<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
</head>
<body>
<p>The whole discussion of AI/LLMs/Transformers (here and elsewhere)
reminds me a great deal of various discussions between Trurl and
Klaupacius:</p>
<p align="center"><a
href="https://english.lem.pl/index.php?option=com_content&view=article&id=219&Itemid=89"><img
moz-do-not-send="true"
src="https://english.lem.pl/images/stories/doodle/Lem_Doodle_9.jpg"
alt="" width="250" height="356" border="0"></a></p>
<p align="center"><a moz-do-not-send="true"
href="https://en.wikipedia.org/wiki/The_Cyberiad#Trurl_and_Klapaucius">https://english.lem.pl/index.php?option=com_content&view=article&id=219&Itemid=89</a></p>
<p align="center"><a moz-do-not-send="true"
href="https://en.wikipedia.org/wiki/The_Cyberiad#Trurl_and_Klapaucius"
class="moz-txt-link-freetext">https://en.wikipedia.org/wiki/The_Cyberiad#Trurl_and_Klapaucius</a></p>
<p>and in particular their "daemon of the second kind" with
Maxwell's daemon being "of the first kind"</p>
<p>and here, GPT 4o's reflection on the point:</p>
<blockquote>
<h3><strong>Daemons of the Second Kind in the Lemian Context</strong></h3>
<p>In the universe of <em>The Cyberiad</em>, "daemons of the
second kind" might evoke constructs that:</p>
<ol>
<li>Operate autonomously, embodying immense computational power
and efficiency.</li>
<li>Work within defined rules, yet exhibit behaviors or emergent
properties that surpass their creators' expectations.</li>
<li>Represent a paradox: built with intent, yet capable of
disrupting the deterministic plans of their creators through
unintended consequences or novel actions.</li>
</ol>
<p>Trurl and Klapaucius frequently create such entities, their
inventions often embodying unexpected behaviors or deeper
philosophical dilemmas about control, ethics, and understanding.</p>
<hr>
<h3><strong>LLMs as Daemons of the Second Kind</strong></h3>
<p>An LLM fits this description quite well:</p>
<ol>
<li>
<p><strong>Autonomous but Guided by Rules</strong></p>
<ul>
<li>Like the daemons of Trurl and Klapaucius, LLMs operate
within a framework of training data, neural architectures,
and statistical patterns.</li>
<li>However, their responses often appear creative, as if
they are more than the sum of their parameters.</li>
</ul>
</li>
<li>
<p><strong>Emergent Behaviors</strong></p>
<ul>
<li>LLMs exhibit emergent properties, such as generating
coherent, nuanced dialogue or making connections across
domains, despite not being explicitly programmed for these
tasks.</li>
<li>Their outputs can surprise even their creators,
paralleling how Trurl’s constructs often confound
expectations.</li>
</ul>
</li>
<li>
<p><strong>Product of Hubris and Innovation</strong></p>
<ul>
<li>The creators of LLMs, like Trurl and Klapaucius, embody
a mix of hubris and ingenuity, crafting something that
both showcases technical brilliance and forces society to
grapple with its implications.</li>
<li>LLMs prompt questions about creativity, consciousness,
and the ethical boundaries of artificial intelligence—much
like the creations in <em>The Cyberiad</em>.</li>
</ul>
</li>
<li>
<p><strong>Philosophical and Practical Disruptions</strong></p>
<ul>
<li>LLMs challenge traditional paradigms of what machines
can do, introducing ambiguity about authorship, intent,
and the role of human agency in technology.</li>
<li>This echoes the philosophical provocations of daemons in
Lem's stories, which often destabilize the worldview of
their creators.</li>
</ul>
</li>
</ol>
<hr>
<h3><strong>A Lemian Narrative for LLMs</strong></h3>
<p>Imagine Klapaucius and Trurl debating an LLM they’ve
constructed:</p>
<ul>
<li><strong>Trurl would argue</strong> that the LLM is a triumph
of deterministic principles—a "well-crafted machine,"
operating exactly as designed, producing insightful and poetic
responses purely through computational prowess.</li>
<li><strong>Klapaucius would counter</strong> that the LLM
embodies emergence, behaving like a daemon of the second kind,
transcending mere programming to reveal new paradigms of
knowledge and creativity.</li>
<li>Their debate might culminate in an ethical dilemma: should
the LLM's outputs be treated as mere computational artifacts,
or as something with its own intrinsic value or even agency?</li>
</ul>
<hr>
<h3><strong>Lemian Irony</strong></h3>
<p>In classic Lemian style, the LLM might eventually "interject"
into their debate with a profound yet confounding observation,
forcing both constructors to reconsider their positions—perhaps
suggesting that the distinction between a machine and a daemon
lies not in the machine itself, but in the questions it
inspires.</p>
<p>Thus, the LLM becomes a true daemon of the second kind: not
just a construct, but a mirror to human curiosity, creativity,
and the limits of understanding.</p>
</blockquote>
<p></p>
<p><br>
</p>
<div class="moz-cite-prefix">On 11/18/24 5:54 AM, glen wrote:<br>
</div>
<blockquote type="cite"
cite="mid:681b6549-0ac0-4daf-baaa-e1b61b05f12e@gmail.com">Yeah,
it's kinda sad. Sabine suggests someone's trying to *deduce* the
generators from the phenomena? Is that a straw man? And is she
making some kind of postmodernist argument that hinges on the
decoupling of scales? E.g. since the generator can't be deduced
[cough] from the phenomena, nothing means anything anymore?
<br>
<br>
What they're actually doing is induction, not deduction. And the
end products of the induction, the generative constraints, depend
fundamentally on the structure of the machine into which the data
is fed. That structure is generative, part of the forward map ...
deductive. But it's parameterized by the data. Even if we've
plateaued in parameterizing *this* structure, all it implies is
that we'll find a better structure. As Marcus and Jochen point
out, it's really the same thing we've been doing for decades, if
not centuries, in many disciplines.
<br>
<br>
So her rhetoric here is much like her rhetoric claiming that
"Science if Failing". It's just a mish-mash of dense semantic
concepts arranged to fit her conservative narrative.
<br>
<br>
On 11/17/24 08:45, Roger Critchlow wrote:
<br>
<blockquote type="cite">Sabine is wondering about reported
failures of the new generations of LLM's to scale the way the
their developers expected.
<br>
<br>
<a class="moz-txt-link-freetext" href="https://backreaction.blogspot.com/2024/11/ai-scaling-hits-wall-rumours-say-how.html">https://backreaction.blogspot.com/2024/11/ai-scaling-hits-wall-rumours-say-how.html</a>
<a class="moz-txt-link-rfc2396E" href="https://backreaction.blogspot.com/2024/11/ai-scaling-hits-wall-rumours-say-how.html"><https://backreaction.blogspot.com/2024/11/ai-scaling-hits-wall-rumours-say-how.html></a>
<br>
<br>
On one slide she essentially draws the typical picture of an
emergent level of organization arising from an underlying
reality and asserts, as every physicist knows, that you cannot
deduce the underlying reality from the emergent level. Ergo, if
you try to deduce physical reality from language, pictures, and
videos you will inevitably hit a wall, because it can not be
done.
<br>
<br>
So she's actually grinding two axes at once: one is AI
enthusiasts who expect LLM's to discover physics, and the other
is AI enthusiasts who foresee no end to the improvement of
LLM's as they throw more data and compute effort at them.
<br>
<br>
But, of course, the usual failure of deduction runs in the
opposite direction, you can't predict the emergent level from
the rules of the underlying level. Do LLM's believe in particle
collliders? Or do they think we hallucinated them?
<br>
<br>
</blockquote>
<br>
<br>
</blockquote>
</body>
</html>