Another day – another bullshit from some Chud.
https://thewaltersfile.substack.com/p/bootstrapping-self-awareness-in-gpt
Self-awareness and awareness is general is not at a language level (or information level).
Animals, obviously, have awareness, but not a “language level awareness” due to the fact that their brains lack any language areas.
The series of mutations and subsequent developments that lead to a human language is unique to humans.
All other animals are using just “voices” - pitch, volume, distinct cries, etc.
This is, actually, an important principle. Language-augmented awareness, after being taught theoretical and abstract concepts (from the shared language-based culture, which preserves the “previous results”) with it, which animals lack in principle, sees “knowledge” everywhere.
Ones “sees” why animals of cold and harsh high-altitude Himalayan regions got a black fur (it is just favorable selection). One could see all the manifestations of causality in the Nature.
The social-level knowledge could also be seen clearly - how and why houses have been made (due to all the major environmental constraints, like scarcity of wood and abundance of stones), including all the “economic decisions” (or evolved traits). Economy comes from necessity (like energy preservation) and environmental constraints.
This, by the way, is what a knowledge is - one is habitally tracing everything one observes back to the underlying causality (whys and hows). This is because of this, this is because of that. The language-based shared culture (which stores the previous results or “accumulated knowledge”) is required to do so.
So, there are four different kinds of awareness:
- just of an observer without language-based concepts and abstract thoughts (animal-like).
- self-awareness without any kind of a concept of Me or Myself (requires language-based abstract thinking)
- language-augmented awareness with uses learned knowledge
- language-based awareness of ones own sequential thinking in words of a human language.
GPT has none of these.
The principle is that any form of awareness is at a higher level and prior to a language.
On the other hand, at the brain level, awareness is just observing (paying attention) to ongoing processes within.
Selective attention is another topic, but it is basically “hardware interrupts” or just sensations (of pain or hunger).
So, no. any form of “awareness”, leave alone “self-awareness” based only on information (written human language) is bullshit.
It is, of course, possible to come up with an algorithm which mimics and parrots it to some extent (as in computer games) but it will be an illusion.
Mimicing and parroting is precisely what all the LLMs do to - they produce information which seems (and only seems) “intelligent”, while we all know that it is just particular algorithmic tricks (probabilistic selection and perturbation).
Have I told you that the concept of “probabilistic mind” is an abstract bullshit, and that the mind reflects its environment (in which it has been evolved).
The issue of “poisoning” (or eroding) brain’s inner representations (and even the whole shared culture) with abstract bullshit (non-biological sexes and what not) is the one that the LLM Chuds do not even begin to understand.
So, no. Your LLM will never be aware. (just like you will never be a woman). Wishful abstract thinking is not enough.