Matthew Sheffield
@matthew.flux.community
🟦 Writer, researcher, podcaster
🟦 As seen in @flux.community, NYT, WaPo, NBC, Variety ...
🟦 Post about politics, cognitive science, gender, media
New book: What Republicans Know https://flux.community/book/what-republicans-know
🟦 As seen in @flux.community, NYT, WaPo, NBC, Variety ...
🟦 Post about politics, cognitive science, gender, media
New book: What Republicans Know https://flux.community/book/what-republicans-know
You have just stated the argument of the thread and essay.
I am against biological essentialism.
I am against biological essentialism.
January 31, 2026 at 11:47 PM
You have just stated the argument of the thread and essay.
I am against biological essentialism.
I am against biological essentialism.
Ah man, my sarcasm filter wasn't working there.
January 31, 2026 at 11:45 PM
Ah man, my sarcasm filter wasn't working there.
It's about how your mind corrects for the external lighting. The actual dress is blue and black.
I discuss how the perceptions differ in the essay, but here's an article about just that: www.livescience.com/50842-dress-...
I discuss how the perceptions differ in the essay, but here's an article about just that: www.livescience.com/50842-dress-...
Science of 'the Dress': Why We Confuse White & Gold with Blue & Black
Three new studies reveal the science behind "the dress,” and what it says about color perception.
www.livescience.com
January 31, 2026 at 11:19 PM
It's about how your mind corrects for the external lighting. The actual dress is blue and black.
I discuss how the perceptions differ in the essay, but here's an article about just that: www.livescience.com/50842-dress-...
I discuss how the perceptions differ in the essay, but here's an article about just that: www.livescience.com/50842-dress-...
Yeah, that's a big problem.
My current working definition of consciousness is something like: A behavioral capability to autonomously determine meaning, context, and intent through internal somatic and abstract reasoning processes without requiring external protocols or instincts.
My current working definition of consciousness is something like: A behavioral capability to autonomously determine meaning, context, and intent through internal somatic and abstract reasoning processes without requiring external protocols or instincts.
January 31, 2026 at 11:15 PM
Yeah, that's a big problem.
My current working definition of consciousness is something like: A behavioral capability to autonomously determine meaning, context, and intent through internal somatic and abstract reasoning processes without requiring external protocols or instincts.
My current working definition of consciousness is something like: A behavioral capability to autonomously determine meaning, context, and intent through internal somatic and abstract reasoning processes without requiring external protocols or instincts.
Yes it's a huge problem. Being more precise on that term is one of the things I'm trying to do in the essay.
January 31, 2026 at 11:04 PM
Yes it's a huge problem. Being more precise on that term is one of the things I'm trying to do in the essay.
It's not fully public what he is doing, and it may be different now.
He had been talking about incorporating some aspects of "good old-fashioned AI" with transformer architectures. It's definitely going to be better, but yes, we need deictic robots paired to LLMs, and more.
He had been talking about incorporating some aspects of "good old-fashioned AI" with transformer architectures. It's definitely going to be better, but yes, we need deictic robots paired to LLMs, and more.
January 31, 2026 at 10:56 PM
It's not fully public what he is doing, and it may be different now.
He had been talking about incorporating some aspects of "good old-fashioned AI" with transformer architectures. It's definitely going to be better, but yes, we need deictic robots paired to LLMs, and more.
He had been talking about incorporating some aspects of "good old-fashioned AI" with transformer architectures. It's definitely going to be better, but yes, we need deictic robots paired to LLMs, and more.
Yann LeCun is who you are likely thinking of. He's on a better track, but really what you have to do is create embodied systems rather than hard-coded descriptions of how things interrelate. He'll have more success, but it will not be quite what I'm describing.
January 31, 2026 at 10:41 PM
Yann LeCun is who you are likely thinking of. He's on a better track, but really what you have to do is create embodied systems rather than hard-coded descriptions of how things interrelate. He'll have more success, but it will not be quite what I'm describing.
Yes neuroplasticity is a very important piece of evidence against neurodeterminism.
January 31, 2026 at 10:28 PM
Yes neuroplasticity is a very important piece of evidence against neurodeterminism.
There's a lot more in the essay, with lots of arguments, examples, and explorations of other conceptions of minds, plus trips to Mary's Room and the Chinese Room as well.
Please read, share, and subscribe. Thanks!
/end
Please read, share, and subscribe. Thanks!
/end
It’s like this: Why perceptions are our realities
Minds do not create experience, experience creates minds
plus.flux.community
January 31, 2026 at 10:27 PM
There's a lot more in the essay, with lots of arguments, examples, and explorations of other conceptions of minds, plus trips to Mary's Room and the Chinese Room as well.
Please read, share, and subscribe. Thanks!
/end
Please read, share, and subscribe. Thanks!
/end
A truly conscious AI entity will have these attributes. No amount of compute and no amount of data can be generative of full minds. Meta-deixis cannot replace somatic deixis.
Selfhood is self-made. Intelligence is not a ghost 𝘪𝘯 the machine, it is a ghost 𝘰𝘧 the machine.
Selfhood is self-made. Intelligence is not a ghost 𝘪𝘯 the machine, it is a ghost 𝘰𝘧 the machine.
January 31, 2026 at 10:26 PM
A truly conscious AI entity will have these attributes. No amount of compute and no amount of data can be generative of full minds. Meta-deixis cannot replace somatic deixis.
Selfhood is self-made. Intelligence is not a ghost 𝘪𝘯 the machine, it is a ghost 𝘰𝘧 the machine.
Selfhood is self-made. Intelligence is not a ghost 𝘪𝘯 the machine, it is a ghost 𝘰𝘧 the machine.
Current AI systems are just abstract token shufflers. Their internalities exist in a disembodied state that has no persistence. They cannot touch or learn from the world. They cannot designate and adjudicate what things are like to them.
This is why they are not conscious.
This is why they are not conscious.
January 31, 2026 at 10:24 PM
Current AI systems are just abstract token shufflers. Their internalities exist in a disembodied state that has no persistence. They cannot touch or learn from the world. They cannot designate and adjudicate what things are like to them.
This is why they are not conscious.
This is why they are not conscious.
Bottom line: perception is constructed; minds are dynamic execution states; "felt meaning" comes from embodied reference-making.
Consciousness is a type of experiencing. It's an extra layer on top of somatic reasoning, sentience (proto-awareness of body and others), and selfhood (I-vs-Not-I).
Consciousness is a type of experiencing. It's an extra layer on top of somatic reasoning, sentience (proto-awareness of body and others), and selfhood (I-vs-Not-I).
January 31, 2026 at 10:24 PM
Bottom line: perception is constructed; minds are dynamic execution states; "felt meaning" comes from embodied reference-making.
Consciousness is a type of experiencing. It's an extra layer on top of somatic reasoning, sentience (proto-awareness of body and others), and selfhood (I-vs-Not-I).
Consciousness is a type of experiencing. It's an extra layer on top of somatic reasoning, sentience (proto-awareness of body and others), and selfhood (I-vs-Not-I).
The underlying hardware was the same. The software was almost the same, but the execution state started acting bizarre.
Even if you somehow could magically digitally encode everything that you know and believe, mind uploading could not work unless you also rebuilt your entire cellular structure.
Even if you somehow could magically digitally encode everything that you know and believe, mind uploading could not work unless you also rebuilt your entire cellular structure.
January 31, 2026 at 10:23 PM
The underlying hardware was the same. The software was almost the same, but the execution state started acting bizarre.
Even if you somehow could magically digitally encode everything that you know and believe, mind uploading could not work unless you also rebuilt your entire cellular structure.
Even if you somehow could magically digitally encode everything that you know and believe, mind uploading could not work unless you also rebuilt your entire cellular structure.
The YouTuber PokeTips Mike showed the difference between software, hardware, and execution states in this video.
He took an execution state (a "save file") from Let's Go Eevee and loaded it into Let's Go Pikachu. Hijinks ensued.
He took an execution state (a "save file") from Let's Go Eevee and loaded it into Let's Go Pikachu. Hijinks ensued.
What Happens if you SWITCH Save Files in Pokemon Lets Go Pikachu And Eevee?
YouTube video by PokeTips
youtu.be
January 31, 2026 at 10:22 PM
The YouTuber PokeTips Mike showed the difference between software, hardware, and execution states in this video.
He took an execution state (a "save file") from Let's Go Eevee and loaded it into Let's Go Pikachu. Hijinks ensued.
He took an execution state (a "save file") from Let's Go Eevee and loaded it into Let's Go Pikachu. Hijinks ensued.
This reconception of cognition is very different from what you've often heard. An analogy from video games can illustrate the distinctions.
Pokemon games often are released in co-branded versions with the same game but slightly different characters available.
Pokemon games often are released in co-branded versions with the same game but slightly different characters available.
January 31, 2026 at 10:21 PM
This reconception of cognition is very different from what you've often heard. An analogy from video games can illustrate the distinctions.
Pokemon games often are released in co-branded versions with the same game but slightly different characters available.
Pokemon games often are released in co-branded versions with the same game but slightly different characters available.
Cognition is correlation of subagents. It is a process, not a thing.
This really important to grasp because far too many people think of a mind as something similar to software. Instead, a mind is an execution state, it is what your cells are doing together. Death is when they stop.
This really important to grasp because far too many people think of a mind as something similar to software. Instead, a mind is an execution state, it is what your cells are doing together. Death is when they stop.
January 31, 2026 at 10:21 PM
Cognition is correlation of subagents. It is a process, not a thing.
This really important to grasp because far too many people think of a mind as something similar to software. Instead, a mind is an execution state, it is what your cells are doing together. Death is when they stop.
This really important to grasp because far too many people think of a mind as something similar to software. Instead, a mind is an execution state, it is what your cells are doing together. Death is when they stop.
The somatic token of the sensation is the cognitive representation of what the stimulus is and what it is like.
Somatic tokens are qualia, and they are the basis of cognition rather than its product. They are how minds are made, not an extra sparkle that comes afterwards.
Somatic tokens are qualia, and they are the basis of cognition rather than its product. They are how minds are made, not an extra sparkle that comes afterwards.
January 31, 2026 at 10:18 PM
The somatic token of the sensation is the cognitive representation of what the stimulus is and what it is like.
Somatic tokens are qualia, and they are the basis of cognition rather than its product. They are how minds are made, not an extra sparkle that comes afterwards.
Somatic tokens are qualia, and they are the basis of cognition rather than its product. They are how minds are made, not an extra sparkle that comes afterwards.
Software developers making phone apps very commonly use rotation data to move characters. They don't need to know the raw physics of how gyroscopes work or what proof masses are.
This is how thinking works. As cells scale up their outputs, only the somatic token of cellular experience is left.
This is how thinking works. As cells scale up their outputs, only the somatic token of cellular experience is left.
January 31, 2026 at 10:17 PM
Software developers making phone apps very commonly use rotation data to move characters. They don't need to know the raw physics of how gyroscopes work or what proof masses are.
This is how thinking works. As cells scale up their outputs, only the somatic token of cellular experience is left.
This is how thinking works. As cells scale up their outputs, only the somatic token of cellular experience is left.
What builds somatic-native minds are cognitive subagents that have direct experience of externality. Our cells literally experience physics and chemistry. Their collective pooling of outputs into cognitive processes is what minds are.
It's also why phenomenal experience seems mysterious.
It's also why phenomenal experience seems mysterious.
January 31, 2026 at 10:16 PM
What builds somatic-native minds are cognitive subagents that have direct experience of externality. Our cells literally experience physics and chemistry. Their collective pooling of outputs into cognitive processes is what minds are.
It's also why phenomenal experience seems mysterious.
It's also why phenomenal experience seems mysterious.
Some philosophers and neuroscients like John Searle and @anilseth.bsky.social believe that biology is essential to full cognition. But this is not quite right.
Biology is currently the only substrate that can generate minds, but this isn't a metaphysical barrier, it's a capability one.
Biology is currently the only substrate that can generate minds, but this isn't a metaphysical barrier, it's a capability one.
January 31, 2026 at 10:15 PM
Some philosophers and neuroscients like John Searle and @anilseth.bsky.social believe that biology is essential to full cognition. But this is not quite right.
Biology is currently the only substrate that can generate minds, but this isn't a metaphysical barrier, it's a capability one.
Biology is currently the only substrate that can generate minds, but this isn't a metaphysical barrier, it's a capability one.
The scaling upward of cellular somatic deixis across tissues, organs, and systems is how minds are made. And we know via basal cognition research that all cells communicate through gap junctions. Neurons are just really specialized communicators and rememberers.
January 31, 2026 at 10:14 PM
The scaling upward of cellular somatic deixis across tissues, organs, and systems is how minds are made. And we know via basal cognition research that all cells communicate through gap junctions. Neurons are just really specialized communicators and rememberers.