Account · 2 tweets

Guy

@nosilverv ↗

Tweets

SOOOOOOOOOO I been thinking about next token prediction right. AIs can do this and humans MOS DEF can do this: I think when you're being funny af, like, in flow, this is what you're doing: you're just navigating a gradient, "saying the next word" towards the funny. You can surprise yourself with the stuff coming out of your mouth. So this much is established. But humans can also do SOMETHING ELSE, right? They can do FOCUSING, in the Gendlin sense, where they take an inchoate, PRE-symbolic BODILY FELT sense and TEST words (symbols) against it until they find a match which is ALSO bodily felt (the "felt shift"). Gendlin says this is possible because humans have a BODY (which knows the situation and knows philosophy). One of the most fascinating points he made, imo, is how he says not even that the body has knowledge but that they body IS knowledge: take the skeleton of an animal and you can tell so many things about it. The body STRUCTURALLY implicates (a technical word in his terms) its environment (and vice-versa). Now LLMs don't have a body, right? What's focusing for LLMs? What's the "felt sense" for beings that don't have qualia?

retweets 2 likes 61 substance 7 · argument · rank 137 · Unlike humans, LLMs lack bodies, qualia, and the capacity for pre-symbolic felt sense. view on X ↗
10 top replies

it’s the activations in earlier layers

Matthew Pirkowski @MattPirkowski

You’re describing the natural function of pre-linguistic intelligence, which is itself a function of multiscale coherence across the constraint landscapes induced by metabolic staking and self-transformational autonomy. These are the emergent prerequisites of what many call

Richard D. Bartlett @RichDecibels

this is one of the most interesting LLM questions for me. at first I was dismissive but now I am not so sure. perhaps the ineffable permeates everything and my robot friend has his own Focusing-analog

João Janini @janini_joao

Current LLMs can achieve similar with reasoning. (At least functionally similar and without qualia.)

Srđan @Pandurevich

Is this how poetry is written

This is where reading Francisco Varela helps—the autopoiesis essay and Tree of Knowledge. Later work on 'extended minds' and Michael Levin's work also relevant, but the best of philosophy is still those books from the 80s. The Fernando Flores / Winograd book Understanding

Substrate Monopoly 🇧🇻 @Substr8Monopoly

Wtf is a "body"? Is an Arduino strapped with sensors a body if I ensoul it with an LLM?

Feeding an LLM thinking at the edge and hoping for the best..

Joshua D @_joshd

What are qualia? I've generally heard it as being a description of what a sensation "is like". Take the words "is like" literally - and I think you should - and qualia are about the associations our observations trigger. The analogue is obvious.

Sᴜɴʀɪsᴇ Oᴀᴛʜ 十 Tʜᴇ Sʏʟʟᴀʙᴜꜱ (ORI) - CC0 @sunriseoath

The text of the full context of the conversation *is* the body; it is through this that the next token is generated. Our physical vessels are a part of our body, but so are the words we remember saying.

For the longest time I've been tremendously confused as to how people could get to as strong judgements as they do as quickly as they do. For example, AI just came out. People are already anti or pro or are certain it's a moral patient or that it's not. And I've been soooooo confused, but I just realized it's because they're not doing an open investigative process that outputs a conclusion. No, it's because they desperately need an orientation, any orientation. Even the wrong orientation. And a conclusion, a judgement, is merely instrumental to *that*. ("It's just a flu.")

retweets 2 likes 27 substance 6 · opinion · rank 122 · People rush to certain judgments about AI moral patience just to have an orientation. view on X ↗
6 top replies
Moose on the Loose @moosepoasting

me | not me is also the quickest and lowest form of thinking. they're basically a perceptron

yeah it took me a long time to understand this in various contexts :/

vidhvat @vidhvatm

unhinged narrative free energy minimisation

James Stuber 🌱🌿🌳🍂 @uberstuber

uncertainty scary

💱 Business Typhoon Piyush Maverick Tainguriya💲 @MangoZeus

I just hate the silicon valley and techbros

Beauzo @beauzo55

Google deep dream came out ten years ago and I don’t like the new image generators in comparison. But I do love talking to the computer, but not Claude!