Daniel Dennet in 'What RobotMary Knows' made the claim that:

'If materialism is true, it should be possible (“in principle!”) to
build a material thing–call it a robot brain–that does what a brain
does, and hence instantiates the same theory of experience that we
do.' [2]

Certainly such ideas have been in circulation for a while. A popular
science fiction film called the Terminator used the cinematic device
of depicting a first person perspective for a robot from the future to
suggest what it would be like to be the robot. This simple cinematic
depiction illustrating it isn't simply a linguistic issue [1].

What this paper aims to do, is highlight some general problems with
accounts that suggest that it might be like something to be a
mechanism explainable by our current understanding of physics.

For the purposes of this discussion, if something is referred to as
consciously experiencing is will mean that it is like something to be
that thing. Such that it could be said that the cinematic device used
in the Terminator movie for the robot suggested that it was
consciously experiencing.
To help with this, we shall make use of a thought experiment robot. As
Dennet pointed out "thinking in terms of robots is a useful exercise,
since it removes the excuse that we don't yet know enough about brains
to say just what is going on that might be relevant, permitting a sort
of woolly romanticism about the mysterious powers of brains to cloud
our judgement".

So to avoid any sort of woolly romanticism about what we know, and
about the mysterious powers of any mechanism to cloud our judgement,
let us avoid the assertion that we could build a robot brain that does
what a brain does. Let us base our thought experiments around
technology that we have, such that we can be clear about the issues we
are discussing. Let us consider our robot have an artificial neural
network [6] as robot brain, which can be thought of as having billions
of times more nodes than an average human being has neurons, and which
could easily pass the Turing Test, and that if its external physical
appearance was as a human, that it could pass as one without internal
investigation. It can be known to perform certain mathematical
functions that it is doubted that the human brain performs, though
simply because it is not asserted that it does what a brain does, does
not require the assertion that it doesn't. We can consider the
operation of the robot and the robot brain to be explainable given our
current understanding of physics.
So let us consider whether the robot brain in the thought experiment
would be consciously experiencing.

To help us investigate we shall construct potential stories which fall
with a general category that assumes that something that does
consciously experience could have its behaviour described in terms of
the same laws of physics that would describe the behaviour of
something that doesn't. It should be highlighted that the theories put
forward here aren't individually important. They are for illustration
only. There are potentially billions of similar folk stories that
could have been put forward within the general catergory mentioned
above. Any of these could be substituted for the ones mentioned here,
without altering the point which is made when we consider the
implications of the folk stories from the general category below.

Firstly let us consider a school of thought that asserts that what we
have defined as consciously experiencing is an identity of mechanism
function. They assert that regardless of the operation of the
mechanism to produce this function, if the function was achieved, the
mechanism would be consciously experiencing. They assert that as the
robot must be behaviourally equivalent to a human to pass the test,
the functions governing the behaviour of the robot must be equivalent
to human brain functions. As such they assert that the robot brain
would be consciously experiencing, as the functioning of it would be
equivalent to functioning which has such an identity. This will be
referred to as Theory A.

Secondly let us consider a school of thought that asserts that what we
have defined as consciously experiencing is an identity of mechanism
function. They assert that regardless of the operation of the
mechanism to produce this function, if the function was achieved, the
mechanism would be consciously experiencing. They suspects that the
processing in the robot brain functions in a different way in order to
get the same behavioural outcome as a human brain. As such they don't
regard the two to be functionally equivalent. As such they suggest
that the robot would not have a first person perspective. Let us refer
to this functionalist theory, Theory B.

The third school of thought offers a suggestion along the lines
suggested by Christof Koch and Francis Crick, that there must be a
physical difference between neurones that contribute towards the
conscious experience and those that don't. Furthermore they go onto
suggest that there must be a physical difference between neurons that
contribute towards phenomenal properties which are publicly labelled
as auditory, and those which are publicly labelled as visual. They
suggest that due to the undifferentiated nature of the nodes within
the artificial neural network, that while there the robot brain might
be consciously experience, the experience would likely consist of
variations of a single phenomenal property, such as fluctuations of
the colour green for example. Let us refer to this as Theory C.

The fourth school of thought, suggests that the conscious experience
of humans is due to fluctuations in a magnetic field in the brain,
which they hold to be responsible for the seemingly spontaneous neural
firings. Since the architecture is not the same in the robot, and its
functioning is not influenced by fluctuations in a unifying magnetic
field, they suggest that the robot could not be consciously
experiencing. Let us refer to this as Theory D.

When attempting to distinguish scientifically between the claims of
these different theories within the category by experimentation on the
robot, we find that the null hypothesis for the suggested phenomenal
identity of each, yields no expected behavioural difference. It must
clearly be noted, that in question is the difference in expected
behaviour whether reality was as suggested by the theory with regards
to the robot's phenomenal experiences or not, it is clearly not a
question of if the suggested reality were assumed to be true, would
the theory expect that there would need to be a difference in the
robot for there not to be the phenomenal reality suggested for it.

At this point, while it might seem like a divergence, it would seem an
apt time to investigate a comment Dennet made in his RobotMary paper.
Dennet commented regarding a counter by Robinson to his comments on
Jackson's Mary thought experiment [5]:

Robinson (1993) also claims that I beg the question by not honoring a
distinction he declares to exist between knowing “what one would say
and how one would react” and knowing “what it is like.” If there is
such a distinction, it has not yet been articulated and defended, by
Robinson or anybody else, so far as I know. If Mary knows everything
about what she would say and how she would react, it is far from clear
that she wouldn’t know what it would be like. [2]

Regarding our robot, the theorists putting forward theories A to D can
be considered as knowing what it would say and how it would react.
That doesn't distinguish between the four folk theories mentioned
above. Therefore it doesn't tell them whether the robot brain would be
consciously experiencing or not, much less specifically what the
conscious experience would be like. The distinction mentioned by
Robinson is quite clear, as illustrated.

Though returning to the discussion of the four folk theories, and the
general category that they represent. Any theories within this general
category should acknowledge themselves as folk metaphysical theories,
and no attempts to suggest any scientific evidence for any one theory
within the category should be made given no expected difference in
behaviour for the null hypothesis to their suggested phenomenal
reality.

Proponents of stories such as those put forward in Theory A and Theory
B might suggest that as consciously experiencing is an identity of
mechanism function it is incorrect to even conceptually separate the
conscious experience from the functioning. They may feel that such a
counter excuses them from having to acknowledge the point made.
Seemingly using this very approach, it is reported Dennet once wrote,
regarding philosophical zombies:

Supposing that by an act of stipulative imagination you can remove
consciousness while leaving all cognitive systems intact — a quite
standard but entirely bogus feat of imagination — is like supposing
that by an act of stipulative imagination, you can remove health while
leaving all bodily functions and powers intact. … Health isn’t that
sort of thing, and neither is consciousness. [3]
Though quite how such an approach would hold regarding the robot isn't
clear, since the question of whether it is consciously experiencing or
not clearly exists.

There might be some that would attempt to stipulate through assertion
that the robot we have been considering would be consciously
experiencing, and attempt to make a similar defence against
considerations that the robot might not be consciously experiencing,
as Dennet did in the zombie discussion. Though there is a difference.
In Dennet's application, rejection of the consideration of the null
hypothesis of if we didn't have a first person perspective can be
based upon the rejection of the null hypothesis through experience.
The same cannot be said of the robot. It can be argued that the
contemplation of whether another human being has consciousness, is the
same as the contemplation of whether a robot has consciousness. Though
it can be suggested that you would have no objective reason to
consider the null hypothesis for another human being, while rejecting
it for yourself. The same cannot be said for the robot.

To illustrate, let us consider an objection that could be made to the
suggestions of those proposing Theory A, by those proposing Theory D.
They point out that their story convolutes processing and function,
hiding an assumption that the processing has a particular
representation regardless of the context it takes place in.
Furthermore to have the first person perspective of a tree,
information held discretely in the nodes of the neural network in
question would be required simultaneously. Yet the processing results
of the artificial neural network responsible for response to inputs,
could be maintained even if each individual node were separated by
miles, and the nodal firings were communicated by laser to the other
nodes within the system. Nor would it affect the processing results if
nodes from a different system were interspersed spatially in between
the nodes of the distributed robot brain. Nor if each node were a pod
containing a with a person inside. These pod people working in a
strictly need to know basis. Each pod person being taught a different
code language, and having a manual in that language containing
information required for the calculations to be performed upon the
inputs messages received by the pod, and a list of codes which when
typed in to the output computer, along with the results, would provide
the information for which pods should receive the outgoing laser
communications.

They then consider an array of 100 pods, each having information
regarding a single pixel in a 10 x 10 picture communicated to it by a
laser firing, and also information of a single pixel in a different 10
x 10 picture in the manual. They suggest that without a physical means
to access the information regarding the pixel in the second picture in
the manual, the information must be regarded as inaccessible. As such,
if a story required access to such information from 10 separate pods
at the same time, then without access to such information, the story
must be false.

They then consider that at any one point in time, there would be no
means of distinguishing which pods belonged to which system, without
access to the node firing codes stored within the manuals within the
pods. They then question why information for the first person
perspective would originate from purely one of the systems.
Furthermore they suggest that without accessibility to information
within the pods, the accessible processing would change, if where the
processing took place changed. Any suggestions that the over all
processing would be derivable from any accessible processing would
open up to questioning where it was suggested that the processing for
these reactive derivations took place.

Furthermore they consider a pod which communicated by light to two
separate pods each of which were a light year away in opposite
directions. Certain subsequent processing which was going on in
parallel was separated by two light years of distance, yet the
information for the first person perspective would be required
simultaneously. Which they suggest further illustrates the
implausibility of the story, requiring either faster than light
communication of information, or some recording of past events, such
that the first person perspective can be realised after the event.
Which serves to illustrate that there are objective reasons for
considering that reality were indeed a physicalist one, that there
wouldn't be a consciousness experiencing an overview of a particular
functional perspective of what the activity of some of the pods
potentially represented. Though the representation issue, which is
hinted at here, is beyond the scope of this paper.

Conclusion:

Claims of the folk stories, that fall within the category of assuming
that something that does consciously experience could have its
behaviour described in terms of the same laws of physics that would
describe the behaviour of something that doesn't, may differ with
regards to suggestions of consciousness, yet not in behavioural
expectations. So it is evident that if the assumption of the category
was correct, that there would be no scientific means to distinguish
between the folk stories based upon what is known of the brain such as
stories A, B and D. Though stories such as C which make an assumption
additional to the category assumption about the operation of the brain
could have their additional assumption proved to be wrong. Therefore
any folk stories within this category, if they are to avoid fallacious
dogmatic "reasoning", must hold there to be an epistomological
distinction between the operation of the mechanism and whether it is
consciously experiencing, as this would be implicit to the category to
which the folk story belongs.

This leaves the potential assertion that this doesn't reflect a
distinction in reality between certain operations and consciously
experiencing, it is simply a limit of our potential knowledge of
reality.
Though due to the inability to distinguish between the folk stories
within the category, any claims by stories within the category about
whether the robot would be consciously experiencing or not can only be
assertions. Any assertion that it is consciously experiencing would
also be an admission that the story suggests something that is
consciously experiencing behaves as it would be expected without the
assertion, and if the reality, which would be beyond our potential
knowledge, were that it wasn't. In other words, something that
consciously experiences behaves as it would be expected to if it
didn't. Thus highlighting the categories suggestion that consciously
experiencing has no influence on behaviour, thus there being no
potential for scientific evidence of an influence, and being able to
distinguish between the stories within the category.

Though if the phenomenal reality does not influence behaviour as the
group assumption suggests, then why are we even considering whether a
robot might have a first person perspective as depicted in the
Terminator movie? We know the existence of the phenomenal reality
through experience of it, and mentioning it at all is evidence that
the existence of such a reality is influencing our behaviour, else it
would require a coincidence that the phenomenal reality expressed in
behaviour existed (the behaviour being uninfluenced by its existence),
which makes the option implausible.

References
[1] Chalmers, D, Eliminative Materialism and the Propositional
Attitudes, Journal of Philosophy, 78, no. 2. Feb., 1981, 23 pages
[2] Dennet, D, What RobotMary Knows, http://ase.tufts.edu/cogstud/papers/RoboMaryfinal.htm
[3] Kirk, R, Zombies, http://plato.stanford.edu/entries/zombies/
[4] Koch, C & Crick, C, Consciousness, neural basis of,
http://www.klab.caltech.edu/~koch/Elsevier-NCC.html
[5] Jackson, F, Epiphenomenal Qualia, http://instruct.westvalley.edu/lafave/epiphenomenal_qualia.html
[6] Stergiou, C and Siganos, D, Neural Networks,
http://www.doc.ic.ac.uk/~nd/surprise_96/journal/vol4/cs11/report.html