if perception is a brain function, then it has lost its established meaning,
— Mww

(...) Can you cite the established meaning? — Kenosha Kid

That would be whatever says perception means a thing of its own and not a brain function. In taking exception to your rendering, it is not incumbent on me to supply an alternative, but to argue as to how yours doesn’t work. Socratic dialectics, donchaknow.

Case in point, in your entry to ......

None of this is to say I am dubious about the existence of any object that purportedly causes my perception of it. — Kenosha Kid

.....is found no disconnect between saying perception over brain function, insofar as external objects in general actually are sufficient causality both perception and brain function, which implies interchangeability without contradiction. However, in the next.....

Nonetheless, I am proceeding only in extremely high confidence in the hypothesis that my experience of the red flower is caused by an external object with certain properties that cause that experience. — Kenosha Kid

....is found the necessary causality not given in the first, re: certain properties. As these certain properties as assumed to belong to the object, they are not given from brain function, but they are nonetheless perceived as residing in, or appended to, the object as the means of its distinction from objects in general.

This is called direct realism, or epistemological monism.
————

Subsequently, in this...

nothing of the external world is, for instance, in my brain. — Kenosha Kid

....is found that those conditions sustaining epistemological monism are apparently false, insofar as herein it is said there is nothing of those given properties of that object found in the brain. Now we see that it is the case that the loss of, or the non-existence of, the monistic properties of the object in the brain, says absolutely nothing whatsoever about the loss of, or non-existence of, those very same properties given from the perception of them.

The logical deduction from all that is twofold:

The theorem:
Properties never did belong to the object, therefore the loss of them is irrelevant, and, perception of objects is distinct from the brain function with respect to them.

The proof:
Imagine yourself a dendrite, just under the skin. A very young dendrite, perhaps an infant dendrite. At least inexperienced. You know what your job is, but haven’t yet had a chance to impress the boss. You know, His Esteemed Grayness sitting all lofty up there, surrounded by bone and long flowing locks. Suddenly, this protrusion of unknown origin (nod to Eric Bloom, aka, Blue Oyster Cult) slaps you right in the dentrical face. YEA!!! You finally....AT LAST....get to send a message. You’ve trained for this since mamma met papa, but you’re cool....you do your job.

First......What do you tell the boss?
Second......do you, as dendrite, really expect the boss to send a return message saying.....thanks, but I already knew all about it.
Third....if the second becomes the case, wouldn’t you wonder, in your dendrite manner.....WTF am I doing here then????
(Then you get all depressed and pathologically stupid and shoot yourself with an overdose of calcium ions, bursting yourself, taking your cell body with you. Your neighbor dendrites look at each other and wonder as well.....he’s got a point. What are we doing here anyway, if the boss already knows what we send him.)

The intent here is to ask....what is NOT sent upstream?

Yep, I agree. If it seems you are looking at a red apple or listening to a piece of music, you are looking at a red apple or listening to music.

But then let's say you're not, what you thought was a red apple was a red ball and what you thought was music was birds chirping. It's still the case that you experienced these things are a red apple and as music, even if it turns out what was there was something different from what you initially perceived.

The "seeming" language is used as an excuse to get rid of some hard problems, like having experience. Easier to do biology if you can get rid of experience theoretically.

We're probably going far too deep into what is essentially a complaint that the term 'sense-datum' has already been taken and can't, for weird reasons, be transported between theories, or between a theory and a more descriptive discussion, leading to a proliferation of terminology and an inability to define clearly where one ends and another begins. — Kenosha Kid

There will probably be 'analogues' of sense data in some indirect realist and some antirealist styles of thinking about perception. There probably won't be in direct perception accounts - embodied cognition/active inference (I think) tends to side with there not being any object like sense data. Dennett also isn't a sense data theorist. Ultimately that isn't because either way of rejecting sense data as a concept rejects the notion of brain states representing world states by some coupling process, it's that sense data as a concept itself stakes out a claim regarding the process that couples mind and world. If you have an entity like sense data or qualia kicking around in your perceptual theory, that will either say things about your theory or about how you view perceptual theories in general. It says things about the abstract lens through which you view perception.

leading to a proliferation of terminology and an inability to define clearly where one ends and another begins. — Kenosha Kid

Exactly. Using sense data as you've been doing contributes to the mess. Unless of course you're actually subscribing to or think in rough accord with sense data theories of perception. It seems to me that you think in sufficient accord with sense data theories that you're happy treating the concept as transportable between perceptual theories, as if what a 'sense datum' allegedly refers to is a theory neutral posit.

An example would be a more complete theory that can account for the existence or illusion of sense data and qualia. This would be as opposed to one that shows that such ideas are purely artefacts of bad stroky bearding. — Kenosha Kid

There's some phenomenology which does try to account for the illusion. You have MP's discussion of raw sensory data in Phenomenology of Perception, Heidegger's discussion of propositional as-structures in... Basic Problems or Logic the Question of Truth? Even Dennett's Cartesian Theatre concept takes aim at sense data, which is roughly the substrate that qualia predicate (sense data bears qualia), sense data is the video reel for the Cartesian Theatre.

Edit: I'm sure there's others!

If you take MP to be suggesting that either qualia or sense data are illusions, you're using some very limited definition of those terms that might coincide with Dennett's usage.

Doing that just magnifies strawmen.

If you take MP to be suggesting that either qualia or sense data are illusions — frank

I ain't. AFAIK he has a pretty realist take towards phenomena, but doesn't take a realist stance toward qualia and their 'experiential chunk' substrates. He has an issue with colour as a property of an experience, not with phenomena and colour.

Good point but the nonphysicalist has to admit s/he can't explain qualia in nonphysical terms and now we're in neti neti territory, Where do we go from here, sir/madam? — Agent Smith

Take it as a given that we see colors, hear sounds, feel pains and see how to fit that with the rest of our understanding of the world.

Well the senses themselves don't cogitate. So there's no puzzle by itself here. — Manuel

What use are the senses without cogitation? What use is cogitation without senses? Brains evolved later from nerve nets. Feelings existed before integrating them into the whole of the brain.

I said "reading", not "watching videos". I come here to discuss with fellow well-read members, Harry, not to teach anyone what they can learn themselves. There's just too much 'idle (uninformed) speculation driven by intellectual laziness' going on lately. — 180 Proof

Not sure how reading what other people write can shed light on your own mind. Seems to me that the mind is fundamental and anyone that has one can reflect on its properties themselves without being influenced by what other people write. I come here to discuss with fellow free-thinkers that can think for themselves and come to their own conclusions, and not only about what other people write.

Source regarding MP's direct realism:

When I perceive a thing such as a fireplace, it is not the concordance of its various
appearances that leads me to believe in the existence of the fireplace as the geometrical plan
and common signification of all of these perspectives. On the contrary, I perceive the thing
in its own clarity (PP 191).

The constancy of colour is merely an abstract moment of the constancy of things, and the
constancy of things is established upon the primordial consciousness of the world as the
horizon of all our experiences (PP 326)

Source regarding MP's disbelief in qualia and sense data:

At the outset of the study of perception, we find in language the notion of sensation, which seems immediate and obvious: I have a sensation of redness, of blueness, of hot or cold. It will, however, be seen that nothing could in fact be more confused, and that because they accepted it readily, traditional analyses missed the phenomenon of perception (PP50)

Pure sensation will be the experience of an undifferentiated, instantaneous, dotlike impact. It is unnecessary to show, since authors are agreed on it, that this notion corresponds to nothing in our experience, and that the most rudimentary factual perceptions that we are acquainted with, in creatures such as the ape or the hen, have a bearing on relationships and not on any absolute terms. (PP51)

IE, we don't experience qualia. Under what conditions could we experience a quale?

The pure quale would be given to us only if the world were a spectacle and one’s own body a mechanism with which some impartial mind made itself acquainted.3 Sense experience, on the other hand, invests the quality with vital value, grasping it first in its meaning for us, for that heavy mass which is our body, whence it comes about that it always involves a reference to the body (PP176).

I ain't. AFAIK he has a pretty realist take towards phenomena, but doesn't take a realist stance toward qualia and their 'experiential chunk' substrates. He has an issue with colour as a property of an experience, not with phenomena and colour. — fdrake

You're loading the term "qualia" with some sort of absolute independence. MP would say that nothing that's phenomenally accessible to us has that status.

If you insist that qualia must be this odd metaphysically independent entity, you're making the word useless for yourself and those who use it that way.

It remains an innocuous way to talk about the phenomenal character of consciousness for everyone else.

You're loading the term "qualia" with some sort of absolute independence. MP would say that nothing that's phenomenally accessible to us has that status.

If you insist that qualia must be this odd metaphysically independent entity, you're making the word useless for yourself and those who use it that way.

It remains an innocuous way to talk about the phenomenal character of consciousness for everyone else. — frank

Source?

IE, we don't experience qualia. Under what conditions could we experience a quale? — fdrake

MP is telling you that perception is organized around ideas. The fireplace is an idea, not something arising from visual data.

If we don't agree on that, we're probably at a deadend.

frank

Source — fdrake

For what? That MP says perception can't be atomized?

For what? — frank

For these:

MP is telling you that perception is organized around ideas. The fireplace is an idea, not something arising from visual data. — frank

MP would say that nothing that's phenomenally accessible to us has that (absolute independence) status. — frank

What use are the senses without cogitation? What use is cogitation without senses? Brains evolved later from nerve nets. Feelings existed before integrating them into the whole of the brain. — Harry Hindu

I suppose a very simple species would avoid pain sensations. An ingrained habit from evolution. Unless one would postulate a cognition from the most basic organic entities, which, who knows?

What good is cogitation without the senses? Well, not entirely senseless, but look at, say, deaf-blind people, they can read by only pressing their fingers over bumps on a page and get an extremely rich story out of that.

So the senses can be extremely poor compared with the cognitive reply.

I don't know what kind of source you're looking for. It's explicit in the quote you provided.

I'm wondering what in the world you think that quote means?

What do you think he means by "pure" qualia?

I don't know what kind of source you're looking for. It's explicit in the quote you provided. — frank

Which quote?

What do you think he means by "pure" qualia?

If you answer that it might help me understand how you're interpreting him.

How on earth do you not see that he's saying that perception is infused with ideas?

It's a unit of pure sensation. Which is a category he undermines repeatedly. Pure sensation I take as meaning intrinsic, nonrelational properties of (hypothetical/criticised) sensory units. Extending the previous P176 quote:

“The light of a candle changes its appearance for a child when, after a burn, it stops attracting the child’s hand and becomes literally repulsive.2 Vision is already inhabited by a meaning (sens) which gives it a function in the spectacle of the world and in our existence. The pure quale would be given to us only if the world were a spectacle and one’s own body a mechanism with which some impartial mind made itself acquainted.3 Sense experience, on the other hand, invests the quality with vital value, grasping it first in its meaning for us, for that heavy mass which is our body, whence it comes about that it always involves a reference to the body. (P176 quote extended)

The pure quale would only arise if abstracted from the practical context of experience and the perceptual relationship a body has with its environment. Purity is construed as a lack of investment of the property in an practical or meaningful context - nonrelationality - specifically with a body - being intrinsic.

With the construal that intrinsic as meaning "deriving wholly from internal processes or its own constitution" and nonrelational means "properties that apply only to the object and are one place predicates". The red of the fire in the hearth is welcoming, the red of a house fire is not. MP is articulating a unity between the welcoming feeling and red and the panic feeling and red (and the broader experience they're embedded in).

He very clearly dunks on pure sensation, and construes qualia as the experiential properties of pure sensations - and I agree with you that he dunks on the atomisation of experience (into sense units that bear properties...).

“The perceptual ‘something’ is always in the middle of something else, it always forms part of a ‘field’. A really homogeneous area offering nothing to be cannot be given to any perception. The structure of actual perception alone can teach us what perception is. The pure impression is, therefore, not only undiscoverable, but also imperceptible and so inconceivable as an instant of perception. If it is introduced, it is because instead of attending to the experience of perception, we overlook it in favour of the object perceived. A visual field is not made up of limited views. But an object seen is made up of bits of matter, and spatial points are external to each other. An isolated datum of perception is inconceivable, at least if we do the mental experiment of attempting to perceive such a thing. But in the world there are either isolated objects or a physical void.”

“I shall therefore give up any attempt to define sensation as pure impression. Rather, to see is to have colours or lights, to hear is to have sounds, to sense (sentir) is to have qualities. To know what sense-experience is, then, is it not enough to have seen a red or to have heard an A? But red and green are not sensations, they are the sensed (sensibles), and quality is not an element of consciousness, but a property of the object. Instead of providing a simple means of delimiting sensations, if we consider it in the experience itself which evinces it, the quality is as rich and mysterious as the object, or indeed the whole spectacle, perceived. This red patch which I see on the carpet is red only in virtue of a shadow which lies across it, its quality is apparent only in relation to the play of light upon it, and hence as an element in a spatial configuration. ”

(P53/P54, bold and italics are mine)

Red and green are properties of "the object", what is the object? The fire. The 'perceptual something in the middle of something else' is a body's perceptual relationship inhabting the world. And its red cannot be thought of as part of an 'isolated datum', since such things are 'imperceptible' - ie, they are not perceptions, whatever the conception of red is that attributes it to an isolated datum (red quale to sense datum of fire) according to MP it cannot be perceived. IE, it doesn't form part of his perceptual theory.

(@Kenosha Kid because they might enjoy MP talking about homogenous qualities as unrealistic, like the red wall thought experiment KK wrote maybe?)

Analysis, then, discovers in each quality meanings which reside in it. It may be objected that this is true only of the qualities which form part of our actual experience, which are overlaid with a body of knowledge, and that we are still justified in conceiving a ‘pure quality’ which would set limits to a pure sensation. But as we have just seen, this pure sensation would amount to no sensation, and thus to not feeling at all. The alleged self-evidence of sensation is not based on any testimony of consciousness, but on widely held prejudice. We think we know perfectly well what ‘seeing’, ‘hearing’, ‘sensing’ are, because perception has long provided us with objects which are coloured or which emit sounds. When we try to analyse it, we transpose these objects into consciousness. We commit what psychologists call ‘the experience error’, which means that what we know to be in things themselves we immediately take as being in our consciousness of them. We make perception out of things perceived.

(@Srap Tasmaner & @Isaac might enjoy that quote because of how it bears on the use of the word model... do we use models to see, or do we see models?)

This is from the introduction, it's a ground clearing that takes place to orient MP's readers that no, he isn't talking about experience and its properties of distinct experience chunks as people seem to. In fact he articulates holist and embodied/enactive intuitions that undermine those concepts.

Can you source your interpretation now please, frank?

How on earth do you not see that he's saying that perception is infused with ideas? — frank

I don't. MP's a direct realist in the sense that the environment of body is the direct object of that body's perception. The 'infusion of ideas' occurs in perception, which is a relationship between that body and its environment. Because there's no intermediary or middle man, perception is in direct contact with the world - the imagining and the seeing and the actions and the feelings all get bunged together as components of a direct contact, rather than having perception itself only operate upon sensory inputs, perception itself contains the processes by which a body elicits and organises its sensory input.

Take it as a given that we see colors, hear sounds, feel pains and see how to fit that with the rest of our understanding of the world. — Marchesk

Indeed, we do see colors, hear sounds, feel pain. You speak as if these are distinguishing features of a nonphysical mind. Let's see what the next coupla decades of neuroscience research tell us. Hope I don't die (too soon).

It's a unit of pure sensation. Which is a category he undermines repeatedly. Pure sensation I take as meaning intrinsic, nonrelational properties of (hypothetical/criticised) sensory units. — fdrake

Absolutely. Why were you disagreeing with me earlier? That's exactly what I was saying. MP is saying that a black dot, for instance, can't be perceived in isolation. There has to be something non-black around it.

This is an insight much older than MP. It at least goes back to Hegel. MP is just taking the time to apply the principle to perception. So if he's identifying an illusion, it's that a bit of perception could be independent.

To now claim that he's saying that qualia are illusions, you'd have to make this independence a necessary feature of the concept of qualia.

It most definitely isn't.

I have now repeated exactly what I told you before you started asking me for sources. *pulls trigger on invisible loaded gun at temple*

Hope you have a good day. Stay warm.

To now claim that he's saying that qualia are illusions, you'd have to make this independence a necessary feature of the concept of qualia. — frank

Can you please give a source that spells out a candidate holistic conception of qualia that you're gesturing towards?

Can you please give a source that spells out a candidate holistic conception of qualia that you're gesturing towards? — fdrake

I don't understand. What's a holistic conception of qualia?

What's a holistic conception of qualia? — frank

A conception of qualia which isn't the name for the isolated 'red' of the fire, or the 'red-welcoming' mixed quale shorn from its generative environment. Which alternative conception of qualia to the one you imagined I was attacking did you have in mind, and can you give me a paper that talks about it?

MP is saying that a black dot, for instance, can't be perceived in isolation. There has to be something non-black around it.

This is an insight much older than MP. It at least goes back to Hegel. MP is just taking the time to apply the principle to perception. So if he's identifying an illusion, it's that a bit of perception could be independent. — frank

You’re leaving out the temporal , or diachronic aspect of MP’s model. figures must appear against backgrounds, yes. But the whole figure-background gestalt ensemble changes with each new act of perception.

Merleau-Ponty explains that to attend to any experience is not merely to shine a neutral light on it, but to articulate a new sense, the ‘active constitution of a new object'. It is to identify a new figure and in doing so, to transform the sense of the previous figure along with its background.

“Attention, therefore, as a general and formal activity, does not exist.” Rather than there being a general capacity for neutral observation, a universal kind of attention necessary for any moment of consciousness, “it is literally a question of creation. “ “Attention is “a change of the structure of consciousness, the establishment of a new dimension of experience, the setting forth of an a priori... To pay attention is not merely further to elucidate pre-existing data, it is to bring about a new articulation of them by taking them as figures”.

This is the first definition in the SEP article on qualia:

"Qualia as phenomenal character. Consider your visual experience as you stare at a bright turquoise color patch in a paint store. There is something it is like for you subjectively to undergo that experience. What it is like to undergo the experience is very different from what it is like for you to experience a dull brown color patch. This difference is a difference in what is often called ‘phenomenal character’. The phenomenal character of an experience is what it is like subjectively to undergo the experience. If you are told to focus your attention upon the phenomenal character of your experience, you will find that in doing so you are aware of certain qualities. These qualities — ones that are accessible to you when you introspect and that together make up the phenomenal character of the experience are sometimes called ‘qualia’. C.S. Peirce seems to have had something like this in mind when he introduced the term ‘quale’ into philosophy in 1866 (1866/1982, para 223)."

The SEP doesn't address whether the phenomenal character of experience is atomic or holistic. We start with this basic idea and then decide whether we can pull the elements of perception apart (whether we can know about a black dot extracted from it's environment).

In the same way, if we're talking about personal identity, we just start with the idea. If we later decide that it's best to talk about Dasein instead of extracting selfhood out into an Aristotelian substance blob, that's fine. We haven't done away with selfhood, just the INDEPENDENCE of it.

I'm actually pretty frustrated at this point.

Yep. I agree.

"Qualia as phenomenal character. Consider your visual experience as you stare at a bright turquoise color patch in a paint store. There is something it is like for you subjectively to undergo that experience. What it is like to undergo the experience is very different from what it is like for you to experience a dull brown color patch. This difference is a difference in what is often called ‘phenomenal character’. The phenomenal character of an experience is what it is like subjectively to undergo the experience. If you are told to focus your attention upon the phenomenal character of your experience, you will find that in doing so you are aware of certain qualities. These qualities — ones that are accessible to you when you introspect and that together make up the phenomenal character of the experience are sometimes called ‘qualia’. C.S. Peirce seems to have had something like this in mind when he introduced the term ‘quale’ into philosophy in 1866 (1866/1982, para 223)." — frank

I mean I know what qualia are.

The canonical example there is the colour of a coloured patch, and they're 'available to introspection', if you read through PP that's exactly the framing MP is taking shots at. You have an isolated stimulus as a canonical example rather than a gestalt one, a focus on introspective scission as a fuel for quale examples (the qualities that 'together make up the phenomenal character of the experience' - together implying an individuating/atomising introspective operation to split the fire's red from the fire) rather than pre-introspective unity (all the figure ground examples, the shadow on a red example).

I thought you would be able to provide me with a relational account of individual qualia, or at least a reference to it. If there is indeed a relational account of individual qualia and they aren't born by an underlying 'sensory datum', I'd be more inclined to agree they're consistent with MP's points.

information more or less unconsciously ascends through areas responsible for particular processes such as recognition of lines, shapes, positions, objects etc. in the visual system — Enrique

...with you so far...

then somehow impinges upon a specially adapted neural network's CEMI field where percepts (if you don't want to talk about the quantum underpinnings, I won't get into it) are ultrasychronized on a relatively large scale via phase locking to contribute towards domains of the perceptual field which of course extensively integrate via synesthesia-like mechanisms. — Enrique

...eh?

I have trouble discerning where these CEMI fields might be located and thought perhaps you could have some ideas... — Enrique

...none, I'm afraid.

...once you're familiar with the theory. — Enrique

Do you have a citation I could look into?

It seems a bit of a stretch to me to say the brain effectively contains images from the methods in the study. — fdrake

I agree. I was quite careful to be circumspect about the degree of progress along the road.

I think my grumblings would derail the thread. Nevertheless I've put them in this hidden box. — fdrake

I can't do the hidden box thing, so people will just have to lump it. Picking apart the methodology is as least as interesting as the actual topic.

( 1 ) since the model linking FMRI signals and the extracted feature from the layers doesn't seem to have a neural mechanism associated with it, the overall algorithm run doesn't have a demonstrated 'port to the wetware', so to speak. It doesn't seem established to run in the brain. I think it's thus evidence for the weaker claim that 'it's possible to reconstruct some images from fmri signals' rather than 'fmri signals encode images in a way similar to what is portrayed in the paper' — fdrake

Yes. Agreed. I hope I wasn't read as making such a claim, I'm indicating a direction I think the research is going rather than claiming a milestone has been reached.

The way I see it, if a DNN can 'learn' to decode fMRI scans into something - anything - then it's proof of principle (which is all I think the authors were going for) for a method by which we can develop candidate models. We obviously then need to actually test those candidates.

( 2 ) The subjective appraisal procedure for accuracy had a strange design and metric:

For the subjective assessment, we conducted a behavioral experiment with another group of 9 raters (4 females and 5 males, aged between 19 and 36 years). On each trial of the experiment, the raters viewed a display presenting a reconstructed image (at the bottom) and two candidate images (at the top; its original image and a randomly selected image), and were asked to select the image similar to the one presented at the bottom from the two candidates. Each trial continued until the raters made a response. For both types of assessments, the proportion of trials, in which the original image was selected as more similar one was calculated as a quality measure.


It measures which of two presented images was 'more similar' (subjectively) to the subjects (an ordinal value) and then the number of agreements was presented as a % accuracy. From the set up they described:

In both objective and subjective assessments, each reconstructed image was tested with
all pairs of the images among the same types of images (natural-images,
geometric-shapes, and alphabetical-letters for images from the image presentation
sessions, and natural-images and geometric-shapes for images from the imagery
session ; e.g., for the test natural-images, one of the 50 reconstructions was tested with
49 pairs consisted of one original image and another image from the rest of 49, resulting
in 50 × 49 = 2,450 comparisons).


Chance is 50% accuracy. Effectively this is a simulation of whether a human could relabel the image generated into the training data corresponding to the type of the original image -eg, inferred lion features with lions when given a single alternative. Considering the pixel crosscorrelation of test features and images was reported as 66%, even mildly increasing the number of comparison images (at the expense of 'complete cases' of comparisons) could very well undermine the claim to 95% accuracy.

If you look at the images, you can do a fair guess of which image is in which labelled category from just the background and colour space (lions are kinda yellow). Willing to bet the similarity is a priming effect of seeing the images on the same screen rather than labelling the DNN's feature as a perceptual type. Eg which of the reconstructed images of subject 2 is a lion and which is a mouse without knowing which is which beforehand?

I'm sure there's an argument there that perceptual feature visual content is not the same thing as an inferred label of the perceptual feature visual content - but that's an argument which should not have to happen. Should've been taken away in the controls of the subjective experiment - or run another to see if people labelling images with categories (is this brown smudge a lion or a mouse?) produces analogous accuracy measurements (and we all know it wouldn't based on the sample reconstructed images). — fdrake

I hadn't picked up on that, it is weird. Other approaches don't use this method - see Rosca for example https://arxiv.org/pdf/1706.04987.pdf which is sort of a seminal paper of this approach (for me anyway - I'm not necessarily up to date with this) Rosca used a dual network (adversarial and variational) which I thought was a standard training methodology.

( 3 ) The experimental design is there to generate test and training data for the neural network, insert bucket of ecological validity concerns here. A person's brain processing is as much devoted to a single image at a time as is possible and they are stationary. — fdrake

Yes, but given the relative isolation of hierarchically separate cortices I don't think this would be too disruptive to a generalisation of the results. The regions being examined are very specialised and specific to their tasks. It's not that I'm unsympathetic to the ecological validity concerns, only that I think they're not big enough here to get much in the way.

Of course, if anyone were to make any claims about AI 'reading our minds' resulting from this, that would be utterly ridiculous. But what kind of world would we live in where media outlets would be stupid enough to do that!

https://time.com/5874444/science-of-nightmares/

Oops!

( 4 )

For test datasets, fMRI samples corresponding to the same stimulus or imagery were
averaged across trials to increase the signal-to-noise ratio of the fMRI signals. To
compensate for a possible difference of the signal-to-noise ratio between training and
test samples, the decoded features of individual DNN layers were normalized by
multiplying a single scalar so that the norm of the decoded vectors of individual DNN
layers matched with the mean norm of the true DNN feature vectors computed from
independent 10,000 natural images. Then, this norm-corrected vector was subsequently
provided to the reconstruction algorithm. See Supplementary Methods for details of the
norm-correction procedure.


Gives me the willies - is it normal to manipulate the test data in a manner you didn't do to the training data? Effectively what's been inferred on is the average FMRI space-time series, but the model was fit on non-averaged ones. At what point would that decision be made? Is it standard? Did I misread it? — fdrake

I think that was the point. They wanted to be able to test the network performance by looking at the difference. I don't know how 'standard' it is, but I've seen the approach before. Here for example
https://www.frontiersin.org/articles/10.3389/fncom.2021.754587/full (where you might prefer the lack of priming opportunities!)

To be honest, I could have picked a much better paper to give the impression I wanted to give, but it was late and that one was the first on my Zotero!