[FRIAM] deducing underlying realities from emergent realities

[glen]

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?

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.

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.

On 11/17/24 08:45, Roger Critchlow wrote:
> Sabine is wondering about reported failures of the new generations of LLM's to scale the way the their developers expected.
> 
> 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>
> 
> 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.
> 
> 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.
> 
> 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?
> 


-- 
ꙮ Mɥǝu ǝlǝdɥɐuʇs ɟᴉƃɥʇ' ʇɥǝ ƃɹɐss snɟɟǝɹs˙ ꙮ