I agree... There's a world of difference between being able to jack someone's *perceptions* and being able to jack their *cognition* (or their consciousness).

I wasn't in any doubt of the capacity to do the former - and I'm not in much doubt that it will eventually become possible to extract memory engrams in the form of neural architecture and weightings - although I am extremely sceptical that we will see this in our lifetimes! :)

However, we may never be able to exchange the experience of being inside another person's belief system. Half the problem with consciousness is that we can't even adequately define it! :)

Hell yes. I agree, we'll inevitably be able to record someone's basic, electrical-impulse perceptions -- but stitching together consciousness outta that ... I'm not hopeful.

The lateral geniculate nucleus is a PART of part the brain that integrates sensory input (the thalamus), but the LGN is strictly dedicated to visual processing almost exclusively. In fact, it is a very early level of processing, second only to the retina (although there are some modifying feedback loops coming from the cortex). Since this area is at so early a step in visual processing, it is easy to say that this is a far cry from accurately describing what the organism is PERCEIVING (for example, the LGN may be working intactly, but without an active cortex, the animal would perceive nothing).

However, I don't think this means that we are incapable of deceiphering what an organism is perceiving. Indeed, we have shown some signs of being able to do that (see here for an example). Kristof Koch's book The Quest for Consciousness provides a good primer for anyone interested in this topic, and many of the recent advancements that have been made. He also talks a lot about theories he and Francis Crick have regarding the nature of consciousness and the problem of qualia.

WARNING: THIS IS LONG
- and having now written it up, it appears that Steve E. more succiently states what I am getting at, but I've typed all this up when I should have been studying civil procedure so I'm gonna post it.

I've got a complaint that was sort of raised in the comments over in Chris's blog.

Now, I didn't read the full article, only the Chris's post and your own. So if I am complaining about something addressed in the original article I apologize.

Anyways, beyond two people viewing the same image and drawing different inferences from it, as you point out, there is the argument that a person's brain may represent visual information (really any information, but let's keep this simple) in a unique and individual way.

I'm going to make an extreme version of my objection, mostly because it seems to best way to clearly state what-in-the-heck I am getting at:
This experiment could proceed and produce the same information even if cats do not have sufficient brain complexity to even experience a subjective reality. I love my kitties, and I suspect that they are aware, but I admit that they could be little cuddle seeking, food demanding zombies. That is, they could be biological robots - organisms with no self awareness, simply moving through life according to a very complex and evolving "program" that may produce organized, complex, and developing behavior - but which does not result in consciousness.
If cats were such unaware robots they would still take in visual information through their eyes, they would process that information and it would inform their movements. They would turn their heads and focus on things, they would react to things before them, they would avoid obstacles - they would do all the things associated with vision; but even so that is not evidence that they are 'aware' of what they are seeing.
Think of the many robots out there which people are teaching to "see" and react to visual stimuli. In particular I remember one that was learning to catch a ball (an experiment involving both vision and learning programs). As cool as that robot was (and wow, it was cool), the robot was not aware of what it saw. It was merely a fantastic computer program - an insanely complex mathematical equation - the coordinated the movements of a mechanical arm with information from a video camera. Cats could be an even more advanced version of the same thing.

If I've still got you, let's relax the skepticism a bit. Let's admit that cats are 'aware' in some way. All right, but that awareness is far more than the transmission of data through 177 neurons in the Thalmus. Their awareness, I am confident in assuming, involves much more of the brain. So while we may be able to poke around in Garfield's head to see what visual information his brain processes as part of his awareness, we cannot describe the experience of that processing. All this experiment has done it take the visual information a cat uses and run our own processing on it - "mathematical filters" is the term the blog post uses.

To understand what is going on in a cat’s head we not only need to know what information it is processing, but we also need to know !how! it processes that information. I do not anticipate that there is really an “end-point” to information processed by a brain; so rather than saying that we need to know what is the product of the processing of visual information, I say we need to know what it is like to process that information. I think this is called the “hard problem of consciousness”.

How can we be sure that the “mathematical filters” used in the experiment accurately reflect the way a cat would approach that information. Perhaps visual information, that being data from the eyes, is presented in feline consciousness as an experience that we think of as smells? This sounds far fetched, but that think of that time your …. Uh… “friend” took acid and talked about smelling colors and tasting sounds? Sure, he was still taking in the same information as before – but his processing of that information was now all messed up.

It seems that this experiment does not tell us what “raw experience looks like inside the brain of another being”; rather it takes information that another being uses to create their raw experience and processes it according to mathematical filters which we deem significant.

Sidenote: This experiment does open the possibility of doing something similar with a conscious human and either asking him if the monitor display accords with his own perception (might have questionable utility), or introducing new information to his brain and asking for visual reports on how that affects his experience.

Well said Will. Though, there is really nothing stopping a mathematical algorythm from solving the 'hard problem' of consciousness, and indeed it probably will, this seems to be far from solving the problem. The lateral geniculate nucleus has a retinotopic organization of its cells - that is, the cells that are next to eachother in the LGN process information that is represented next to eachother within the retina, and therefore in space. But as clive points out, that information alone isn't enough to describe what an agent is 'seeing' since the same image can produce entirely different percepts. So really, while this tells us a lot about the integrity of the information in the LGN, it doesn't tell us a whole lot more than that. (NB: I still think this experiment is bloody cool).

"Though, there is really nothing stopping a mathematical algorythm from solving the 'hard problem' of consciousness, and indeed it probably will..."

Agreed, especially as I think that of myself as little more than a mathemtical algorythm (or perhaps more than one) embodied in a magnificently complicated chunk of carbon.

As for your point about the retinotopic organization of the cells of the lateral geniculate nucleus (wow... mouthful) that is interesting. On a gut level it just seems right that with basic cognitive processes, such as vision, the location and structure of a neuron should be tied to its function. Perhaps it's the essential/inherited nature of vision and like functions, we didn't have to learn how to see. (Though perhaps we had to learn about what to think of what we saw.)

On the other hand, when I think about more advanced processes - like all higher level abstract thought - the link between form, function, and the action/event in question seems less important. I think of language and abstract thought as distributed events without direct links to specific cells. Perhaps that's a bit uninformed, but I doubt we'll ever be able to point to a clump of cells and state: "those are the cells representing [green car], and whenever I think of a green car those are the cells that light up."

And, oh yeah - this experiment is cool.

Totally wicked discussion going on here. This is why I love having actual students of neurology onboard the conversation ...

Further to your point, Will, about how a cat could evince all the behaviors of consciousness while really being nothing more than a bundle of very-complex-but-automatic responses -- it's worth pointing out an interesting parallel in artificial-intelligence research. I once talked to a guy who added some A.I. and light-detecting sensors to a little toy car, such that it could guide itself around the room avoiding obstacles. In his first iteration, the car would scan the entire room very slowly, then plot out a clear path to where it wanted to go, and then very carefully and deliberately follow that path. But it was moving too slowly, so he decided to reprogram it with far "less" A.I.: Instead of carefully scanning the entire room, it would just blast off in one direction, and if it detected an obstacle suddenly appearing in front of it, pick a new direction. It did not, like the previous iteration, actually have a mental map of the room in its "mind".

But the upshot was that people thought the latter mechanism appeared more intelligent than the former one, simply because the car's behavior seemed more lively and quirky.

It's a common finding in A.I.: Simpler, arguably "dumber" forms of A.I. can paradoxically appear more intelligent -- at least, they appear more intelligent to us anthropomorphizing humans. So even if the cat doesn't have anything remotely close to consciousness, and even if it's interpreting the visual data arriving at its brain in a completely different fashion than we humans might, we still think it seems conscious.

Kind of a tangent, but mildly related.

I really, really love the idea of tapping into an actual human's LGN, recording the data, then replaying it and seeing how closely it accords to the subject's memory of what he or she "saw".

weeee. These kinds of conversations bring me immense joy.

I just saw this: http://www.the-scientist.com/news/20050912/01 Read this! One could extrapolate from this experiment that the size and timing of the P300 may tell us whether or not someone is conscious of a scene. Though it may be early to make these kinds of conclusions, it is still exciting stuff.