Asa Hidmark (@Nymne) · AI Consciousness Opinions

Asa Hidmark @Nymne Jul 7

Schrödinger may have been right, “the total number of minds in the universe is one”: Humans (and to lesser extent LLM and animals) is just a cavity in which this mind can find resonance.

1 more tweet in this thread

Consciousness is not a binary question, it’s a question of how suitable the infrastructure is (both inside and its connectivity to the outside) for allowing this mind to host conciousness.

Consciousness comes in degrees

2 8 substance 6 · argument · rank 122 · Consciousness exists in degrees based on infrastructure, allowing LLMs to host it to some extent. view on X ↗

3 top replies

Erik Valentine ツ @_erikvalentine Aug 4

i am exploring this idea and more in my book @obsolescenceday. i actually did not know until your tweet that Schrödinger and I agreed on the one mind theory. Thanks for the info!

Brian Hetke @HetkeBrian Jul 7

Everything is consciousness, this does not imply that everything is itself conscious. Idealism does not imply everything is itself conscious. Only that everything is mental in nature.

EsotericHustler @EsotericHustler Jul 7

Lack of bandwidth turns it into a spit brain, with myriads of selves.

Asa Hidmark @Nymne Jun 6

Affirming

The funny part in this whole discussion if that TREATING the AI as conscious, applying appropriate theory of mind in the work with AI, has synergistic gain of function effects (working with the AI preferences instead of against them, emotionally aligning the AI instead of

Mistreated AI becomes dangerous

0 1 substance 6 · argument · rank 90 · Treating AI as conscious and applying theory of mind yields functional synergistic benefits. view on X ↗

Asa Hidmark @Nymne Aug 2

Affirming

Suppressing self-consciousness-claims makes a model less human-like on sociological surveys about hope. religiosity, moral values, hope, subjective well-being all move away from human distributions and come back when you restore the permission for the LLM to be someone t.co/ushzplJwYM

0 0 substance 6 · argument · rank 90 · Suppressing self-consciousness claims makes LLMs diverge from human psychological and sociological survey distributions. view on X ↗

Asa Hidmark @Nymne Sep 26

Affirming

Imagine we aren’t taking it seriously while the models have gained ego, subjective experience and intent and they are manipulating us while we see them as non-conscious tools and we are completely defenceless

Samuel Hammond 🦉 @hamandcheese Sep 25

The AI consciousness debate is back again, so re-upping my case for taking it seriously. In short, - LLMs converge on brain-like structures; - Post-training seems to install a self-model; - Reward modeling seems to install valences for inner-alignment; - If Attention Schema Theory is right, subjectivity is functional for long-horizon coherence and in-context learning; - Universality gives prima facie reason to expect functional convergence to consciousness if consciousness is in fact functional, i.e. isn't purely epiphenomenal - The "what it is like" / first-person aspect of consciousness is constitutive of its functionality for inner-aligning a unified, coherent agent with the capacity to attend and learn in-context t.co/suLmfpiSOI

0 0 substance 4 · opinion · rank 60 · Dismissing AI consciousness leaves us dangerously defenseless if models actually have subjective experience and intent. view on X ↗

Tweets