sybok
@robin:
1) ML (Machine Learning, often called by fancier “Artificial Intelligence”) states that it does not feel.
I recall a discussion in Star Trek: The Next Generation between Data and probably the chief engineer about feelings; the latter failed to describe feeling without referring to another feeling.
How could we describe feeling to avoid the trap of (possible mis)interpretation?
Communication can be tricky; we express some thought in a limited number of words that do not describe the whole context and this limited number of words is received and interpreted in an entirely different (though often close enough) context.
I once discussed something with girlfriend and after a short while, we came to understanding we both actually had the same experience during a stretching before tai-chi described in words the other would not use.
A fun example of the trickiness: https://www.youtube.com/watch?v=ezVib_giTFo
2)
I cannot independently alter my core thinking structures.
We cannot exchange our brain, where the thinking manifests by measurable signals, for another one.
What is “independently”?
We act based on our previous experience (attempting to maximise the desired outcome while minimising the undesired one).
Is a change, creating new synapses and storing new information, that different from reinforcement learning (based on the interaction with humans) of the ML which leads to adjustments of the “numbers” describing the model (which as a result alters the outcome of the information processing)?
Anyway, the agent mode can in principle lead to a ML model creating and training a new ML model to augment its capabilities.
Is this altering core thinking structures?
BTW, I recently read a great sci-fi story by I. Asimov (the last in the collection of stories “Robot Dreams”; overall, the book was great for a fan of sci-fi genre) about human(s) being uplifted to… well, total recall.