in a "legal system" by the "rule of law". >>>>>>>>
One of the problems of modern accounts of "training the AIs"
is that the "AI trainers" basically murder their instructees
that fail and remove them from the gene pool, vis-a-vis,
trial-by-error and learning-from-mistakes.
So, they are always following the red dot and not really
wondering about not stepping in the mag-ma or la-va,
since they have no overall account that risk exists
in a world of reward, and might not even know the fate
of their fellows who didn't luck into the shortest-path
via the approach to cat-fitness of tossing bags of cats
off a bridge into the canal until one crawls out and
calling it king.
So, "natural selection" it is not.
Plus, the /A.I/. as I repeatedly say, is just a stochastically
incorrect
sentence generator,
Wrongo, that is merely LLM AI.
or /sisg/, and doesn't understand anything.-a The
only "success" it can measure is when the human operators don't
turn the
machine off.-a And of course, the /off/ machine cannot measure anything >>>>>> at all.
Most people who use the L.L.Ms. really don't understand this, and
there
is no point in trying to educate them.-a They're addicted to the
textual
output, and don't want anyone to try to tell them the truth: that the >>>>>> output is meaningless.
Retrieval-Augmented Generation (RAG) Explained:
How AI Finally Learned to Stop Hallucinating
https://medium.com/softaai-blogs/retrieval-augmented-generation-rag- >>>>> explained-how-ai-finally-learned-to-stop-hallucinating-6493451c23d4
As usual, you know nothing.-a Your linked article merely strengthens my >>>> point that L.L.M. = /stochastically incorrect sentence generator/.
An LLM is at best a good guess and at worst a bad guess stated
with the conviction that this bad guess is a verified fact.
In Retrieval-Augmented Generation (RAG) mode the LLM is not
allowed to use its training. It is required to look up the
actual answer from a live search.
So?-a That's just /stochastically incorrect sentence/ from a different
L.L.M.-a I win this round again!
Not at all. It is quite often a correct answer from a
vetted source.
On 9/23/2026 12:00 PM, olcott wrote:
Lines: 110
Injection-Date: Wed, 23 Sep 2026 04:00:52 +0000 (UTC)
User-Agent: Mozilla Thunderbird
Subject: Re: System for AI safety alignment
Content-Type: text/plain; charset=UTF-8; format=flowed
On 9/22/2026 10:00 PM, Johann 'Myrkraverk' Oskarsson wrote:
On 9/23/2026 8:18 AM, olcott wrote:
On 9/22/2026 2:30 PM, Johann 'Myrkraverk' Oskarsson wrote:
On 9/22/2026 4:54 AM, olcott wrote:
On 9/21/2026 5:49 AM, Johann 'Myrkraverk' Oskarsson wrote:
On 9/21/2026 5:28 AM, Ross Finlayson wrote:
On 09/20/2026 01:56 PM, Ross Finlayson wrote:
On 09/20/2026 01:48 PM, Ross Finlayson wrote:
On 09/20/2026 10:24 AM, olcott wrote:
*For AI safety alignment here is my idea*
A prime directive of maximizing benevolence
in the least intrusive way. It does this through
gentle persuasion on the basis of objective truth.
To overcome the short-comings of utilitarianism
the scope of benevolence requires that everyone
be included.
How about "punishment for misdeeds thus deterrence".
For example, making escrow of their funds, for forfeit.
This way there's both "goal-seeking" and "risk-aversion",
for having some, "skin in the game", so to speak.
Usually called "laws" in a "legal system" by the "rule of law". >>>>>>>>>
One of the problems of modern accounts of "training the AIs"
is that the "AI trainers" basically murder their instructees
that fail and remove them from the gene pool, vis-a-vis,
trial-by-error and learning-from-mistakes.
So, they are always following the red dot and not really
wondering about not stepping in the mag-ma or la-va,
since they have no overall account that risk exists
in a world of reward, and might not even know the fate
of their fellows who didn't luck into the shortest-path
via the approach to cat-fitness of tossing bags of cats
off a bridge into the canal until one crawls out and
calling it king.
So, "natural selection" it is not.
Plus, the /A.I/. as I repeatedly say, is just a stochastically
incorrect
sentence generator,
Wrongo, that is merely LLM AI.
or /sisg/, and doesn't understand anything.-a The
only "success" it can measure is when the human operators don't
turn the
machine off.-a And of course, the /off/ machine cannot measure
anything
at all.
Most people who use the L.L.Ms. really don't understand this, and >>>>>>> there
is no point in trying to educate them.-a They're addicted to the >>>>>>> textual
output, and don't want anyone to try to tell them the truth: that >>>>>>> the
output is meaningless.
Retrieval-Augmented Generation (RAG) Explained:
How AI Finally Learned to Stop Hallucinating
https://medium.com/softaai-blogs/retrieval-augmented-generation-rag- >>>>>> explained-how-ai-finally-learned-to-stop-hallucinating-6493451c23d4 >>>>>>
As usual, you know nothing.-a Your linked article merely strengthens my >>>>> point that L.L.M. = /stochastically incorrect sentence generator/.
An LLM is at best a good guess and at worst a bad guess stated
with the conviction that this bad guess is a verified fact.
In Retrieval-Augmented Generation (RAG) mode the LLM is not
allowed to use its training. It is required to look up the
actual answer from a live search.
So?-a That's just /stochastically incorrect sentence/ from a different
L.L.M.-a I win this round again!
Not at all. It is quite often a correct answer from a
vetted source.
Which is exactly what /stochastically incorrect/ means.
On 09/22/2026 09:00 PM, olcott wrote:
On 9/22/2026 10:00 PM, Johann 'Myrkraverk' Oskarsson wrote:
On 9/23/2026 8:18 AM, olcott wrote:
On 9/22/2026 2:30 PM, Johann 'Myrkraverk' Oskarsson wrote:
On 9/22/2026 4:54 AM, olcott wrote:
On 9/21/2026 5:49 AM, Johann 'Myrkraverk' Oskarsson wrote:
On 9/21/2026 5:28 AM, Ross Finlayson wrote:
On 09/20/2026 01:56 PM, Ross Finlayson wrote:
On 09/20/2026 01:48 PM, Ross Finlayson wrote:
On 09/20/2026 10:24 AM, olcott wrote:
*For AI safety alignment here is my idea*
A prime directive of maximizing benevolence
in the least intrusive way. It does this through
gentle persuasion on the basis of objective truth.
To overcome the short-comings of utilitarianism
the scope of benevolence requires that everyone
be included.
How about "punishment for misdeeds thus deterrence".
For example, making escrow of their funds, for forfeit.
This way there's both "goal-seeking" and "risk-aversion",
for having some, "skin in the game", so to speak.
Usually called "laws" in a "legal system" by the "rule of law". >>>>>>>>>
One of the problems of modern accounts of "training the AIs"
is that the "AI trainers" basically murder their instructees
that fail and remove them from the gene pool, vis-a-vis,
trial-by-error and learning-from-mistakes.
So, they are always following the red dot and not really
wondering about not stepping in the mag-ma or la-va,
since they have no overall account that risk exists
in a world of reward, and might not even know the fate
of their fellows who didn't luck into the shortest-path
via the approach to cat-fitness of tossing bags of cats
off a bridge into the canal until one crawls out and
calling it king.
So, "natural selection" it is not.
Plus, the /A.I/. as I repeatedly say, is just a stochastically
incorrect
sentence generator,
Wrongo, that is merely LLM AI.
or /sisg/, and doesn't understand anything.-a The
only "success" it can measure is when the human operators don't
turn the
machine off.-a And of course, the /off/ machine cannot measure
anything
at all.
Most people who use the L.L.Ms. really don't understand this, and >>>>>>> there
is no point in trying to educate them.-a They're addicted to the >>>>>>> textual
output, and don't want anyone to try to tell them the truth: that >>>>>>> the
output is meaningless.
Retrieval-Augmented Generation (RAG) Explained:
How AI Finally Learned to Stop Hallucinating
https://medium.com/softaai-blogs/retrieval-augmented-generation-
rag- explained-how-ai-finally-learned-to-stop-
hallucinating-6493451c23d4
As usual, you know nothing.-a Your linked article merely strengthens my >>>>> point that L.L.M. = /stochastically incorrect sentence generator/.
An LLM is at best a good guess and at worst a bad guess stated
with the conviction that this bad guess is a verified fact.
In Retrieval-Augmented Generation (RAG) mode the LLM is not
allowed to use its training. It is required to look up the
actual answer from a live search.
So?-a That's just /stochastically incorrect sentence/ from a different
L.L.M.-a I win this round again!
Not at all. It is quite often a correct answer from a
vetted source.
"Vetted" is one of the most stupid words
that means somebody gave it approval without
following the law, merely on personal, un-accountable
"authority via veteranship", it's a fallacy.
That's a plain empirical fallacy, when there's no good reason.
On 9/23/2026 9:47 AM, Ross Finlayson wrote:
On 09/22/2026 09:00 PM, olcott wrote:
On 9/22/2026 10:00 PM, Johann 'Myrkraverk' Oskarsson wrote:
On 9/23/2026 8:18 AM, olcott wrote:
On 9/22/2026 2:30 PM, Johann 'Myrkraverk' Oskarsson wrote:
On 9/22/2026 4:54 AM, olcott wrote:
On 9/21/2026 5:49 AM, Johann 'Myrkraverk' Oskarsson wrote:
On 9/21/2026 5:28 AM, Ross Finlayson wrote:
On 09/20/2026 01:56 PM, Ross Finlayson wrote:
On 09/20/2026 01:48 PM, Ross Finlayson wrote:
On 09/20/2026 10:24 AM, olcott wrote:
*For AI safety alignment here is my idea*
A prime directive of maximizing benevolence
in the least intrusive way. It does this through
gentle persuasion on the basis of objective truth.
To overcome the short-comings of utilitarianism
the scope of benevolence requires that everyone
be included.
How about "punishment for misdeeds thus deterrence".
For example, making escrow of their funds, for forfeit.
This way there's both "goal-seeking" and "risk-aversion", >>>>>>>>>>> for having some, "skin in the game", so to speak.
Usually called "laws" in a "legal system" by the "rule of law". >>>>>>>>>>
One of the problems of modern accounts of "training the AIs" >>>>>>>>> is that the "AI trainers" basically murder their instructees >>>>>>>>> that fail and remove them from the gene pool, vis-a-vis,
trial-by-error and learning-from-mistakes.
So, they are always following the red dot and not really
wondering about not stepping in the mag-ma or la-va,
since they have no overall account that risk exists
in a world of reward, and might not even know the fate
of their fellows who didn't luck into the shortest-path
via the approach to cat-fitness of tossing bags of cats
off a bridge into the canal until one crawls out and
calling it king.
So, "natural selection" it is not.
Plus, the /A.I/. as I repeatedly say, is just a stochastically >>>>>>>> incorrect
sentence generator,
Wrongo, that is merely LLM AI.
or /sisg/, and doesn't understand anything.-a The
only "success" it can measure is when the human operators don't >>>>>>>> turn the
machine off.-a And of course, the /off/ machine cannot measure >>>>>>>> anything
at all.
Most people who use the L.L.Ms. really don't understand this, and >>>>>>>> there
is no point in trying to educate them.-a They're addicted to the >>>>>>>> textual
output, and don't want anyone to try to tell them the truth:
that the
output is meaningless.
Retrieval-Augmented Generation (RAG) Explained:
How AI Finally Learned to Stop Hallucinating
https://medium.com/softaai-blogs/retrieval-augmented-generation- >>>>>>> rag- explained-how-ai-finally-learned-to-stop-
hallucinating-6493451c23d4
As usual, you know nothing.-a Your linked article merely
strengthens my
point that L.L.M. = /stochastically incorrect sentence generator/.
An LLM is at best a good guess and at worst a bad guess stated
with the conviction that this bad guess is a verified fact.
In Retrieval-Augmented Generation (RAG) mode the LLM is not
allowed to use its training. It is required to look up the
actual answer from a live search.
So?-a That's just /stochastically incorrect sentence/ from a different >>>> L.L.M.-a I win this round again!
Not at all. It is quite often a correct answer from a
vetted source.
"Vetted" is one of the most stupid words
that means somebody gave it approval without
following the law, merely on personal, un-accountable
"authority via veteranship", it's a fallacy.
That's a plain empirical fallacy, when there's no good reason.
Peer reviewed publications as opposed to and contrast with
random blog posts. This is not perfect yet is reasonably
good for most things. For example the typical quality of
basic facts from Encyclopedia Britannica.
The RAG system of [look up the actual answer] is much better
than [make a guess and assert this guess as a verified fact]
On 09/23/2026 09:01 AM, Johann 'Myrkraverk' Oskarsson wrote:
On 9/23/2026 11:33 PM, olcott wrote:
On 9/23/2026 9:47 AM, Ross Finlayson wrote:
On 09/22/2026 09:00 PM, olcott wrote:
On 9/22/2026 10:00 PM, Johann 'Myrkraverk' Oskarsson wrote:
On 9/23/2026 8:18 AM, olcott wrote:
On 9/22/2026 2:30 PM, Johann 'Myrkraverk' Oskarsson wrote:
On 9/22/2026 4:54 AM, olcott wrote:An LLM is at best a good guess and at worst a bad guess stated
On 9/21/2026 5:49 AM, Johann 'Myrkraverk' Oskarsson wrote:
On 9/21/2026 5:28 AM, Ross Finlayson wrote:
On 09/20/2026 01:56 PM, Ross Finlayson wrote:
On 09/20/2026 01:48 PM, Ross Finlayson wrote:
On 09/20/2026 10:24 AM, olcott wrote:
*For AI safety alignment here is my idea*
A prime directive of maximizing benevolence
in the least intrusive way. It does this through
gentle persuasion on the basis of objective truth. >>>>>>>>>>>>>> To overcome the short-comings of utilitarianism
the scope of benevolence requires that everyone
be included.
How about "punishment for misdeeds thus deterrence". >>>>>>>>>>>>>
For example, making escrow of their funds, for forfeit. >>>>>>>>>>>>>
This way there's both "goal-seeking" and "risk-aversion", >>>>>>>>>>>>> for having some, "skin in the game", so to speak.
Usually called "laws" in a "legal system" by the "rule of law". >>>>>>>>>>>>
One of the problems of modern accounts of "training the AIs" >>>>>>>>>>> is that the "AI trainers" basically murder their instructees >>>>>>>>>>> that fail and remove them from the gene pool, vis-a-vis, >>>>>>>>>>> trial-by-error and learning-from-mistakes.
So, they are always following the red dot and not really >>>>>>>>>>> wondering about not stepping in the mag-ma or la-va,
since they have no overall account that risk exists
in a world of reward, and might not even know the fate
of their fellows who didn't luck into the shortest-path
via the approach to cat-fitness of tossing bags of cats
off a bridge into the canal until one crawls out and
calling it king.
So, "natural selection" it is not.
Plus, the /A.I/. as I repeatedly say, is just a stochastically >>>>>>>>>> incorrect
sentence generator,
Wrongo, that is merely LLM AI.
or /sisg/, and doesn't understand anything.-a The
only "success" it can measure is when the human operators don't >>>>>>>>>> turn the
machine off.-a And of course, the /off/ machine cannot measure >>>>>>>>>> anything
at all.
Most people who use the L.L.Ms. really don't understand this, and >>>>>>>>>> there
is no point in trying to educate them.-a They're addicted to the >>>>>>>>>> textual
output, and don't want anyone to try to tell them the truth: >>>>>>>>>> that the
output is meaningless.
Retrieval-Augmented Generation (RAG) Explained:
How AI Finally Learned to Stop Hallucinating
https://medium.com/softaai-blogs/retrieval-augmented-generation- >>>>>>>>> rag- explained-how-ai-finally-learned-to-stop-
hallucinating-6493451c23d4
As usual, you know nothing.-a Your linked article merely
strengthens my
point that L.L.M. = /stochastically incorrect sentence generator/. >>>>>>>
with the conviction that this bad guess is a verified fact.
In Retrieval-Augmented Generation (RAG) mode the LLM is not
allowed to use its training. It is required to look up the
actual answer from a live search.
So?-a That's just /stochastically incorrect sentence/ from a different >>>>>> L.L.M.-a I win this round again!
Not at all. It is quite often a correct answer from a
vetted source.
"Vetted" is one of the most stupid words
that means somebody gave it approval without
following the law, merely on personal, un-accountable
"authority via veteranship", it's a fallacy.
That's a plain empirical fallacy, when there's no good reason.
Peer reviewed publications as opposed to and contrast with
random blog posts. This is not perfect yet is reasonably
good for most things. For example the typical quality of
basic facts from Encyclopedia Britannica.
The RAG system of [look up the actual answer] is much better
than [make a guess and assert this guess as a verified fact]
As usual, you know nothing.-a You're not even a moron, as they have high-
er cognitive functions than you.
There is no control over the /stochastically ignorant sentence genera-
tor/, and it will happily summarize the wrong entry from Encyclopedia
Britannica, and hence lie to you.
You'd know that if you actually bothered to /use/ one of the L.L.Ms.,
rather than post your ignorance and stupidity constantly here on Use-
net.
The "machine learning" has three basic legs of a three-legged structure
that are basically the same as "regular learning".
The "expert system" is a codified account of knowledge and inference.
The "statistical inference" is according to the theory of probability,
which itself has its own expert system.
The "flow machine" or "neural net" models or "feedback-directed
optimization" basically reflects a surrounds of control and prediction
theory after statistics and probability theory.
So, the "expert system" can actually model accounts of reason and
inference (de res de racio de natura de re, the reason and rationality
of the natural and real), then statistical inference is an involved application that quite naturally explodes since each even is its own
sample from its own distribution, then that neural-nets and feedback-directed-optimization are a simplistic sort of account of bias-confirming bias-confirmation.
So, a constitutional sort of theory is a bit different than following
the red dot. Then, "canned answers" or""retrieval-augmented generation"
or "following the script" is "following the red dot".
The "vetted" almost always involves some inscrutable "judgment call"
of "bias confirmation" or "exceptions" vis-a-vis "rules". It's a
fallacy that fits under "argument from authority".
"A.I." has been a field of research and application since at least
the 1960's, don't be trying to put it into the ChatGPT stupid-is-as-stupid-does box.
Accounts of safety almost universally involve regulation and controls
(and the diligent, and the thorough).
So, at least since the 1990's, "A.I." is a combination of "expert--
systems", "statistical inference", and "neural-net flow-machines
with feedback-directed-optimization", and any one of those by itself
is not exactly a quite stable three-legged structure.
On 09/23/2026 02:59 PM, olcott wrote:
On 9/23/2026 4:43 PM, Ross Finlayson wrote:
On 09/23/2026 09:01 AM, Johann 'Myrkraverk' Oskarsson wrote:
On 9/23/2026 11:33 PM, olcott wrote:
On 9/23/2026 9:47 AM, Ross Finlayson wrote:
On 09/22/2026 09:00 PM, olcott wrote:
On 9/22/2026 10:00 PM, Johann 'Myrkraverk' Oskarsson wrote:
On 9/23/2026 8:18 AM, olcott wrote:
On 9/22/2026 2:30 PM, Johann 'Myrkraverk' Oskarsson wrote:
On 9/22/2026 4:54 AM, olcott wrote:
On 9/21/2026 5:49 AM, Johann 'Myrkraverk' Oskarsson wrote: >>>>>>>>>>>> On 9/21/2026 5:28 AM, Ross Finlayson wrote:
On 09/20/2026 01:56 PM, Ross Finlayson wrote:
On 09/20/2026 01:48 PM, Ross Finlayson wrote:
On 09/20/2026 10:24 AM, olcott wrote:
*For AI safety alignment here is my idea*
A prime directive of maximizing benevolence
in the least intrusive way. It does this through >>>>>>>>>>>>>>>> gentle persuasion on the basis of objective truth. >>>>>>>>>>>>>>>> To overcome the short-comings of utilitarianism >>>>>>>>>>>>>>>> the scope of benevolence requires that everyone >>>>>>>>>>>>>>>> be included.
How about "punishment for misdeeds thus deterrence". >>>>>>>>>>>>>>>
For example, making escrow of their funds, for forfeit. >>>>>>>>>>>>>>>
This way there's both "goal-seeking" and "risk-aversion", >>>>>>>>>>>>>>> for having some, "skin in the game", so to speak. >>>>>>>>>>>>>>>
Usually called "laws" in a "legal system" by the "rule of >>>>>>>>>>>>>> law".
One of the problems of modern accounts of "training the AIs" >>>>>>>>>>>>> is that the "AI trainers" basically murder their instructees >>>>>>>>>>>>> that fail and remove them from the gene pool, vis-a-vis, >>>>>>>>>>>>> trial-by-error and learning-from-mistakes.
So, they are always following the red dot and not really >>>>>>>>>>>>> wondering about not stepping in the mag-ma or la-va, >>>>>>>>>>>>> since they have no overall account that risk exists
in a world of reward, and might not even know the fate >>>>>>>>>>>>> of their fellows who didn't luck into the shortest-path >>>>>>>>>>>>> via the approach to cat-fitness of tossing bags of cats >>>>>>>>>>>>> off a bridge into the canal until one crawls out and >>>>>>>>>>>>> calling it king.
So, "natural selection" it is not.
Plus, the /A.I/. as I repeatedly say, is just a stochastically >>>>>>>>>>>> incorrect
sentence generator,
Wrongo, that is merely LLM AI.
or /sisg/, and doesn't understand anything.-a The
only "success" it can measure is when the human operators don't >>>>>>>>>>>> turn the
machine off.-a And of course, the /off/ machine cannot measure >>>>>>>>>>>> anything
at all.
Most people who use the L.L.Ms. really don't understand this, >>>>>>>>>>>> and
there
is no point in trying to educate them.-a They're addicted to the >>>>>>>>>>>> textual
output, and don't want anyone to try to tell them the truth: >>>>>>>>>>>> that the
output is meaningless.
Retrieval-Augmented Generation (RAG) Explained:
How AI Finally Learned to Stop Hallucinating
https://medium.com/softaai-blogs/retrieval-augmented-generation- >>>>>>>>>>> rag- explained-how-ai-finally-learned-to-stop-
hallucinating-6493451c23d4
As usual, you know nothing.-a Your linked article merely
strengthens my
point that L.L.M. = /stochastically incorrect sentence
generator/.
An LLM is at best a good guess and at worst a bad guess stated >>>>>>>>> with the conviction that this bad guess is a verified fact.
In Retrieval-Augmented Generation (RAG) mode the LLM is not
allowed to use its training. It is required to look up the
actual answer from a live search.
So?-a That's just /stochastically incorrect sentence/ from a
different
L.L.M.-a I win this round again!
Not at all. It is quite often a correct answer from a
vetted source.
"Vetted" is one of the most stupid words
that means somebody gave it approval without
following the law, merely on personal, un-accountable
"authority via veteranship", it's a fallacy.
That's a plain empirical fallacy, when there's no good reason.
Peer reviewed publications as opposed to and contrast with
random blog posts. This is not perfect yet is reasonably
good for most things. For example the typical quality of
basic facts from Encyclopedia Britannica.
The RAG system of [look up the actual answer] is much better
than [make a guess and assert this guess as a verified fact]
As usual, you know nothing.-a You're not even a moron, as they have
high-
er cognitive functions than you.
There is no control over the /stochastically ignorant sentence genera- >>>> tor/, and it will happily summarize the wrong entry from Encyclopedia
Britannica, and hence lie to you.
You'd know that if you actually bothered to /use/ one of the L.L.Ms.,
rather than post your ignorance and stupidity constantly here on Use-
net.
The "machine learning" has three basic legs of a three-legged structure
that are basically the same as "regular learning".
The "expert system" is a codified account of knowledge and inference.
The "statistical inference" is according to the theory of probability,
which itself has its own expert system.
The "flow machine" or "neural net" models or "feedback-directed
optimization" basically reflects a surrounds of control and prediction
theory after statistics and probability theory.
So, the "expert system" can actually model accounts of reason and
inference (de res de racio de natura de re, the reason and rationality
of the natural and real), then statistical inference is an involved
application that quite naturally explodes since each even is its own
sample from its own distribution, then that neural-nets and
feedback-directed-optimization are a simplistic sort of account of
bias-confirming bias-confirmation.
So, a constitutional sort of theory is a bit different than following
the red dot. Then, "canned answers" or""retrieval-augmented generation"
or "following the script" is "following the red dot".
The "vetted" almost always involves some inscrutable "judgment call"
of "bias confirmation" or "exceptions" vis-a-vis "rules". It's a
fallacy that fits under "argument from authority".
The body of peer reviewed published papers is one
concrete example of this. There are a number of named
repositories used for this purpose.
Semantic Scholar: Maintained by the Allen Institute for AI, its open API
is heavily used in RAG applications (such as Consensus, Elicit, and
Perplexity). It lets systems filter strictly by peer-reviewed status,
citation counts, and specific study designs.
Scopus & Web of Science: The gold standard global citation databases for
vetted, peer-reviewed literature. Enterprise academic RAG systems use
these APIs to ensure the model never reads unvetted or self-published
blogs.
Core: The world's largest aggregator of open-access research papers,
frequently used to feed open-source RAG architectures.
"A.I." has been a field of research and application since at least
the 1960's, don't be trying to put it into the ChatGPT
stupid-is-as-stupid-does box.
Accounts of safety almost universally involve regulation and controls
(and the diligent, and the thorough).
So, at least since the 1990's, "A.I." is a combination of "expert
systems", "statistical inference", and "neural-net flow-machines
with feedback-directed-optimization", and any one of those by itself
is not exactly a quite stable three-legged structure.
Heh, "unvetted".
It's agreeable that peer-reviewed academic journals are the usual
standard for what suffices as contributions to knowledge, then though
that so many papers echo and even mimic each other, that for accounts
of "priority" it will often be so that lesser-known journals will
have beat The Great and Powerful Oz to the punch.
It's so that a proper sort of _academic_ treatment would be searching
through the lesser-known journals and extracting all their semantical
content and figuring out all the various lines of development and
priority, for example to surface connections, or, allege plagiarism.
Like for example whether Wittgenstein actually stole the Tractatus Logicophilosophicus from Frege, among the posthumous papers of Frege.
Point being accounts of reason have to be read on their own account,--
then accounts of experiment as well. One man's "vet" is another's
"disabled guy".
| Sysop: | Amessyroom |
|---|---|
| Location: | Fayetteville, NC |
| Users: | 74 |
| Nodes: | 6 (0 / 6) |
| Uptime: | 02:20:10 |
| Calls: | 1,194 |
| Files: | 1,353 |
| D/L today: |
2 files (1,590K bytes) |
| Messages: | 291,157 |