Greg Detre
Wednesday, 14 November, 2001
@8.30 in the Summer Common Room
in summary
yes � problems about the self (computational insights + ideas, less so in neuroscience)
no � with respect to qualia (despite the work in neuroscience)
there�s no single problem of consciousness
2 main questions
problems about the self, self-consciousness and the �I�
problems about qualia, experience
she�s going to focus on the more difficult option of qualia
over-optimistic timescale in neuroscience � by the end of this century
Fodor � cats among pigeons � in a review of a book on linguistics
nobody has the slightest idea or even what it would be like to have the slightest idea how the material could be conscious
McGinn � just aren�t capable (our cognitive apparatus), just as dogs can�t understand multiplication
she thinks that McGinn may be right, but it�s far too early to tell
neuroscientists might argue that we�ve learnt a lot in the last 50, even 20 years
yes, but � we�ve found mind-body correlations
e.g. Lockethetis and Showell � binocular rivalry
special spectacles, black patch over right eye, horizontal grid over left � see a horizontal grid
swap the patch/grid, see a horizontal grid
take away the black patch, horizontal/vertical grid in opposite eyes
what do you see? squares? no, you shift back between the two � you don�t integrate the two � like the Necker cube
it can be speeded up and slowed down, but it�s largely involuntary
usually complete dominance for a while, then the other eye
try the same experiment with monkeys, trained to raise one paw with horizontal/vertical
the monkey raises one hand then the other in sequence � assume that a corresponding phenomenon is going on
unless you�re a follower of John McDowell, then you probably think that monkeys have some experience
while horiz grid stimulus presented, certain cells active in the visual cortex, other cells active only during vertical grid
another group active only when the monkey was raising one hand, and another group correlated with raising the other hand � and this is in the visual cortex
Descartes wouldn�t have liked this quote from Francis Crick: �FW is located in or near the anterior cingulate sulcus�
but that�s still just a correlation, not an explanation (necessary perhaps but not sufficient)
e.g. in 500 years, if you could tell which part of the brain would light up for any given phenomenal event. but let�s say that every different phenomenal event is randomly, unpredictably, unintelligibly scattered across the brain, then you wouldn�t feel that you had an explanation � explanation needs you to come up with counterfactual conditions. he would not be able to say, �if those cells had not been stimulated, then you wouldn�t have felt that experience�.
systematicity + isomorphism � give you intelligibility + predictability + counterfactual conditions
isomorphism � there is some sense in which there is a similarity of structure between the experience and the brain mechanism, i.e. systematical structural mapping relating experiences to brain events, and different brain events to other brain events and different experiences to other experiences
nusemann + salzmann � middle temporal visual cortex in monkeys,
orientation detectors, respond selectively to certain orientation, and also movement detectors � present different visual stimuli and see which cells fire in response. beforehand, they trained the monkeys to move their eyes in the direction of perceived movement
so then they trained stimulating those cells, and watching what the animals did with their eyes
found a geometric (360�) representation topographically organised, as evidenced by the monkeys eyes
of course, they only took a sample, but they could predict either what the cells would fire for a given stimulus, as well as what the monkey would do with its eyes if you were to stimulate those cells
Paul Churchland hypothetical experiment
sketched out a taste space. we know there are 4??? different receptors on the tongue, assume 4 different nerve fibres. some evidence for specific firing rates. imagine that�s true. gives a multi-dimensional quantisable space that could be mapped out onto the neurophysiology of taste. would have to do psychological experiments, asking people how similar they think pairs of tastes are. suppose that you found (and he predicts) intersubjective agreement about which tastes are similar and which aren�t, and that that agreement can be seen reflected in the neurons. this is a mapping of phenomenological and neurological space.
he thinks that to taste roast turkey just is to have these fibres firing at such and such a rate
unlike Crick, he at least realises that this is not a straightforward empirical hypothesis
Hardy � in 1912 was giving a lecture, and pointed to something as intuitively obvious, one of the students asks if it�s obvious, so Hardy furrows his brow, chews his chalk, walks round the quad a couple of times, comes back and says, �yes, it�s obvious�
nothing�s obvious here at the moment
it�s not just a case of just more neurophysiological experiments (even adding a few more perceptual categories to neuroscience) in order to understand these things
i.e. how there could be any consciousness at all?
(which is a very different question from �given that there is consciousness, why is it like this rather than that?)
requires a fundamental conceptual revolution at least as fundamental/creative/deep as Maxwell�s field theory (e.g. where people argued about the nature of light). Maxwell�s theory enabled you to generate statements about light as both corpuscles and waves, and the circumstances under which you�d get one or the other
she doesn�t think Penrose�s or Chalmers� suggestions are promising
dichotomy within philosophy over realism, e.g. Putnam started out as a scientifically-minded realist, and put forward functionalism � now he�s rejected functionalism and all notions of independent reality
she prefers the more realist approach to science
subtle but important shift in the status of mental events
she�s less keen on evading the nomological route
her answers of isomorphism + systematicity don�t help with fodor�s problem � she thinks that Fodor
creative consciousness � very different use of �consciousness�
monkey horizontal/vertical grid � who was that???
with the monkey, where in the visual cortex were the cells relating to raising one hand??? posterior parietal???
but what more would the 500-year
future neurophysiologist need??? surely even systematicity + isomorphism wouldn�t
be an explanation of consciousness???
what�s the difference between Churchland�s eliminativism and materialism/identity theory??? is his theory a functional one??? how can we talk of the same taste and its corresponding mapping between different people???
I think you have to view human-neurophysiologied zombies as being ridiculous � but computer functionally-equivalent zombies???
isomorphism as �structural mapping�??? in which case, I want to find an isomorphism between phenomenal and gross neurophysiological, or even cellular neurophysiological (e.g. pattern recognition in neural networks as inherent connectionist property)
but the pre-Maxwell arguments about the nature of light (corpuscles vs waves) � well, our current understanding sees light as a combination of both � how�s that progress??? oops, that�s wrong (see notes). but the problematic notion of light really centres around the same phenomenological explanatory gap, so we haven�t managed any intertheoretic reduction of the hard central question about the nature of light, i.e. why does light feel/seem �light/bright/coloured� to us???
what�s missing about Chalmers� suggestions??? they�re pretty vague, after all�
if �nobody has the slightest idea or even what it would be like to have the slightest idea how the material could be conscious� then will we know such a theory if we find it
where does she stand on GOFAI vs connectionism???
what�s her view on creativity???
our concepts and representations are ragged, overlapping, multi-modal, messy - i think this is because there aren't really distinct atoms, just layers and distributed ensembles - symbolic approaches are necessarily neat, because that's the way we think - we can't help but build up out of modules - beyond a minimal level of complexity, we have no choice - we think in serial terms, but mother nature's solutions are massively parallel. given all this, what chance does GOFAI have of achieving human-level (though not necessarily human-like) AI???
although of course one could imagine a multiple drafts model comprised of symbolic units, how would they self-organise??? yes, we could specify the overall structure and interconnectedness using a GA in some way, but is there a means to do it within a single agent's lifespan, i.e. to learn???
I�m looking for literature on why connectionist models of mind might be fundamentally incompatible/rendered compatible somehow with rational objectivity
there�s more systematicity + isormorphism in the lowest levels in the NS than in the higher levels though??? (I think he�s wrong/doesn�t understand what she means by isomorphism, relative to our phenomenal experience)
Stratten � first wore the inverted spectacles � never became fully competent/normal, at least not within the week
newborn chick starves if you put spectacles that reflect the light 5� to the left � they will starve to death and will never peck in the right place
Pellonis and Levash(???) inspired Churchland�s taste-mapping idea
you can be conscious if born without your cerebellum
difficulties for connectionism include representation of hierarchy, sequence/order
to what extent and what kind of consciousness is necessary/important for creativity?
in the evaluation stage � creativity is part of the concept � it comes in the self-evaluation (is that idea valuable or not?), self-criticism + self-amendment
e.g. Picasso, �je cherche pas; je trouve�
even Mozart who supposedly composed pieces entire, the reality is that he did correct some stuff along the way, though nowhere near as much as someone like Beethoven
does she think that language is important for creativity???
I think she thinks that linguistic ability affects your cognitive processes
or that there�s language/structure in the world
if a computer comes up with a better-than-human solution, is it creative?
Deep Blue isn�t creative (even within its domain)
thinking about it, isn�t creativity intimately related to aesthetics???
and the programmers aren�t really the creative ones there either (especially not in terms of chess, because they couldn�t, even collectively, beat Kasparov), other than as engineers � and that creativity doesn�t translate into chess-space creativity
does creativity involve identifying a goal?
well, only humans can talk about that goal, but a dog jumping for a bone has a goal, and she�s not sure that creativity necessarily involve identifying a goal anyway
book by Andrew Harrison � Making and thinking � why is it that we don�t sit and cogitate for hours then just write/draw/sculpt something straight out whole? �because things happen in the external world that you hadn�t planned
is that because our representations aren�t accurate/detailed enough to know how it�s going to turn out??? well, our linguistic representations don�t distort on implementation, but we�re unable to hold/remember it all in our head at once though
Gardner � 7 creative minds of the 20th century (Einstein, Eliot, Martha Graham, Stravinsky, Picasso, Gandhi, �) � they were all impossible swine to live with, and that�s no accident � being considerate takes time, and being friendly takes time away from concentrated learning about your domain, and also you might not respond so positively when criticised/huge ego
is AI going to help with this???
well, neurosci already is very computational, and so computation ideas will definitely be involved
and they help with problems of self + freedom
but a satisfactory computational explanation of qualia�? doesn�t see how � everything else, yes, maybe
sloman + dennett � think that distinguishing between qualia is just a computational question, that�s all that there is here