Loading ...
Sorry, an error occurred while loading the content.

Another example for images in the brain

Expand Messages
  • Eray Ozkural
    http://news.bbc.co.uk/2/hi/science/nature/471786.stm http://www.jneurosci.org/cgi/content/abstract/19/18/8036 I know that there are people who still deny that
    Message 1 of 13 , Mar 6, 2009
    • 0 Attachment
      http://news.bbc.co.uk/2/hi/science/nature/471786.stm
      http://www.jneurosci.org/cgi/content/abstract/19/18/8036

      I know that there are people who still deny that there are images in
      the brain. So I thought I would send this old research.

      I wonder how badly some "philosophers" understand neuroscience. I
      wonder how they think it makes sense to argue against representations
      in the brain.

      Bialek's decoding is of course way more accurate than this, if it were
      applied to the whole visual field it would make for exact decoding.

      I was telling you guys that soon enough we will be recording the
      visual input of humans (the stuff several research teams are working
      on) and who knows, long enough in the future and maybe we will be able
      to directly decode the 3d scene information from the visual system.
      (And even insert 3d information)

      --
      Eray Ozkural, PhD candidate. Comp. Sci. Dept., Bilkent University, Ankara
      Research Assistant, Erendiz Supercomputer Inc.
      http://groups.yahoo.com/group/ai-philosophy
      http://myspace.com/arizanesil http://myspace.com/malfunct
    • iro3isdx
      ... I guessing that you don t understand why people deny that there are images in the brain. For if you understood what they are denying, you would realize
      Message 2 of 13 , Mar 6, 2009
      • 0 Attachment
        --- In ai-philosophy@yahoogroups.com, Eray Ozkural <erayo@...> wrote:

        > http://news.bbc.co.uk/2/hi/science/nature/471786.stm
        > http://www.jneurosci.org/cgi/content/abstract/19/18/8036

        > I know that there are people who still deny that there are images in
        > the brain. So I thought I would send this old research.

        I guessing that you don't understand why people deny that there are
        images in the brain. For if you understood what they are denying,
        you would realize that the cited work does not in any way contradict
        their denial.
      • Eray Ozkural
        ... No, no, no. Those same idiots would still deny that there are any representations in the brain even if we decoded the output and long term memory of every
        Message 3 of 13 , Mar 6, 2009
        • 0 Attachment
          On Sat, Mar 7, 2009 at 1:30 AM, iro3isdx <xznwrjnk-evca@...> wrote:
          > --- In ai-philosophy@yahoogroups.com, Eray Ozkural <erayo@...> wrote:
          >
          >> http://news.bbc.co.uk/2/hi/science/nature/471786.stm
          >> http://www.jneurosci.org/cgi/content/abstract/19/18/8036
          >
          >> I know that there are people who still deny that there are images in
          >> the brain. So I thought I would send this old research.
          >
          > I guessing that you don't understand why people deny that there are
          > images in the brain.  For if you understood what they are denying,
          > you would realize that the cited work does not in any way contradict
          > their denial.

          No, no, no.

          Those same idiots would still deny that there are any representations
          in the brain even if we decoded the output and long term memory of
          every single neuron in the visual cortex.

          For instance the 3d / 2 1/2 d scene representations in the brain could
          be decoded as well, with extra effort.

          What is it that they are denying, Mr. Genius?

          The (decade old) research shows that the images (these being 2d
          images) are represented through a distributed neural coding just as we
          would expect.

          In what sense is there no two dimensional image in the brain?

          --
          Eray Ozkural, PhD candidate. Comp. Sci. Dept., Bilkent University, Ankara
          Research Assistant, Erendiz Supercomputer Inc.
          http://groups.yahoo.com/group/ai-philosophy
          http://myspace.com/arizanesil http://myspace.com/malfunct
        • jgkjcasey
          ... I don t think the fact that images are mapped onto the LGN of a cat proves that is what the cat sees anymore than the mapping of a webcam image into the
          Message 4 of 13 , Mar 6, 2009
          • 0 Attachment
            --- In ai-philosophy@yahoogroups.com, Eray Ozkural <erayo@...> wrote:
            >
            > On Sat, Mar 7, 2009 at 1:30 AM, iro3isdx <xznwrjnk-evca@...> wrote:
            > > --- In ai-philosophy@yahoogroups.com, Eray Ozkural <erayo@> wrote:
            > >
            > >> http://news.bbc.co.uk/2/hi/science/nature/471786.stm
            > >> http://www.jneurosci.org/cgi/content/abstract/19/18/8036
            > >
            > >> I know that there are people who still deny that there
            > >> are images in the brain. So I thought I would send this
            > >> old research.
            > >
            > > I guessing that you don't understand why people deny
            > > that there are images in the brain. For if you
            > > understood what they are denying, you would realize
            > > that the cited work does not in any way contradict
            > > their denial.
            >
            > No, no, no.
            >
            > Those same idiots would still deny that there are any
            > representations in the brain even if we decoded the
            > output and long term memory of every single neuron in
            > the visual cortex.
            >
            > For instance the 3d / 2 1/2 d scene representations in
            > the brain could be decoded as well, with extra effort.
            >
            > What is it that they are denying, Mr. Genius?
            >
            > The (decade old) research shows that the images (these
            > being 2d images) are represented through a distributed
            > neural coding just as we would expect.
            >
            > In what sense is there no two dimensional image in the
            > brain?

            I don't think the fact that images are mapped onto the
            LGN of a cat proves that is what the cat sees anymore
            than the mapping of a webcam image into the computer
            memory is proof that is what the computer program sees.

            "seeing" is what is done with the image. This is true
            for a computer program and I would suggest is also
            true for the brain.


            John
          • Michael Olea
            ... So ... Seeing is behavior? ;-
            Message 5 of 13 , Mar 6, 2009
            • 0 Attachment

              On Mar 6, 2009, at 7:53 PM, jgkjcasey wrote:

              "seeing" is what is done with the image.

              So ...

              Seeing is behavior?

              ;-\


            • Marvin Minsky
              ... In psychology, the word behavior means externally observable activity. I think most everyone would agree that Seeing (and Thinking) are Mental Activities
              Message 6 of 13 , Mar 6, 2009
              • 0 Attachment

                On Mar 7, 2009, at 1:08 AM, Michael Olea wrote:


                On Mar 6, 2009, at 7:53 PM, jgkjcasey wrote:

                "seeing" is what is done with the image.

                So ...

                Seeing is behavior?

                In psychology, the word 'behavior' means externally observable activity.

                I think most everyone would agree that Seeing (and Thinking) are Mental Activities (which most of us would also regard as either resulting from or actually being subset of in-brain activities.
              • Michael Olea
                ... You know, no doubt, about the experiments in which subjects wear goggles that turn the visual field upside down. After a while, a few days of wearing them,
                Message 7 of 13 , Mar 6, 2009
                • 0 Attachment
                  On Mar 6, 2009, at 10:19 PM, Marvin Minsky wrote:

                  >
                  > On Mar 7, 2009, at 1:08 AM, Michael Olea wrote:
                  >
                  >>
                  >> On Mar 6, 2009, at 7:53 PM, jgkjcasey wrote:
                  >>
                  >>> "seeing" is what is done with the image.
                  >>
                  >> So ...
                  >>
                  >> Seeing is behavior?
                  >
                  > In psychology, the word 'behavior' means externally observable
                  > activity.
                  >
                  > I think most everyone would agree that Seeing (and Thinking) are
                  > Mental Activities (which most of us would also regard as either
                  > resulting from or actually being subset of in-brain activities.


                  You know, no doubt, about the experiments in which subjects wear
                  goggles that turn the visual field upside down. After a while, a few
                  days of wearing them, the world suddenly appears right-side up.
                  Then, when the goggles are removed, the world appears, for a time,
                  upside down.
                • Bhupinder Singh Anand
                  Wouldn t this discussion be more productive if we could agree that my perception (or that of any homo sapien) of what my TV sees is the image that it shows
                  Message 8 of 13 , Mar 7, 2009
                  • 0 Attachment
                    Wouldn't this discussion be more productive if we could agree that my perception (or that of any homo sapien) of what my TV 'sees' is the image that it shows on it's screen? So, if the brain is essentially a computer, shouldn't my perception of what an individual brain 'sees' be, similarly, defined as the image that is seen on any screen (or technology) that I use to project such an image?

                    Wouldn't the quality of such projection - which would be the limiting factor in my perception - then reduce to a technological and engineering problem, rather than a philosophical one, for my species?

                    Regards,

                    Bhup
                  • Michael Olea
                    ... Cool paper, Eray; very much in the tradition of Rieke, Warland, Bialek, and their colleagues. It shows at least 3 things: 1) High mutual information (in
                    Message 9 of 13 , Mar 7, 2009
                    • 0 Attachment

                      On Mar 6, 2009, at 2:15 AM, Eray Ozkural wrote:

                      http://news.bbc.co.uk/2/hi/science/nature/471786.stm
                      http://www.jneurosci.org/cgi/content/abstract/19/18/8036

                      I know that there are people who still deny that there are images in
                      the brain. So I thought I would send this old research.



                      Cool paper, Eray; very much in the tradition of Rieke, Warland, Bialek, and their colleagues.

                      It shows at least 3 things:

                      1) High mutual information (in the strict Shannon sense) between the spiking activity of neurons in LGN and video signals.

                      Of course, it would be astounding if were otherwise, but this begins to quantify the relationships.

                      2) Much of the video signal can be recovered using linear reconstruction filters.

                      Bialek et al. got similar results from their studies of the spike trains of individual neurons: whereas the (probabilistic) map from signal to spike train -- P[spike(t_i) | signal(t)] -- was highly non-linear, the (probabilistic) map from spike train to signal -- P[signal(t) | spike(t_i)] -- was nearly linear. This makes "reading the neural code" tractable.

                      3) Much of the video signal can be recovered using a simple "rate code" (spikes per second) that ignores details of inter-spike timing.

                      Again, Bialek  and colleagues found in their studies that approximately half of the information (in bits) spike trains convey about a signal can be recovered from spike rate alone, but more detailed information about the signal becomes available when details of inter-spike timing, not just average spike rate, are taken into account. So that is an open question (or was as of 1999) for the Stanley, Li, and Dan paper - how much better does the signal reconstruction become if spike timing is taken into account? The difficulty is that to do that using reverse correlation requires lots of data. Bialek and those guys were able to get those huge data sets recording from the H1 neurons in flies because the recording setup was very stable. They could leave the electrode in there for hours at a time, pausing the experiment every now and then to feed the fly.

                      What is not clear, however, at least to me, is what sort of bearing this work has, if any, on the "analog/propositional" debate between the likes of Zenon Pylyshin and Stephen Kosslyn. This debate was about the mechanisms involved in recalling visual information.


                      I wonder how badly some "philosophers" understand neuroscience. I
                      wonder how they think it makes sense to argue against representations
                      in the brain.


                      I suspect it is more a matter of arguing against particular kinds of representations (and sterile quibbling over the word "representation" itself).

                      Bialek's decoding is of course way more accurate than this, if it were
                      applied to the whole visual field it would make for exact decoding.

                      I was telling you guys that soon enough we will be recording the
                      visual input of humans (the stuff several research teams are working
                      on) and who knows, long enough in the future and maybe we will be able
                      to directly decode the 3d scene information from the visual system.
                      (And even insert 3d information)


                      Well, maybe, but do not forget the results that specific neural codes vary from individual to individual, and even within a single individual vary over time. So to build reconstruction filters would not only take gobs of data, but the filters you built for Joe would not work on Casey, and would not work on Joe a year, say, later.

                      Also, you know that famous binary image of a dalmatian in dappled light and shade? At first encounter most people do not see the dog right away (or at all, until told there is a dog in the picture). But once having seen the dog, seeing the picture again months and even years later they see the dog right away. My point is that what people see depends on what they have seen in the past. Reconstruction of a 3d scene is an ill-posed problem in the sense that there are many solutions consistent with any given visual input -- you could not deduce from the visual system alone what 3d scene people see, but would have to take also into account that "massive common-sense database" even five-year olds draw upon on the path from photons to phenomenology.

                      -- Michael

                    • iro3isdx
                      ... There are several kinds of statement one might make: (1) The brain contains information about the world; (2) The brain contains a representation of the
                      Message 10 of 13 , Mar 7, 2009
                      • 0 Attachment
                        --- In ai-philosophy@yahoogroups.com, Eray Ozkural <erayo@...> wrote:

                        > Those same idiots would still deny that there are any representations
                        > in the brain even if we decoded the output and long term memory of
                        > every single neuron in the visual cortex.

                        There are several kinds of statement one might make:
                        (1) The brain contains information about the world;
                        (2) The brain contains a representation of the world;
                        (3) The brain contains an image of the world.

                        As far as I know, (1) is not at all controversial. And it seems
                        to me that the work you cited provided only evidence for (1).

                        (2) and (3) are more contentious. Moreover, it seems silly to
                        argue about them. Surely what matters is (1), together with how
                        that information is used.

                        > What is it that they are denying, Mr. Genius?

                        An image is something to be looked at, that bears a likeness to
                        what it is an image of. To say that the brain has an image is to
                        suggest that there is a homunculus inside the brain that is
                        looking at (perceiving) the image. And that explains nothing.
                      • Murat
                        There you are a ph.d. thesis proposing a model for the visual cortex within the same lines of research: http://www.mrtc.mdh.se/publications/0854.pdf cheers
                        Message 11 of 13 , Mar 9, 2009
                        • 0 Attachment
                          There you are a ph.d. thesis proposing a model for the visual cortex within the same lines of research:

                          http://www.mrtc.mdh.se/publications/0854.pdf

                          cheers

                          --- In ai-philosophy@yahoogroups.com, Eray Ozkural <erayo@...> wrote:
                          >
                          > http://news.bbc.co.uk/2/hi/science/nature/471786.stm
                          > http://www.jneurosci.org/cgi/content/abstract/19/18/8036
                          >
                          > I know that there are people who still deny that there are images in
                          > the brain. So I thought I would send this old research.
                          >
                          > I wonder how badly some "philosophers" understand neuroscience. I
                          > wonder how they think it makes sense to argue against representations
                          > in the brain.
                          >
                          > Bialek's decoding is of course way more accurate than this, if it were
                          > applied to the whole visual field it would make for exact decoding.
                          >
                          > I was telling you guys that soon enough we will be recording the
                          > visual input of humans (the stuff several research teams are working
                          > on) and who knows, long enough in the future and maybe we will be able
                          > to directly decode the 3d scene information from the visual system.
                          > (And even insert 3d information)
                          >
                          > --
                          > Eray Ozkural, PhD candidate. Comp. Sci. Dept., Bilkent University, Ankara
                          > Research Assistant, Erendiz Supercomputer Inc.
                          > http://groups.yahoo.com/group/ai-philosophy
                          > http://myspace.com/arizanesil http://myspace.com/malfunct
                          >
                        • Marvin Minsky
                          ... Wow. I couldn t see what was either the aim or the outcome of this research. It seems to be a theory of how some parts of a network can be tuned to
                          Message 12 of 13 , Mar 9, 2009
                          • 0 Attachment
                            On Mar 9, 2009, at 5:34 PM, Murat wrote:

                            > There you are a ph.d. thesis proposing a model for the visual cortex
                            > within the same lines of research:
                            >
                            > http://www.mrtc.mdh.se/publications/0854.pdf

                            Wow. I couldn't see what was either the aim or the outcome of this
                            research. It seems to be a theory of how some parts of a network can
                            be tuned to detect the orientation of edges, but I couldn't find any
                            summary.

                            To me, one important problem is to find (and represent) the boundaries
                            of significant visual regions. Another is to find certain important
                            kinds of relationships between boundaries—for example, like those
                            discussed in Section 3 of http://web.media.mit.edu/~minsky/papers/PR1971.html
                            . But I couldn't find any such discussions.

                            Then, why did you include this reference? Did it help you to answer
                            some questions you had?
                          • Murat
                            My only intention was to provide another example in the computational nueroscience lines perspective dealing with the visual regions which considers a
                            Message 13 of 13 , Mar 10, 2009
                            • 0 Attachment
                              My only intention was to provide another example in the computational nueroscience lines perspective dealing with the visual regions which considers a synthesis problem in a complementary fashion to the -comparably analysis sort of - approach in those work Eray pointed out.

                              The style of the manuscript is a bit unusual to me, but yes, it attempts to model the lower layer computation (edge detection etc.) that is carried out in the visual cortex.

                              AFAIC, they have enough experimental evidence (collected through animals etc. ) to come up with such models however, I am not sure that they have this for the higher level processing parts. If I' m not wrong, we can sort of know the regions which use the outputs of these lower layers and through examination of some interesting diseases can characterize some properties but they still lack even an architecture.

                              Maybe a complete analysis of the brain is not that necessary for the synthesis of an AI but I'd bet it'd help a lot. And for this,
                              I suppose we need to develop better machines that would help us better the ways we experiment on brains such as high time resolution MRIs, better source located EEGs, or maybe invasive yet lower risk techniques.


                              --- In ai-philosophy@yahoogroups.com, Marvin Minsky <minsky@...> wrote:
                              >
                              >
                              > On Mar 9, 2009, at 5:34 PM, Murat wrote:
                              >
                              > > There you are a ph.d. thesis proposing a model for the visual cortex
                              > > within the same lines of research:
                              > >
                              > > http://www.mrtc.mdh.se/publications/0854.pdf
                              >
                              > Wow. I couldn't see what was either the aim or the outcome of this
                              > research. It seems to be a theory of how some parts of a network can
                              > be tuned to detect the orientation of edges, but I couldn't find any
                              > summary.
                              >
                              > To me, one important problem is to find (and represent) the boundaries
                              > of significant visual regions. Another is to find certain important
                              > kinds of relationships between boundaries—for example, like those
                              > discussed in Section 3 of http://web.media.mit.edu/~minsky/papers/PR1971.html
                              > . But I couldn't find any such discussions.
                              >
                              > Then, why did you include this reference? Did it help you to answer
                              > some questions you had?
                              >
                            Your message has been successfully submitted and would be delivered to recipients shortly.