Discussion Query:
In his presentation on steganography
last time, Bhushan pointed out
that the message to be hidden could be "anything that can be
represented as a bit stream." You may recall that right at the end of
seminar I said that next meeting we would begin with the question, is
there any message whatsoever that cannot be represented as a bit
stream? This question obviously goes well beyond steganography, indeed
well beyond computer vision, but should be of concern to any serious
computer science student or practitioner.
Since we are not meeting this week,
and have a bit of unexpected time
on our hands, here is what I would like everyone in the seminar to do.
Please think about this question and write out your opinion, along with
some supportive reasoning or explanation. Put it in an email addressed
to me, I will gather the responses and post them all on our website.
Your reply could be as short as a paragraph, or longer if you wish, and
may or may not include references, that's up to you. Maybe you think
the answer is easy and direct, and requires only a sentence or two. Or
maybe you think it needs more detailed consideration to explain your
judgment.
Please send your email to me no later
than Monday Nov 9 9:00am. I will
hold all replies till then, and then post them all on our website. It
should make for some interesting reading!
-PS
---------------------------------------------------------------------------------------
Replies posted in order received:
Received 11/3/2009 11:03am
From: Peter Scott
In my opinion, the answer here depends in part on how we are to
understand "represent" and "message." Let me use common-sense
definitions, that a bitstream is a representation if it comes
arbitrarily close to fully specifying the information contained in the
message, and that by message we mean the content of any act of transfer
of information. With these understandings, I think there may be at
least one class of messages that cannot be represented as a bitstream:
the internal signals of consciousness. These are the primitive,
subjective messages from what Freud called The Id or Schopenhauer
called The Will. As Nagel argues in a wonderful paper called "What Is
It Like To Be A Bat?" the more we try to objectivize subjective
experience, the further we get from its representation. So until/unless
I see a good way around this problem, I will continue to believe that
bitstreams cannot capture the experience of my smelling a rose, or a
bat being a bat. Perhaps everything else, including images, sounds, and
conscious processes, but not everything.
-------------------------------------------------------------------------------------
Received 11/3/2009 11:11am
From: Bill Rapaport
Here are some quick observations.
First, let's consider messages that are "external" to an agent's mind.
My first reaction to the claim that there are (external) messages that
cannot be represented by a bitstream was that perhaps analog messages
(e.g., pictures, sounds) might not be thus representable. But if
you're willing to allow digital versions of them (and why wouldn't
you?), then I think that it's highly unlikely that there are any
messages that couldn't be represented in a bitstream, and probably
there aren't any.
Now let's consider messages that are "internal" to an agent.
Certainly, any physically realizable signal, e.g., a nerve signal,
ought to be representable by a bitstream. Again, we might have to
settle for a digital version of it, but that seems acceptable to
me. Even a chemical signal could be representable by a string
representing its chemical analysis, perhaps.
What of "internal signals of consciousness"? Well, I'm not sure
what you have in mind here. The current literature on
consciousness seems to be pretty much in agreement that there are two
kinds of consciousness: "access" consciousness and "phenomenal"
consciousness. The former is what psychologists study and that
can be expressed in a functional-computational theory. E.g., John
Anderson's ACT-R system is said by him to be conscious in this sense,
because there are messages that are globally accessible to all parts of
the cognitive architecture.
Phenomenal consciousness is the subjective feeling that is associated
with (some or all) conscious states; it also goes by the name
"qualia". Some researchers claim that it is not reducible to, or
explainable in terms of, access consciousness (Chalmers). Others
claim that it is (McDermott). Still others claim that there's no
such thing, but that we only think there is (Dennett).
I would say that internal signals of access consciousness, because they
are reducible to neuron firings, would be bitstream representable.
As for qualia, only those who believe that they exist and are not
reducible to neuron firings would need to answer the question.
But because no one has a theory about what they are (well, I'm working
on one, but...), I think this question needs to remain open.
-------------------------------------------------------------------------------------
Received 11/3/2009 11:51pm
From: Niranjan Kamat
In my opinion all the data,logic,etc should be bit wise representable.
Bits mean
0/1. It either means that something exists or doesnt exist. There is no
other
option than for something logically/physically to occur or not occur.
There
cannot be a third option of any entity instantiating. As Shakespeare
said 'To be
or not to be'. What else can there be? And since there is none the
logic /data
can be split up into smaller sets of logic which finally has to come
down to 0s
and 1s.
-------------------------------------------------------------------------------------
Received 11/6/2009
From: Ranjan Bangalore Seetharama
In my opinion it is human written scripts that cannot be converted to
bit-stream as they are very difficult to interpret.
The difficulty of recognition varies with a number of factors:
* Restrictions on the number of writers.
* Constraints on the writer: entering characters in
boxes, lifting the pen between characters, observing a certain stroke
order, entering strokes with a specific shape.
* Constraints on the language: limiting the number
of symbols to be recognized, limiting the size of the vocabulary,
limiting the syntax and/or the semantics.
The above information is taken from
http://cslu.cse.ogi.edu/HLTsurvey/ch2node7.html
In the above cited reference Section 2.5.2 speaks of the difficulties
and Section 2.5.4 speaks of current state of Softwares.
------------------------------------------------------------------------------------
Received 11/6/2009 6:01pm
From: Qiang Gao
In terms of your question, “is there any message whatsoever that cannot
be represented as a bit stream?”, it is hard to give direct answer. But
I would like to make a comment:
You could represent any message as a bit stream, but after that some
information or features in the message may be lost. For an example,
suppose that we have an audio message of someone saying something,
after convert it into bit stream, the original meaning the message
carries may be lost and you could never know what the guy means.
Another example comes out from my mind is: if a guy sign his name on a
blank paper, how can you convert it into a bit stream? You may take a
digital picture of the signature first. But if you do that, how can you
preserve the pressure information of the signature which is an
important feature to validate whether the signature is genuine or not?
Does it make sense? Or did I miss something?
------------------------------------------------------------------------------------
Received 11/6/2009 8:54pm
From: Ning Zhang
I agree that anything can be represented as a bit stream. Everything I
can think of can be represented as a bit stream nowadays, such as
pictures, music, video, characters and so much more. I could not think
of anything that cannot be represented
as a bit stream.
-------------------------------------------------------------------------------------
Received 11/8/2009 11:27pm
From: Mukul Apte
According to me, light cannot be represented as a bitstream so as to
exactly recreate it at the receiving end. If the message is created
using light waves (wave nature of photons) it cannot be represented as
a bit stream.
I do not mean to say that message is encoded in light (i.e. bit
represented as bright/dim light, r/g/b light etc.), what I mean is
light itself is the message (i.e. phases of light are used to represent
the message).
This is because even if the light is decode at the receiving end, it
will not be coherent to the light encoded at the source. No 2 sources
of light are ever coherent, they will have some phase difference,
(except LASERs).
-------------------------------------------------------------------------------------
Received 11/8/2009 11:44pm
From: Bhushan Chitte
According to me, if the message is in the form of light, it can't be
encoded in bits. What I am thinking is light as message and not as a
carrier as in optical fibers. Reproducing the exact phase and other
properties of light is not possible. Also, we can't encode things like
taste and smell which we can feel and want to transmit on some
occations.
-------------------------------------------------------------------------------------
Received 11/9/2009 1:33am
From: Vikas Choudhary
Q: Is there any message whatsoever that cannot be represented as a bit
stream?
A: Message containing
Irrational numbers (number which cannot be expressed
as a fraction m/n) ex: sqaure root(5)
Fractions like 1/3,22/7(pi) having infinite binary
expansion.
Consequently, the number representation and arithmetic operation
performed involving such numbers are never infinitely accurate.Hence,
such numbers can never be represented as bit stream accurately.
------------------------------------------------------------------------------------
Received 11/9/2009 1:35am
From: Porchelvi Vijayakumar
Modern Comptuers implement binary systems internally to represent any
kind of data. In order to process the data using a computer we devise
techniques which help us to represent this data in binary form. Most
data naturally exhibits a continous form (i.e analog signal of a
specific property like the brightness of an image).This signal could
easily be sampled and converted to digital values. Thus in conclusion I
believe that, if any message or information can be represented as
analog signal, then it could be represented using bits and hence a
binary stream.
------------------------------------------------------------------------------------
Received 11/9/2009 1:49am
From: Lynn S. Lobo
I don't believe that there is any data that cannot be in some way
represented by a bit stream.Even continuous data can ( with some loss)
can still be conveyed by reasonable standards.Even the most complex of
numbers/ characters can be in some way or the other encoded as a bit
stream.Even Analog signals can be represented with close to perfect
precision.Although if one had to encode a continuous signal without any
loss of information (or all the information in that signal needs to be
captured), then a bit stream would not be able to handle it however I
cannot think of data that would be like this.
-----------------------------------------------------------------------------------
Received 11/9/2009 5:09am
From: Santhosh Kandalu
I think every character in a message has a ASCII values which
represented in
binary bits('0' or '1'). So bit stream can be written for those
characters in
that message. Unless and until that message has some special greek
letters like
'α','β' etc, those characters cannot have
ASCII values and such message cannot be
represented in bit stream. This is my opinion.
------------------------------------------------------------------------------------
Received 11/9/2009 11:58am
From: Jaehan Koh
I guess, the real-world object message cannot be represented by bit
streams. The abstracted version of it or the model of it can be
expressed by a bit stream.
Also, the secret code message should not be represented by a bit stream
intentionally not to be decoded by an enemy. In that case, I guess,
information for decoding the hidden message might be uniformly
distributed over some intervals or hidden in the chunks of bits along
with fake bits that are used to make it impossible to decode.
-----------------------------------------------------------------------------------
Received 11/9/2009 2:47pm
From: Kedar Sarmalkar
I would like to say that there is no message w.r.t audio, text, image,
video that cannot be represented as bit stream. However if considering
a message as to be give as input to a machine or a computer the “smell”
can be one of the message which would be really difficult to represent
as bit stream. As there are wide variety of smells which be really
difficult to represent as bit stream. Also same might be true with
emotions.
-----------------------------------------------------------------------------------
Received 11/10/2009 2:53pm
From: Praveen Santhanam
In my opinion All computer inputs known today have a bit stream
representation possible. If we are to consider something like emotions
or sense of touch to become a computer input, it might be hard to
obtain a bit stream representation possible. I would like to comment
that everything may have a bit stream representation possible but it
might not necessarily be the best way to go.
-----------------------------------------------------------------------------------