jamest wrote:If nothing else, this discussion has highlighted some of the rational problems faced by those wishing to explain human behaviour in terms of a 'brain model'. For obvious [and explained] reasons, for example, we cannot have brain models that depend upon brains making assumptions about a reality beyond their own internal states. And neither can we have brain models that depend upon the brain giving external meaning to its own internal states. What is needed, then, is a brain model whereby the brain's internal responses to the external environment suffice to produce appropriate behaviour/response to that environment without the brain knowing about the external realm. I might budge on some issues, but on the issue of the brain knowing nothing more than its own internal states, I remain anchored to the spot.
The problem I see in your thinking there is that you have made huge assumptions about the nature of 'assumptions' and 'meaning', as if you had certain information that these cannot be 'tags' relating to representational relations between brain states and 'the world'.
An assumption is the application of a generalised pattern classification to a situation about which there is limited information. If the data is incomplete pick an interpretation that fits some of it. Do you have a sound argument against NNs doing that, firing in recognition of an incomplete stimulus pattern?
As for 'meaning' it is accuracy, saliency and survival. A NN that responds to an angry expression has 'meaning' that influences behaviour. The same expression recognised in the presence of other recognitions may render the detection more or less relevant - more or less 'meaningful'.
jamest wrote:Currently, my main issue is that the brain does seem to know about a reality external to its own states. And what is my reason or evidence for this claim? - 'Our' thoughts and words.
As we keep explaining, 'knowing' is the brain states. 'Knowing that your mother is looking at you' is a neurological response. There is no obvious need to add a superfluous entity of a Cartesian observer of the 'knowing' process.
jamest wrote:... If you reduce the 'I' to the brain, then intrinsically, whatever 'I' claim that I am thinking and feeling and saying, must be attributed to the brain itself. And the fact is this: practically all humans (except a small minority) think, feel & act, as though the world beyond itself is real. This should be obvious, as most of our responses are emotionally-driven: we fear the consequences of fucking up, for example.
My mind brain is currently awhirl with lots of questions for which 'brain models' seem insufficient to answer. For example:
1) If we require a 'brain model' whereby the brain's internal responses to the external environment suffice to produce appropriate response amidst that environment, then why does the brain require 'emotion' to produce the appropriate behaviour?
2) If the brain [evidently] needs emotion to produce 'appropriate' (self-serving) behaviour, and brain states are just responses to the environment, then from whence cometh 'emotion'? Certainly, the environment cannot be responsible for a self-serving attitude, or associated emotions that drive it.
3) If the response to the external environment is a consequence of a self-serving attitude, then how can we say that the brain's internal responses to the external environment suffice to produce appropriate response amidst that environment?
What is 'emotion'? Why is it not a set of NN 'tags' that represent processes in the brain that produce the appropriate response?
'Fear' makes you want to flee. We can account for the NN mechanisms that produce a flight response to danger cues. but how does the brain represent its own state? It can't see its own neurons and the most useful 'model' is something simple that fits with the other capabilities of the brain.
So the brain recognises a large scale pattern of neural activity and physical behaviour as 'Fear' in much the same way as it recognises the sensory signals from trunk, branches and leaves as 'Tree'. The big bag of 'fear responses' is represented as 'the emotion of fear', which serves to make sense of the individual's state, rather than drive it. You don't run because you feel afraid, you run because flight has historically (and pre-historically) been a good-enough response to such cues. The cues produces the responses. The responses are 'understood' as fear.
Looking at it subjectively, we become aware that we are fearful and that prompts certain bias in decision making. It focusses attention.
Do you see how this fits the model? The brain identifies potential threats as activation of various recognition circuits. This pattern of activation triggers certain physiological responses (adrenaline, faster heart rate etc). The brain recognises these responses as 'fear' and can thus predict its own behaviour, allowing for the possibility of modifying it for better outcomes.
jamest wrote:I have more questions like these. The point that I'm trying to make, is that ultimately, a model of the brain will make no sense unless it addresses such questions. No model will ever be acclaimed that fails to address reasoning... whilst simultaneously failing to impress, empirically. Unfortunately, all of this chatter about finite state machines et al, isn't really getting us anywhere. It's just conjecture that fails to address the rationale behind the issues, and has no sufficient empirical backing, either, to be of any worth in this discussion.
What was supposed to be a philosophical debate has turned into a pseudoscientific debate, that is oblivious to any rational concerns... so it seems.
Empirical data underpins the philosophy here. It demonstrates that NNs are self-organising classifiers that recognise patterns. It demonstrates how the patterns become 'learned' in the physical substance of the brain.
Introspection shows that the subjective self is not instigator of emotions of thoughts, or any aspect of mind, once we peer beneath the naive R1 façade.
Reason shows that the pieces fit together, that such concepts as meaning, knowledge, assume and aware, relate to representational relations between brain and mind that do fit the neurological models. NNs have those relations to the world.
