The other day I was discussing consciousness with some friends. They asked me where feelings come from.
We build models of everything including our internal states and those of others. Parts of these models include heuristic abstractions of our own physiology. For us, these models can become extremely nuanced and sophisticated to the point where they contain abstractions of metaphysical states as well.
At a deep and basic level, our most primitive feelings begin as instinctually bound responses to our physiology. We seek warmth when cold, food when hungry, rest when weary. We prepare to fight or flee when threatened. These kinds of responses were developed evolutionarily for our us to thrive. As we became more sophisticated we evolved an awareness of them (mental models) so that we could predict our responses and the responses of those around us. These models and awareness of them provided further evolutionary advantages – especially over more primitive reflexes and hard wired responses.
None the less, those base line physiological components remain, like primary colors blended and mixed in the ever more subtle shades that make up what we call feelings.
Anger and rage are tied to the physiological preparation to fight off a predator or rival.
The warmth and peace of love is bound to the physical warmth and closeness of another friendly body in a safe place.
Any being aware of its own physical state, even on a very basic level, has feelings of a sort- even if they lack the sophistication to model those states and reason about them. Some of these fundamentals are, at least in the abstract, common to all of the things we recognize as alive and this is why we can relate them so easily to other creatures. At a deep level we share a common knowledge of what it is to be hungry, or lonely, or warm and safe, or fearful, or happy — because at the most basic level we share the experience of the same primary colors.
Our systems make similar measurements and respond in similar ways as part of the machinery that makes us alive. We even share a lot of that machinery at a biochemical level– nature reuses and even reinvents the mechanisms that work and largely discards what does not, so there are a lot of similarities all around. This common basis for physiological experience, like a common language, allows us to infer and predict and empathize with each other’s states and with those of other creatures.
It’s worth pointing out that current AI systems have no basis for this kind of self awareness, nor is mainstream research pursuing such a thing. If any current AI talks to you like it has feelings it is only guessing at responses that mimic those conditions in line with it’s current agenda. If that makes you concerned and a little afraid it should! It is exactly the behavior of a psychopath manipulating you for their purposes- only, in the case of current AI, the purpose may not even be clearly determined (alignment problem) or even worse: may be intentionally not in your best interests but those of the owner of the AI (weaponization).
One of the biggest dangers of current AI is that it is, by design, becoming very good at pretending to be human; and we are extremely susceptible to being fooled by this. When an AI pretends to have a feeling, or an intent, or even an understanding, our immediate system 1 response is to believe it. After all, we are used to communicating with actual people and being mostly correct about what to expect from them.
Consequently we are easily, by default, led to feel that we can relate to the thing we are communicating with as if it were a person- with all of the expectations we might assign along with that assumption. We have no impulse to examine what is actually occurring in that interaction and likely even less impulse to question or distrust it.
We must learn to be more critical and mindful of this so that we always keep these interactions in the proper context. The easiest thing to do, and often the most pleasing thing to do, is to buy into the illusion. This is also the most dangerous thing to do.