like, without disrespect to faine, i simply don't understand why this would be important, considering that we _verifiably_ do not have any disintermediated access to our own embodiment!
like, without disrespect to faine, i simply don't understand why this would be important, considering that we _verifiably_ do not have any disintermediated access to our own embodiment!
Just read an otherwise pretty decent book critical of ai (mostly on data harvesting grounds) that felt compelled to begin by not only endorsing this "embodied vs not" metaphysics of mind, but accused people who think machine intelligence is possible of "the fallacy of cartesian dualism"
Felt like a bit of a projection tbh lol, even as someone otherwise very sympathetic to the arguments about data harvesting information from databases of prisoners and jail bookings
Embodiment is hugely important! We have direct access to visual, audio, etc stimula of the real world that actually exists. Not simplified "tokens". That's the difference between an actual human reporter, and an AI "reporter" who has to rely on stuff other people have written
"we" do not have direct access to video or audio. our experience of these things is constructed by our brains.
Ego tunnel goes brr
why is this so funny help
I already know about the homunculus argument, that is why it is funny.
Now who's a Cartesian?
i mean, it is absolutely true that, e.g., auditory processing and visual processing operate on fundamentally different timescales, and that the "simulteneity" of our sensorium is a constructed phenomenon. same with the extremely low resolution of our vision, and editing out saccades, etc.
I have just never found the idea that eg saccades mean that our perceptions are not of the world or that we are like observers of an inward show. Tyler Burge went on forever in this vein (against McDowell, in 2005) but it seems like an unjust slide from science to metaphysics
I'm not even sure how to conceptualize the kind of direct perception that we are denied to have on the basis of the fact that we perceive via sense organs (I think that's all you need to get it going; the fancy brains stuff is just extra detail)
Sensory processing happens as early as at the sense organs. There is always processing that doesn’t require the CNS. Source: studied the sensory periphery in grad school.
i think that the first argument -- that the perceived "moment" is constructed from sensory experiences and behaviors that both precede and follow the experience of "making a decision" is pretty troubling.
Well I think that's because people are unreasonably attached to the concept of free will
to be fair, i think that "free will" survives this; it just requires that the idea be applied to a temporal sliding window occupied by a gestalt whose parts act at different speeds.
if i'm reading you right i think i end up somewhere similar; in particular, i think the continuity of the "self" is especially illusory.
I don't! *falls asleep*
We focus a lot on our own interiority, which is fundamentally not comparable to that of others, and less on how we appear to others, which is comparable
The fact that I have about 98% of my email since 1994 in notmuch (I lost a chunk of 1999) is not quite the same as having boxes of diaries but it definitely makes you confront the idea of a continuous self.
just also, if ”we” are observers of an inward show, then are mice that as well, or say mantis shrimp? Recall reading that mice have a predator evasion pattern that is processed in the retina (rapidly approaching object: bad)
i don't mean that there is no verifiable truth, but that we are very much little homunculi who live inside a sensorium which has properties which are convenient for permitting us to react to the real world but have no relation to the way our underlying cognition actually processes its inputs.
What do you think accounts for the contents of our thoughts on this picture? How is it that we are able to think _about_ objects and properties in the world?
Yes there isn't a statement that we aren't homunculi. But, the human-homunculus and the token based LLM homunculus are dramatically different, and if you hypothesize that intelligence is a single thing, the interface to the world could be sufficient to fully explain the differences in behavior
That sensorium is part of me, as much as the homunculus (is that the conscious part of the mind?) The conscious part of the mind is not the whole of me, or separate from the rest of me
With this in mind, no pun intended, what would direct access look like?
Being able to copy that stimula out of our consciousness and examine it externally, I would imagine.
We have pretty damn direct access to what pixel map we're seeing at any given time, more than you'd necessarily think we would
This is not even a little bit true. For starters, it takes 50-300ms for you to consciously register visual stimuli. There is a hole in the middle of your vision you don’t perceive. Your brain deletes the images from when your eyes flit back and forth. You can only see about 5 degrees in focus
Your brain seamlessly stitches the images from the two eyes together. I could go on. There is nothing direct about “our” access to visual data from our eyes
That’s just…completely untrue.
"organic" is totally an implementation detail
And this is coming from somebody who doesn't believe in dualism or even that humans are anything but stochastic parrots ourselves
"Whatever signals make it through the optic nerve" is not really direct or unmediated.
Plus our brains are doing so much interpolation frame to frame, it's a synthetic reality, which matches up pretty well, but is not actually accurate second to second.
Tell me you have no idea how vision works without saying you have no idea how vision works
How does any of this www.ritsumei.ac.jp/~akitaoka/in... make sense in a "we can access the pixel map" version of vision
we can access the light rays that our brain is getting the hacks the brain uses to assemble a single image from what our eyes send over (flipping the image, filling in the blindspots, not great peripheral vision) is what enables these illusions
Yeah, but we also have actuators that can move atoms as well as bits and enough control over our sensors to set up self contained feedback loops that are by all indications more than just solipsistic. That seems significant.
the word "direct" may not be helping here, but I think it is probably very important whether access to things is mediated by a process that involves acting on things [not concepts] in a way that involves feedback on your action. this doesn't require dualism or that "embodiment" be embodiment in meat
last time we went round with this, I more or less convinced myself that the Chinese room doesn't think in Chinese--not because the guy doesn't understand Chinese (he doesn't matter) but because it didn't acquire Chinese concepts by interacting with the things they represent
As I keep saying, everyone should read this: dspace.mit.edu/handle/1721....
And this very recent thing as well: eschwitz.substack.com/p/minimal-au...
Kantian epistemology
Is this how you imagine how the brain works? The mind is separate from the brain, i.e., Cartesian dualism?
You shouldn't have to be David Hume to get why logic, or eidetic pattern matching, can only get you so far.
Citation needed
Saccades, the cochlear amplifier, any optical illusion
i think the thing is that the embodiment of LLMs are GPUs, and basically only i am brave enough to say the Google Data Center is alive
Important for what, exactly? If you mean to say it's cognitively irrelevant, it's pretty much a falsified statement - intermediation or not. But I assume you're meaning something else ?
I genuinely (maybe weakly) believe that this is down to “people who realize that electrons are indistinguishable IN PRINCIPLE” and people who do not. You can imagine one of the electrons is red and the other one is blue, therefore “the red electron” and “the blue electron” are coherent concepts.
[jeff foxworthy] you might be a dualist if