Showing posts with label acting/choosing. Show all posts
Showing posts with label acting/choosing. Show all posts

Wednesday, April 26, 2023

Is the mind a reliable mirror of reality? The marriage of physics and information theory

 This post is the fifth installment of my passing on to both MindBlog readers and my future self my idiosyncratic selection of clips of text from O’Gieblyn’s book ‘God, Human, Animal, Machine’ that I have found particularly interesting. Here are fragments of Chapter 7 from the  fourth section of her book,  titled "Paradox."

Is the mind a reliable mirror of reality? Do the patterns we perceive belong to the objective world, or are they merely features of our subjective experience? Given that physics was founded on the separation of mind and matter, subject and object, it’s unsurprising that two irreconcilable positions that attempt to answer this question have emerged: one that favors subjectivity, the other objectivity. Bohr’s view was that quantum physics describes our subjective experience of the world; it can tell us only about what we observe. Mathematical equations like the wave function are merely metaphors that translate this bizarre world into the language of our perceptual interface—or, to borrow Kant’s analogy, spectacles that allow us to see the chaotic world in a way that makes sense to our human minds. Other interpretations of physics, like the multiverse theory or string theory, regard physics not as a language we invented but as a description of the real, objective world that exists out there, independent of us. Proponents of this view tend to view equations and physical laws as similarly transcendent, corresponding to literal, or perhaps even Platonic, realities.

The marriage of physics and information theory is often attributed to John Wheeler, the theoretical physicist who pioneered, with Bohr, the basic principles of nuclear fission. In the late 1980s, Wheeler realized that the quantum world behaved a lot like computer code. An electron collapsed into either a particle or a wave depending on how we interrogated it. This was not dissimilar from the way all messages can be simplified into “binary units,” or bits, which are represented by zeros and ones. Claude Shannon, the father of information theory, had defined information as “the resolution of uncertainty,” which seemed to mirror the way quantum systems existed as probabilities that collapsed into one of two states. For Wheeler these two fields were not merely analogous but ontologically identical. In 1989 he declared that “all things physical are information-theoretic in origin.            
            
In a way Wheeler was exploiting a rarely acknowledged problem that lies at the heart of physics: it’s uncertain what matter actually is. Materialism, it is often said, is not merely an ontology but a metaphysics—an attempt to describe the true nature of things. What materialism says about our world is that matter is all that exists: everything is made of it, and nothing exists outside of it. And yet, ask a physicist to describe an electron or a quark, and he will speak only of its properties, its position, its behavior—never its essence.

Wheeler’s answer was that matter itself does not exist. It is an illusion that arises from the mathematical structures that undergird everything, a cosmic form of information processing. Each time we make a measurement we are creating new information—we are, in a sense, creating reality itself. Wheeler called this the “participatory universe,” a term that is often misunderstood as having mystical “connotations, as though the mind has some kind of spooky ability to generate objects. But Wheeler did not even believe that consciousness existed. For him, the mind itself was nothing but information. When we interacted with the world, the code of our minds manipulated the code of the universe, so to speak. It was a purely quantitative process, the same sort of mathematical exchange that might take place between two machines.            

While this theory explains, or attempts to explain, how the mind is able to interact with matter, it is a somewhat evasive solution to the mind-body problem, a sleight of hand that discards the original dichotomy by positing a third substance—information—that can explain both. It is difficult, in fact, to do justice to how entangled and self-referential these two fields—information theory and physics—have become, especially when one considers their history. The reason that cybernetics privileged relationships over content in the first place was so that it could explain things like consciousness purely in terms of classical physics, which is limited to describing behavior but not essence—“doing” but not “being.” When Wheeler merged information theory with quantum physics, he was essentially closing the circle, proposing that the hole in the material worldview—intrinsic essence—could be explained by information itself.

Seth Lloyd, an MIT professor who specializes in quantum information, insists that the universe is not like a computer but is in fact a computer. “The universe is a physical system that contains and processes information in a systematic fashion,” he argues, “and that can do everything a computer can do.” Proponents of this view often point out that recent observational data seems to confirm it. Space-time, it turns out, is not smooth and continuous, as Einstein’s general relativity theory assumed, but more like a grid made up of minuscule bits—tiny grains of information that are not unlike the pixels of an enormous screen. Although we experience the world in three dimensions, it seems increasingly likely that all the information in the universe arises from a two-dimensional field, much like the way holograms work, or 3-D films.
            
When I say that I try very hard to avoid the speculative fringe of physics, this is more or less what I am talking about. The problem, though, is that once you’ve encountered these theories it is difficult to forget them, and the slightest provocation can pull you back in. It happened a couple years ago, while watching my teenage cousin play video games at a family gathering. I was relaxed and a little bored and began thinking about the landscape of the game, the trees and the mountains that made up the backdrop. The first-person perspective makes it seem like you’re immersed in a world that is holistic and complete, a landscape that extends far beyond the frame, though in truth each object is generated as needed. Move to the right and a tree is “generated; move to the left and a bridge appears, creating the illusion that it was there all along. What happened to these trees and rocks and mountains when the player wasn’t looking? They disappeared—or no, they were never there to begin with; they were just a line of code. Wasn’t this essentially how the observer effect worked? The world remained in limbo, a potentiality, until the observer appeared and it was compelled to generate something solid. Rizwan Virk, a video game programmer, notes that a core mantra in programming is “only render that which is being observed.”
            
Couldn’t the whole canon of quantum weirdness be explained by this logic? Software programs are never perfect. Programmers cut corners for efficiency—they are working, after all, with finite computing power; even the most detailed systems contain areas that are fuzzy, not fully sketched out. Maybe quantum indeterminacy simply reveals that we’ve reached the limits of the interface. The philosopher Slavoj Žižek once made a joke to this effect. Perhaps, he mused, God got a little lazy when he was creating the universe, like the video game programmer who doesn’t bother to meticulously work out the interior of a house that[ “the player is not meant to enter. “He stopped at a subatomic level,” he said, “because he thought humans would be too stupid to progress so far.”

Monday, April 24, 2023

Networks and Emergentism

This post is the fourth installment of my passing on to both MindBlog readers and my future self my idiosyncratic selection of clips of text from O’Gieblyn’s book ‘God, Human, Animal, Machine’ that I have found particularly interesting. Chapters 5 and 6 form the third section of her book,  titled "Network."

From Chapter 5:

When it comes to biological systems like forests and swarms, emergent behavior that appears to be unified and intelligent can exist without a centralized control system like a brain. But the theory has also been applied to the brain itself, as a way to account for human consciousness. Although most people tend to think of the brain as the body’s central processing unit, the organ itself has no central control. Philosophers and neuroscientists often point out that our belief in a unified interior self—the illusion, as Richard Dawkins once put it, that we are “a unit, not a colony”—has no basis in the actual architecture of the brain. Instead there are only millions of unconscious parts that conspire, much like a bee colony, to create a “system” that is intelligent. Emergentism often entails that consciousness isn’t just in the head; it emerges from the complex relationships that exist throughout the body, and also from the interactions between the body and its environment.

Although emergentism is rooted in physicalism, critics have often claimed that there is something inherently mystical about the theory, particularly when these higher-level patterns are said to be capable of controlling or directing physical processes...few emergentists have managed to articulate precisely what kind of structure might produce consciousness in machines; in some cases the mind is posited simply as a property of “complexity,” a term that is eminently vague. Some critics have argued that emergentism is just an updated version of vitalism—the ancient notion that the world is animated by a life force or energy that permeates all things…Although emergentism is focused specifically on consciousness, as opposed to life itself, the theory is vulnerable to the same criticism that has long haunted vitalism: it is an attempt to get “something from nothing.” It hypothesizes some additional, invisible power that exists within the mechanism, like a ghost in the machine.

…emergence in nature demonstrates that complex systems can self-organize in unexpected ways without being intended or designed. Order can arise from chaos. In machine intelligence, the hope persists that if we put the pieces together the right way—through either ingenuity or sheer accident—consciousness will simply emerge as a side effect of complexity. At some point nature will step in and finish the job…aren’t all creative undertakings rooted in processes that remain mysterious to the creator? Artists have long understood that making is an elusive endeavor, one that makes the artist porous to larger forces that seem to arise from outside herself.

From Chapter 6:

…once the world was a sacred and holy place, full of chatty and companionable objects—rocks and trees that were capable of communicating with us—we now live in a world that has been rendered mute… some disenchantment narratives place the fall from grace not with the Enlightenment and the rise of modern science but with the emergence of monotheism. The very notion of imago dei, with humanity created in God’s image and given “dominion” over creation, has linked human exceptionalism with the degradation of the natural world.  Is it possible to go back? Or are these narratives embedded so deeply in the DNA of our ontological assumptions that a return is impossible? This is especially difficult when it comes to our efforts to create life from ordinary matter…In the orthodox forms of Judaism and Christianity, the ability to summon life from inert matter is denounced as paganism, witchcraft, or idolatry.

Just as the golems were sculpted out of mud and animated with a magical incantation, so the hope persists that robots built from material parts will become inhabited by that divine breath…While these mystical overtones should not discredit emergence as such—it is a useful enough way to describe complex systems like beehives and climates—the notion that consciousness can emerge from machines does seem to be a form of wishful thinking, if only because digital technologies were built on the assumption that consciousness played no role in the process of intelligence. Just as it is somewhat fanciful to believe that science can explain consciousness when modern science itself was founded on the exclusion of the mind, it is difficult to believe that technologies designed specifically to elide the notion of the conscious subject could possibly come to develop an interior life.
           
To dismiss emergentism as sheer magic is to ignore the specific ways in which it differs from the folklore of the golems—even as it superficially satisfies the same desire. Scratch beneath the mystical surface and it becomes clear that emergentism is often not so different from the most reductive forms of materialism, particularly when it comes to the question of human consciousness. Plant intelligence has been called a form of “mindless mastery,” and most emergentists view humans as similarly mindless. We are not rational agents but an encasement of competing systems that lack any sort of unity or agency. Minsky once described the mind as “a sort of tangled-up bureaucracy” whose parts remain ignorant of one another.

Just as the intelligence of a beehive or a traffic jam resides in the patterns of these inert, intersecting parts, so human consciousness is merely the abstract relationships that emerge out of these systems: once you get to the lowest level of intelligence, you inevitably find, as Minsky put it, agents that “cannot think at all.” There is no place in this model for what we typically think of as interior experience, or the self.

Embodied artificial intelligence is being pursued in laboratories using humanoid robots equipped with sensors and cameras that endow the robots with sensory functions and motor skills. The theory is that these sensorimotor capacities will eventually lead to more advanced cognitive skills, such as a sense of self or the ability to use language, though so far this has not happened.
 

Friday, April 21, 2023

Equivalence of the metaphors of the major religions and transhumanism

This post is the third installment of my passing on to both MindBlog readers and my future self my idiosyncratic selection of clips of text from O’Gieblyn’s book ‘God, Human, Animal, Machine’ that I have found particularly interesting. Chapters 3 and 4 form the second section of her book,  titled "Pattern."

From Chapter 3:

Once animal brains began to form, the information became encoded in neural patterns. Now that evolution has produced intelligent, tool-wielding humans, we are designing new information technologies more sophisticated than any object the world has yet seen. These technologies are becoming more complex and powerful each year, and very soon they will transcend us in intelligence. The ‘transhumanist’  movement believes that the only way for us to survive as humans is to begin merging our bodies with these technologies, transforming ourselves into a new species—what Kurzweil calls “posthumans,” or spiritual machines. Neural implants, mind-uploading, and nanotechnology will soon be available, he promises. With the help of these technologies, we will be able to transfer or “resurrect” our minds onto supercomputers, allowing us to become immortal. Our bodies will become incorruptible, immune to disease and decay, and each person will be able to choose a new, customizable virtual physique.

From Chapter 4:

…how is it that the computer metaphor—an analogy that was expressly designed to avoid the notion of a metaphysical soul - has returned to us ancient religious ideas about physical transcendence and the disembodied spirit?

In his book “You Are Not a Gadget”, the computer scientist Jaron Lanier argues that just as the Christian belief in an immanent Rapture often conditions disciples to accept certain ongoing realities on earth—persuading them to tolerate wars, environmental destruction, and social inequality—so too has the promise of a coming Singularity served to justify a technological culture that privileges information over human beings. “If you want to make the transition from the old religion, where you hope God will give you an afterlife,” Lanier writes, “to the new religion, where you hope to become immortal by getting uploaded into a computer, then you have to believe information is real and alive.” This sacralizing of information is evident in the growing number of social media platforms that view their human users as nothing more than receptacles of data. It is evident in the growing obsession with standardized testing in public schools, which is designed to make students look good to an algorithm. It is manifest in the emergence of crowd-sourced sites such as Wikipedia, in which individual human authorship is obscured so as to endow the content with the transcendent aura of a holy text. In the end, transhumanism and other techno-utopian ideas have served to advance what Lanier calls an “antihuman approach to computation,” a digital climate in which “bits are presented as if they were alive, while humans are transient fragments.

In a way we are already living the dualistic existence that Kurzweil promised. In addition to our physical bodies, there exists—somewhere in the ether—a second self that is purely informational and immaterial, a data set of our clicks, purchases, and likes that lingers not in some transcendent nirvana but rather in the shadowy dossiers “of third-party aggregators. These second selves are entirely without agency or consciousness; they have no preferences, no desires, no hopes or spiritual impulses, and yet in the purely informational sphere of big data, it is they, not we, that are most valuable and real.

He too found an “essential equivalence” between transhumanist metaphors and Christian metaphors: both systems of thought placed a premium value on consciousness. The nature of consciousness—as well as the question of who and what is conscious—is the fundamental philosophical question, he said, but it’s a question that cannot be answered by science alone. This is why we need metaphors.  “religion deals with legitimate questions but the major religions emerged in pre-scientific times so that the metaphors are pre-scientific. That the answers to existential questions are necessarily metaphoric is necessitated by the fact that we have to transcend mere matter and energy to find answers…The difference between so-called atheists and people who believe in “God” is a matter of the choice of metaphor, and we could not get through our lives without having to choose metaphors for transcendent questions.
           
Perhaps all these efforts—from the early Christians’ to the medieval alchemists’ to those of the luminaries of Silicon Valley—amounted to a singular historical quest, one that was expressed through analogies that were native to each era. Perhaps our limited vantage as humans meant that all we could hope for were metaphors of our own making, that we would continually grasp at the shadow of absolute truths without any hope of attainment.
 

Wednesday, April 19, 2023

The Illusion of the Self as Humans become Gods.

This post continues my passing on to both MindBlog readers and my future self my idiosyncratic selection of clips of text from O’Gieblyn’s book ‘God, Human, Animal, Machine’ that I have found particularly interesting.  This post deals with Chapter 2 from the first section of the book, 'Image.'  I’m discontinuing the experiment of including Chat GPT 4 condensations of the excerpts. Here are the clips:

It turns out that computers are particularly adept at the tasks that we humans find most difficult: crunching equations, solving logical propositions, and other modes of  abstract thought. What artificial intelligence finds most difficult are the sensory perceptive tasks and motor skills that we perform unconsciously: walking, drinking from a cup, seeing and feeling the world through our senses. Today, as AI continues to blow past us in benchmark after benchmark of higher cognition, we quell our anxiety by insisting that what distinguishes true consciousness is emotions, perception, the ability to experience and feel: the qualities, in other words, that we share with animals.
            
If there were gods, they would surely be laughing their heads off at the inconsistency of our logic. We spent centuries denying consciousness in animals precisely because they lacked reason or higher thought.”

“Metaphors are typically abandoned once they are proven to be insufficient. But in some cases, they become only more entrenched: the limits of the comparison come to redefine the concepts themselves. This latter tactic has been taken up by the eliminativists, philosophers who claim that consciousness simply does not exist. Just as computers can operate convincingly without any internal life, so can we. According to these thinkers, there is no “hard problem” because that which the problem is trying to explain—interior experience—is not real. The philosopher Galen Strawson has dubbed this theory “the Great Denial,” arguing that it is the most absurd conclusion ever to have entered into philosophical thought—though it is one that many prominent”

The eliminativist philosophers claim that consciousness simply does not exist. Just as computers can operate convincingly without any internal life, so can we. According to these thinkers, there is no “hard problem” because that which the problem is trying to explain—interior experience—is not real. The philosopher Galen Strawson has dubbed this theory “the Great Denial,” arguing that it is the most absurd conclusion ever to have entered into philosophical thought—though it is one that many prominent. Chief among the deniers is Daniel Dennett, who has often insisted that the mind is illusory. Dennett refers to the belief in interior experience derisively as the “Cartesian theater,” invoking the popular delusion—again, Descartes’s fault—that there exists in the brain some miniature perceiving entity, a homunculus that is watching the brain’s representations of the external world projected onto a movie screen and making decisions about future actions. One can see the problem with this analogy without any appeals to neurobiology: if there is a homunculus in my brain, then it must itself (if it is able to perceive) contain a still smaller homunculus in its head, and so on, in infinite regress.
            
            Dennett argues that the mind is just the brain and the brain is nothing but computation, unconscious all the way down. What we experience as introspection is merely an illusion, a made-up story that causes us to think we have “privileged access” to our thinking processes. But this illusion has no real connection to the mechanics of thought, and no ability to direct or control it. Some proponents of this view are so intent on avoiding the sloppy language of folk psychology that any any reference to human emotions and intentions is routinely put in scare quotes. We can speak of brains as “thinking,” “perceiving,” or “understanding” so long as it’s clear that these are metaphors for the mechanical processes. “The idea that, in addition to all of those, there is this extra special something—subjectivity—that distinguishes us from the zombie,” Dennett writes, “that’s an illusion.”

Perhaps it’s true that consciousness does not really exist—that, as Brooks put it, we “overanthropomorphize humans.” If I am capable of attributing life to all kinds of inanimate objects, then can’t I do the same to myself? In light of these theories, what does it mean to speak of one’s “self” at all?


Monday, April 17, 2023

Disenchantment of the world and the computational metaphors of our times.

I am doing a second reading of Meghan O’Gieblyn’s book “Gods, Humans, Animals, Machines” and extracting clips of text that I find most interesting.  I’m putting them in a MindBlog post, hoping they will be interesting to some readers, and also because MindBlog is my personal archive of things I want to remember I've engaged. At least I will know where to look up something I'm trying to recall.

 O’Gieblyn’s text has bursts of magisterial insight interspersed with details of her personal experiences and travails, and the clips try to capture my biased selection of the former. 

The first section of her book “Image”, has two Chapters, and this post passes on Chapter 1,  starting with the result of my asking ChatGPT 4 to summarize my clips in approximately 1000 words. It generated the ~300 words below, and I would urge you to continue reading my clips (976 words), which provide a more rich account. Subsequent posts in this series will omit Chat GPT summaries, unless they generate something that blows me away. 

Here is Chat GPT 4’s summary:

The concept of the soul has become meaningless in modern times, reduced to a dead metaphor that no longer holds any real significance. This is due to the process of disenchantment that has taken place since the dawn of modern science, which has turned the world into a subject of investigation and reduced everything to the causal mechanism of physical laws. This has led to a world that is devoid of the spirit-force that once infused and unified all living things, leaving us with an empty carapace of gears and levers. However, the questions that were once addressed by theologians and philosophers persist in conversations about digital technologies, where artificial intelligence and information technologies have absorbed them.  

Humans have a tendency to see themselves in all beings, as evidenced by our habit of attributing human-like qualities to inanimate objects. This has led to the development of the idea of God and the belief that natural events are signs of human agency. This impulse to see human intention and purpose in everything has resulted in a projection of our image onto the divine, which suggests that metaphors are two-way streets and that it is not always easy to distinguish between the source domain and the target.

The development of cybernetics and the application of the computational analogy to the mind has resulted in the description of the brain as the hardware that runs the software of the mind, with cognitive systems being spoken of as algorithms. However, the use of metaphors like these can lead to a limiting of our understanding of the world and how we interact with it. As cognitive linguist George Lakoff notes, when an analogy becomes ubiquitous, it can be difficult to think around it, and it structures how we think about the world.

And here are the text clips I asked ChatGPT 4 to summarize

It is meaningless to speak of the soul in the twenty-first century (it is treacherous even to speak of the self). It has become a dead metaphor, one of those words that survive in language long after a culture has lost faith in the concept, like an empty carapace that remains intact years after its animating organism has died. The soul is something you can sell, if you are willing to demean yourself in some way for profit or fame, or bare by disclosing an intimate facet of your life. It can be crushed by tedious jobs, depressing landscapes, and awful music. All of this is voiced unthinkingly by people who believe, if pressed, that human life is animated by nothing more mystical or supernatural than the firing of neurons

We live in a world that is “disenchanted.” The word is often attributed to Max Weber, who argued that before the Enlightenment and Western secularization, the world was “a great enchanted garden.” In the enchanted world, faith was not opposed to knowledge, nor myth to reason. The realms of spirit and matter were porous and not easily distinguishable from one another. Then came the dawn of modern science, which turned the world into a subject of investigation. Nature was no longer a source of wonder but a force to be mastered, a system to be figured out. At its root, disenchantment describes the fact that everything in modern life, from our minds to the rotation of the planets, can be reduced to the causal mechanism of physical laws. In place of the pneuma, the spirit-force that once infused and unified all living things, we are now left with an empty carapace of gears and levers—or, as Weber put it, “the mechanism of a world robbed of gods.”
            
If modernity has an origin story, this is our foundational myth, one that hinges, like the old myths, on the curse of knowledge and exile from the garden.

To discover truth, it is necessary to work within the metaphors of our own time, which are for the most part technological. Today artificial intelligence and information technologies have absorbed many of the questions that were once taken up by theologians and philosophers: the mind’s relationship to the body, the question of free will, the possibility of immortality. These are old problems, and although they now appear in different guises and go by different names, they persist in conversations about digital technologies much like those dead metaphors that still lurk in the syntax of contemporary speech. All the eternal questions have become engineering problems.

Animism was built into our design. David Hume once remarked upon “the universal tendency among mankind to conceive of all beings like themselves,” an adage we prove every time we kick a malfunctioning appliance or christen our car with a human name. Our brains can’t fundamentally distinguish between interacting with people and interacting with devices. Our habit of seeing our image everywhere in the natural world is what gave birth to the idea of God. Early civilizations assumed that natural events bore signs of human agency. Earthquakes happened because the gods were angry. Famine and drought were evidence that the gods were punishing them. Because human communication is symbolic, humans were quick to regard the world as a system of signs, as though some higher being were seeking to communicate through “natural events. Even the suspicion that the world is ordered, or designed, speaks to this larger impulse to see human intention and human purpose in every last quirk of “creation.”
    
There is evidently no end to our solipsism. So deep is our self-regard that we projected our image onto the blank vault of heaven and called it divine. But this theory, if true, suggests a deeper truth: metaphors are two-way streets. It is not so easy to distinguish the source domain from the target, to remember which object is the original and which is modeled after its likeness. The logic can flow in either direction. For centuries we said we were made in God’s image, when in truth we made him in ours.”

Shannon removed the thinking mind from the concept of information. Meanwhile, McCulloch applied the logic of information processing to the mind itself. This resulted in a model of mind in which thought could be accounted for in purely abstract, mathematical terms, and opened up the possibility that computers could execute mental functions. If thinking was just information processing, computers could be said to “learn,” “reason,” and “understand”—words that were, at least in the beginning, put in quotation marks to denote them as metaphors. But as cybernetics evolved and the computational analogy was applied across a more expansive variety of biological and artificial systems, the limits of the metaphor began to dissolve, such that it became increasingly difficult to tell the difference between matter and form, medium and message, metaphor and reality. And it became especially difficult to explain aspects of the mind that could not be accounted for by the metaphor.

The brain is often described today as the hardware that “runs” the software of the mind. Cognitive systems are spoken of as algorithms: vision is an algorithm, and so are attention, language acquisition, and memory.
            
In 1999 the cognitive linguist George Lakoff noted that the analogy had become such a given that neuroscientists “commonly use the Neural Computation metaphor without noticing that it is a metaphor.” He found this concerning. Metaphors, after all, are not merely linguistic tools; they structure how we think about the world, and when “an analogy becomes ubiquitous, it is impossible to think around it. ..there is virtually no form of discourse about intelligent human behavior that proceeds without employing this metaphor, just as no form of discourse about intelligent human behavior could proceed in certain eras and cultures without reference to a spirit or deity.”

Wednesday, April 12, 2023

The Physics of Intelligence - and LDMs (Large Danger Models)

I want to pass on my abstracting of an interesting article by Venkatesh Rao, another instance of my using MindBlog as my personal filing system to be sure I can come back to - and refresh my recall of - ideas I think are important.  I also pass on ChatGPT 3.5 and ChatGPT 4's summaries of my summary!

The Physics of Intelligence   -  The missing discourse of AI

There are strong philosophy and engineering discourses, but no physics discourse. This is a problem because when engineers mainline philosophy questions in engineering frames without the moderating influence of physics frames, you get crackpottery…I did not say the physics of artificial intelligence…The physics of intelligence is no more about silicon semiconductors or neurotransmitters than the physics of flight is about feathers or aluminum.

Attention is the focus of one of the six basic questions about the physics of intelligence that I’ve been thinking about. Here is my full list:
What is attention, and how does it work?
What role does memory play in intelligence?
How is intelligence related to information?
How is intelligence related to spacetime?
How is intelligence related to matter?
How is intelligence related to energy and thermodynamics?
 

The first three are obviously at the “physics of intelligence” level of abstraction, just as “wing” is at the “physics of flight” level of abstraction. The last three get more abstract, and require some constraining, but there are already some good ideas floating around on how to do the constraining…We are not talking about the physics of computation in general…computation and intelligence are not synonymous or co-extensive…To talk about intelligence, it is necessary, but not sufficient, to talk about computation. You also have to talk about the main aspects of embodiment: spatial and temporal extent, materiality, bodily integrity maintenance in relation to environmental forces, and thermodynamic boundary conditions.
 

What is attention, and how does it work?

A computer is “paying attention” to the data and instructions in the CPU registers in any given clock cycle…but fundamentally, attention is not a design variable used in complex ways in basic computing. You could say AI begins when you start deploying computational attention in a more dynamic way.

Attention is to intelligence as wing is to flight. The natural and artificial variants have the same sort of loose similarity. Enough that using the same word to talk about both is justified…In AI, attention refers primarily to a scheme of position encoding of a data stream. Transformer models like GPT keep track of the position of each token in the input and output streams, and extract meaning out of it. Where a word is in a stream matters almost as much as what the word is.

You can interpret these mechanisms as attention in a human sense. What is in the context of a text? In text streams, physical proximity (tokens before and after), syntactic proximity (relationship among clauses and words in a formal grammatical sense) and semantic proximity (in the sense of some words, including very distant ones, being significant in the interpretation of others) all combine to create context. This is not that different from how humans process text. So at least to first order, attention in human and artificial systems is quite similar.

But as with wings, the differences matter. Human attention, arguably, is not primarily about information processing at all. It is about energy management. We pay attention to steady our energy into a steady, sustainable, and renewable flow. We engage in low-information activities like meditation, ritual, certain kinds of art, and prayer to learn to govern our attention in specific ways. This is not to make us better at information processing, but for other reasons, such as emotion regulation and motivation. Things like dopamine loops of satisfaction are involved. The use of well-trained attention for better information processing is only one of the effects.

Overall, human attention is more complex and multifaceted than AI attention, just as bird wings are fundamentally more complex mechanically. Attention in the sense of position-encoding for information processing is like the pure lift function of a wing. Human attention, in addition, serves additional functions analogous to control and propulsion type functions.

What role does memory play in intelligence?

The idea of attention leads naturally to the idea of memory. Trivially, memory is a curated record of everything you’ve paid attention to in the past…An obvious way to understand current developments in AI is to think of LLMs and LIMs as idiosyncratically compressed atemporal textual and visual memories. Multimodal models can be interpreted as generalizations of this idea.

Human memory is modulated by evolving risk perceptions as it is laid down, and emotions triggered by existing memories regulates current attention, determining what new data gets integrated into the evolving model (as an aside, this is why human memory exists as a kind of evolving coherent narrative of self, rather than just as a pile of digested stuff).

Biological organism memory is not just an undifferentiated garbage record (LGM) of what you paid attention to in the past; it shapes what you pay attention to in the future very directly and continuously. Biological memory is strongly opinionated memory. If a dog bites you…you can’t afford to separate training and inference in the sense of “training” on a thousand dog encounters…you have to use your encounter with Dog 1 to shape your attentional response to Dog 2. Human memories are like LGMs, except that the training process is regulated by a live emotional regulation feedback loop that somehow registers and acts on evolving risk assessments. There’s a term for this in psychology (predictive coding or predictive processing) with a hand-wavy theory of energy-minimization attached, but I don’t find it fully satisfying.

I have a placeholder name for this scheme, but as yet it’s not very fleshed out. Biological memories are Large Danger Models (LDMs).

Why just danger? Why not other signals and drives towards feeding, sex, interestingness, poetry, and so on? I have a stronger suspicion that danger is all you need to generate human-like memory, and in particular human-like experience of time. Human memory is the result of playing to continue the game, ie an infinite-game orientation. Once you have that in place, everything else emerges. It’s not as fundamental as basic survival.

AIs don’t yet count as human-equivalent to me: they’re in no danger, ever. Since we’re in the brute-force stage of training AI models, we train them on basically everything we have, with no danger signal accompanying any of it…AIs today develop their digested memories with no ongoing encoding or capture of the evolving risk and emotion assessments that modulate human memories. Even human grade schools, terrible as they are, do a better job than AI training protocols…the next big leap should be achievable by encoding some kind of environmental risk signal. Ie, we just need to put AIs in danger in the right way. My speculative idea of LDMs don’t seem that mysterious. LDMs are an engineering moonshot, not a physics mystery.

To lay it out more clearly, consider a thought experiment...Suppose you put a bunch of AIs in robot bodies, and let them evolve naturally, while scrounging resources for survival. To keep it simple, let’s say they only compete over scarce power outlets to charge their batteries. Their only hardcoded survival behavior is to plug in when running low….Let’s say the robots are randomly initialized to pay attention differently to different aspects of data coursing through them. Some robots pay extra attention to other robots’ actions. Other robots pay extra attention to the rocks in the environment. Obviously, the ones that happen to pay attention in the right ways will end up outcompeting the ones who don’t. The sneaky robots will evolve to be awake when other robots are powered down or hibernating for self-care, and hog the power outlets then. The bigger robots will learn they can always get the power outlets by shoving the smaller ones out of the way.

Now the question is: given all the multimodal data flowing through them, what will the robots choose to actually remember in their limited storage spaces, as their starter models get trained up? What sorts of LDMs will emerge? How will the risk modulation emerge? What equivalent of emotional regulation will emerge? What sense of time will emerge?

The thought experiment of LDMs suggests a particular understanding of memory in relation to intelligence: memory is risk-modulated experiential data persistence that modulates ongoing experiential attention and risk-management choices....It’s a bit of a mouthful, but I think that’s fundamentally it.

I suspect the next generation of AI models will include some such embodiment feedback loop so memory goes from being brute-force generic persistence to persistence that’s linked to embodied behaviors in a feedback loop exposed to environmental dangers that act as survival pressures.

The resulting AIs won’t be eidetic idiot savants, and less capable in some ways, but will be able to survive in environments more dangerous than datacenters exposed to the world only through sanitized network connections. Instead of being Large Garbage Models (LGMs), they’ll be Large Danger Models (LDMs).
 

How is intelligence related to information?
 

We generally think about information as either primitive (you just have to know it) or entailed (you can infer it from what you already know)…Primitive information is a kind of dehydrated substance to which you can add compute (water) to expand it. Entailed information can be dehydrated into primitive form. Compression of various sorts exploits different ways of making the primitive/entailed distinction.

When you think of intelligence in relation to information though, you have to get more sophisticated…We think in terms of whether or not new data patterns require new categories, or simply modify the strengths of, and interrelationships among, existing ones…are you looking at something new, or is this just a variant or instance of something you already know?

Information for an intelligent system them, is best understood in an ontological novelty way rather than an information-theoretic way. Because it is not as fundamental an abstraction level, it is more arbitrary, which means how you factor your experience stream into categories is as much a function of the idiosyncrasies of your survival mode as it is a function of the bits-and-bytes level structure of what you’re experiencing…the models are making up weird compressions that are not human-understandable. That’s what information is to an intelligent system: efficient ontologies that reflect how that particular intelligence is situated in its environment.

Or to put it more plainly: information is the hallucinations an intelligence makes up to categorize reality compactly, in order to survive efficiently.

The hallucinations that LLMs make up — papers and citations that don’t exist, weird words that show up in images, strange new symbols, inscrutable compressions — are neither errors in truth-seeking, nor evidence of weird kinds of magic going on. It is just raw information (in the bits and bytes Shannon sense) that has been judged for salience to the basic challenge of survival, and represented accordingly, in evolving memory. It is ontologically compressed experiential data. It’s just extra weird because the models are evolving in a trivial non-dangerous environment. So distinctions between robustly real categories and fragile hallucinations don’t matter. An academic can get fired for making up a citation. An LLM just gets wonder and mystique.

This is why information comes after memory in my physics of intelligence. Your memories are a function of a real-time danger management behavior. Your emergent categories of cognition did not evolve to be true in any rigorous sense. They evolved to help you survive. If someone who names and classifies certain data patterns as “ghosts” survives better than someone who names and classifies them “paranoid projections,” that’s how it goes. Truth is, in a evolutionary sense, the most robust representation of experience that helps you survive efficiently in the most challenging environments.

Models that describe reality in a broad and interesting way emerge when the danger environment tests you enough that you can’t bullshit to survive, but not so much that you must bullshit to survive. Both disconnected daydreams and psychotic breaks are poor foundations for existing in reality.

Spacetime, Matter, and Energy NOTE:  I am not including clips from this section in this post.
 

Here is ChatGPT 3.5's response to my request to summarize the above in 1000 words:

The article discusses the importance of including a physics discourse in the field of AI. The author proposes six basic questions about the physics of intelligence: attention, memory, information, spacetime, matter, and energy and thermodynamics. The concept of attention is related to position encoding of a data stream, and while it has some similarities between human and artificial systems, the differences are significant. Human attention is about energy management and regulation, and it is multifaceted. Memory is a curated record of everything a person has paid attention to in the past, but it is also modulated by evolving risk perceptions and emotions triggered by existing memories. Biological memory is strongly opinionated memory, and the training process is regulated by a live emotional regulation feedback loop that somehow registers and acts on evolving risk assessments. The author suggests that biological memories are Large Danger Models (LDMs) and that danger is all that is needed to generate human-like memory and experience of time.

And here is ChatGPT 4's response, slightly edited to remove one obvious glitch:

The Physics of Intelligence: The Missing Discourse of AI

The discourse on artificial intelligence (AI) is currently dominated by philosophy and engineering but lacks a physics perspective. To truly understand intelligence, it is necessary to consider its physical aspects, such as spatial and temporal extent, materiality, and thermodynamics.

Six questions about the physics of intelligence are proposed:
What is attention, and how does it work?
What role does memory play in intelligence?
How is intelligence related to information?
How is intelligence related to spacetime?
How is intelligence related to matter?
How is intelligence related to energy and thermodynamics?

Attention is fundamental to intelligence. In AI, attention refers to the position encoding of a data stream. While human and artificial attention mechanisms are similar, human attention is more complex and multifaceted, serving additional functions related to energy management, motivation, and emotion regulation.

Memory is the curated record of everything one has paid attention to in the past. Current AI models, such as LLMs (Language Models) and LIMs (Image Models), can be seen as compressed textual and visual memories. Human memory, however, is modulated by risk perceptions, emotions, and experiences. The idea of Large Danger Models (LDMs) is proposed, where memory is risk-modulated experiential data persistence that modulates ongoing experiential attention and risk-management choices. The next generation of AI models may include embodiment feedback loops that link memory to embodied behaviors in a feedback loop exposed to environmental dangers, resulting in more capable AIs.

Intelligence is related to information through efficient ontologies that reflect how an intelligence is situated in its environment. For intelligent systems, information is best understood in terms of ontological novelty rather than information-theoretic ways. Information is the hallucinations an intelligence makes up to categorize reality compactly to survive efficiently. This is why information comes after memory in the physics of intelligence, as memories are a function of real-time danger management behavior. Truth, in an evolutionary sense, is the most robust representation of experience that helps an organism survive efficiently in challenging environments.

The physics of intelligence is not solely about computation but also involves understanding the physical aspects of intelligence, such as attention, memory, and information. Future AI models may need to incorporate embodiment feedback loops and risk modulation to develop more human-like intelligence.
 



Monday, February 20, 2023

Fundamentally rethinking what a mind is and how a brain works.

The February Issue of Trends in Cognitive Science has an open source Opinions article from Lisa Feldman Barrett and collaborators that suggests that new research approaches grounded in different ontological commitments will be required to properly describe brain-behavior relationships. Here is a clip of the introductory text and a graphic clip from the article. Finally, I pass on the concluding remarks on fundamentally rethinking what a mind is and how a brain works.
Most brain imaging studies present stimuli and measure behavioral responses in temporal units (trials) that are ordered randomly. Participants’ brain signals are typically aggregated to model structured variation that allows inferences about the broader population from which people were sampled. These methodological details, when used to study any phenomenon of interest, often give rise to brain-behavior findings that vary unexpectedly (across stimuli, context, and people). Such findings are typically interpreted as replication failures, with the observed variation discounted as error caused by less than rigorous experimentation (Box 1). Methodological rigor is of course important, but replication problems may stem, in part, from a more pernicious source: faulty assumptions (i.e., ontological commitments) that mis-specify the psychological phenomena of interest.

In this paper, we review three questionable assumptions whose reconsideration may offer opportunities for a more robust and replicable science: 

 (1) The localization assumption: the instances that constitute a category of psychological events (e.g., instances of fear) are assumed to be caused by a single, dedicated psychological process implemented in a dedicated neural ensemble (see Glossary). 

 (2) The one-to-one assumption: the dedicated neural ensemble is assumed to map uniquely to that psychological category, such that the mapping generalizes across contexts, people, measurement strategies, and experimental designs. 

 (3) The independence assumption: the dedicated neural ensemble is thought to function independently of contextual factors, such as the rest of the brain, the body, and the surrounding world, so the ensemble can be studied alone without concern for those other factors. Contextual factors might moderate activity in the neural ensemble but should not fundamentally change its mapping to the instances of a psychological category. 

 These three assumptions are rooted in a typological view of the mind, brain, and behavior [1. ] that was modeled on 19th century physics and continues to guide experimental practices in much of brain-behavior research to the present day. In this paper, we have curated examples from studies of human functional magnetic resonance imaging (fMRI) and neuroscience research using non-human animals that call each assumption into question. We then sketch the beginnings of an alternative approach to study brain-behavior relationships, grounded in different ontological commitments: (i) a mental event comprises distributed activity across the whole brain; (ii) brain and behavior are linked by degenerate (i.e., many-to-one) mappings; and (iii) mental events emerge as a complex ensemble of weak, nonlinearly interacting signals from the brain, body, and external world.

 

Concluding remarks

Scientific communities tacitly agree on assumptions about what exists (called ontological commitments), what questions to ask, and what methods to use. All assumptions are firmly rooted in a philosophy of science that need not be acknowledged or discussed but is practiced nonetheless. In this article, we questioned the ontological commitments of a philosophy of science that undergirds much of modern neuroscience research and psychological science in particular. We demonstrated that three common commitments should be reconsidered, along with a corresponding course correction in methods. Our suggestions require more than merely improved methodological rigor for traditional experimental design. Such improvements are important, but may aid robustness and replicability only when the ontological assumptions behind those methods are valid. Accordingly, a productive way forward may be to fundamentally rethink what a mind is and how a brain works. We have suggested that mental events arise from a complex ensemble of signals across the entire brain, as well as the from the sensory surfaces of the body that inform on the states of the inner body and outside world, such that more than one signal ensemble maps to a single instance of a single psychological category (maybe even in the same context. To this end, scientists might find inspiration by mining insights from adjacent fields, such as evolution, anatomy, development, and ecology , as well as cybernetics and systems theory. At stake is nothing less than a viable science of how a brain creates a mind through its constant interactions with its body, its physical environment, and with the other brains-in-bodies that occupy its social world.

Monday, November 07, 2022

Sadder but Wiser? Maybe Not.

It looks like another universally accepted result of psychological research may be wrong - that depressed people have a more accurate reading of their ability to affect outcomes. Barry points to work by Dev et al. that fails to replicate experiments of Alloy and Abramson done 43 years ago that led to the hypothesis of “depressive realism,” that depressed people having a more realistic view of their conditions because they are free of the optimistic bias of their cheerful peers. The new research was unable to find any association between depressive symptoms and outcome bias. While Barry's review notes debate over whether differences in the design of the older and newer experiments may account for the variance in results, there now is certainly less confidence in the original findings.

Wednesday, August 24, 2022

The brain chemistry underlying mental exhaustion.

Emily Underwood does a review of work by Wiehler et al. (open source) on the brain chemistry underlying mental fatigue, also describing several reservations expressed by other researchers. From her description:
The researchers divided 39 paid study participants into two groups, assigning one to a series of difficult cognitive tasks that were designed to induce mental exhaustion. In one, participants had to decide whether letters and numbers flashing on a computer screen in quick succession were green or red, uppercase or lowercase, and other variations. In another, volunteers had to remember whether a number matched one they’d seen three characters earlier...As the day dragged on, the researchers repeatedly measured cognitive fatigue by asking participants to make choices that required self-control—deciding to forgo cash that was immediately available so they could earn a larger amount later, for example. The group that had been assigned to more difficult tasks made about 10% more impulsive choices than the group with easier tasks, the researchers observed. At the same time, their glutamate levels rose by about 8% in the lateral prefrontal cortex—a pattern that did not show up in the other group...

Here is the Wiehler et al. abstract:  

Highlights

• Cognitive fatigue is explored with magnetic resonance spectroscopy during a workday 
• Hard cognitive work leads to glutamate accumulation in the lateral prefrontal cortex 
• The need for glutamate regulation reduces the control exerted over decision-making 
• Reduced control favors the choice of low-effort actions with short-term rewards
Summary
Behavioral activities that require control over automatic routines typically feel effortful and result in cognitive fatigue. Beyond subjective report, cognitive fatigue has been conceived as an inflated cost of cognitive control, objectified by more impulsive decisions. However, the origins of such control cost inflation with cognitive work are heavily debated. Here, we suggest a neuro-metabolic account: the cost would relate to the necessity of recycling potentially toxic substances accumulated during cognitive control exertion. We validated this account using magnetic resonance spectroscopy (MRS) to monitor brain metabolites throughout an approximate workday, during which two groups of participants performed either high-demand or low-demand cognitive control tasks, interleaved with economic decisions. Choice-related fatigue markers were only present in the high-demand group, with a reduction of pupil dilation during decision-making and a preference shift toward short-delay and little-effort options (a low-cost bias captured using computational modeling). At the end of the day, high-demand cognitive work resulted in higher glutamate concentration and glutamate/glutamine diffusion in a cognitive control brain region (lateral prefrontal cortex [lPFC]), relative to low-demand cognitive work and to a reference brain region (primary visual cortex [V1]). Taken together with previous fMRI data, these results support a neuro-metabolic model in which glutamate accumulation triggers a regulation mechanism that makes lPFC activation more costly, explaining why cognitive control is harder to mobilize after a strenuous workday.

Friday, August 05, 2022

Dissecting and improving motor skill acquisition in older adults

 From the introduction of Elvira et al. (open source):

We designed a study intended to identify (i) the main factors leading to differences in motor skill acquisition with aging and (ii) the effect of applying noninvasive brain stimulation during motor training. Comparing different components of motor skill acquisition in young and older adults, constituting the extremes of performance in this study, we found that the improvement of the sequence-tapping task is maximized by the early consolidation of the spatial properties of the sequence in memory (i.e., sequence order), leading to a reduced error of execution, and by the optimization of its temporal features (i.e., chunking). We found the consolidation of spatiotemporal features to occur early in training in young adults, suggesting the emergence of motor chunks to be a direct consequence of committing the sequence elements to memory. This process, seemingly less efficient in older adults, could be partially restored using atDCS by enabling the early consolidation of spatial features, allowing them to prioritize the increase of their speed of execution, ultimately leading to an earlier consolidation of motor chunks. This separate consolidation of spatial and temporal features seen in older adults suggests that the emergence of temporal patterns, commonly identified as motor chunks at a behavioral level, stem from the optimization of the execution of the motor sequence resulting from practice, which can occur only after the sequence order has been stored in memory.
Here is their abstract:
Practicing a previously unknown motor sequence often leads to the consolidation of motor chunks, which enable its accurate execution at increasing speeds. Recent imaging studies suggest the function of these structures to be more related to the encoding, storage, and retrieval of sequences rather than their sole execution. We found that optimal motor skill acquisition prioritizes the storage of the spatial features of the sequence in memory over its rapid execution early in training, as proposed by Hikosaka in 1999. This process, seemingly diminished in older adults, was partially restored by anodal transcranial direct current stimulation over the motor cortex, as shown by a sharp improvement in accuracy and an earlier yet gradual emergence of motor chunks. These results suggest that the emergence of motor chunks is preceded by the storage of the sequence in memory but is not its direct consequence; rather, these structures depend on, and result from, motor practice.

Wednesday, August 03, 2022

Motor learning without movement

Fascinating work from Kim et al. on the influence of the prediction errors that are essential in calibrating actions of our predictive minds:

Significance

Our brains control aspects of our movements without conscious awareness, allowing many of us to effortlessly pick up a glass of water or wave hello. Here, we demonstrate that this implicit motor system can learn to refine movements that we plan but ultimately decide not to perform. Participants planned to reach to a target but sometimes withheld these reaches while an animation simulated missing the target. Afterward, participants unknowingly reached opposite the direction of the apparent mistake, indicating that the implicit motor system had learned from the animated error. These findings indicate that movement is not strictly necessary for motor adaptation, and we can learn to update our actions without physically performing them.
Abstract
Prediction errors guide many forms of learning, providing teaching signals that help us improve our performance. Implicit motor adaptation, for instance, is thought to be driven by sensory prediction errors (SPEs), which occur when the expected and observed consequences of a movement differ. Traditionally, SPE computation is thought to require movement execution. However, recent work suggesting that the brain can generate sensory predictions based on motor imagery or planning alone calls this assumption into question. Here, by measuring implicit motor adaptation during a visuomotor task, we tested whether motor planning and well-timed sensory feedback are sufficient for adaptation. Human participants were cued to reach to a target and were, on a subset of trials, rapidly cued to withhold these movements. Errors displayed both on trials with and without movements induced single-trial adaptation. Learning following trials without movements persisted even when movement trials had never been paired with errors and when the direction of movement and sensory feedback trajectories were decoupled. These observations indicate that the brain can compute errors that drive implicit adaptation without generating overt movements, leading to the adaptation of motor commands that are not overtly produced.

Monday, July 25, 2022

Efficiently irrational: deciphering the riddle of human choice

Highlights of an open source article from Paul Glimcher:
A central question for decision-making scholars is: why are humans and animals so predictably inconsistent in their choices? In the language of economics, why are they irrational?
Data suggest that this reflects an optimal trade-off between the precision with which the brain represents the values of choices and the biological costs of that precision. Increasing representational precision may improve choice consistency, but the metabolic cost of increased precision is significant.
Given the cost of precision, the brain might use efficient value-encoding mechanisms that maximize informational content. Mathematical analyses suggest that a mechanism called divisive normalization approximates maximal efficiency per action potential in decision systems.
Behavioral studies appear to validate this claim. Inconsistencies produced by decision-makers can be well modeled as the byproduct of efficient divisive normalization mechanisms that maximize information while minimizing metabolic costs.

Friday, June 17, 2022

Testerone production in adult men is regulated by an adolescent period sensitive to family experiences.

 From Gettler et al.:

Significance
Testosterone influences how animals devote energy and time toward reproduction, including opposing demands of mating and competition versus parenting. Reflecting this, testosterone often declines in new fathers and lower testosterone is linked to greater caregiving. Given these roles, there is strong interest in factors that affect testosterone, including early-life experiences. In this multidecade study, Filipino sons whose fathers were present and involved with raising them when they were adolescents had lower testosterone when they later became fathers, compared to sons whose fathers were present but uninvolved or were not coresident. Sons’ own parenting behaviors did not explain these patterns. These results connect key social experiences during adolescence to adult testosterone, and point to possible intergenerational effects of parenting style.
Abstract
Across vertebrates, testosterone is an important mediator of reproductive trade-offs, shaping how energy and time are devoted to parenting versus mating/competition. Based on early environments, organisms often calibrate adult hormone production to adjust reproductive strategies. For example, favorable early nutrition predicts higher adult male testosterone in humans, and animal models show that developmental social environments can affect adult testosterone. In humans, fathers’ testosterone often declines with caregiving, yet these patterns vary within and across populations. This may partially trace to early social environments, including caregiving styles and family relationships, which could have formative effects on testosterone production and parenting behaviors. Using data from a multidecade study in the Philippines (n = 966), we tested whether sons’ developmental experiences with their fathers predicted their adult testosterone profiles, including after they became fathers themselves. Sons had lower testosterone as parents if their own fathers lived with them and were involved in childcare during adolescence. We also found a contributing role for adolescent father–son relationships: sons had lower waking testosterone, before and after becoming fathers, if they credited their own fathers with their upbringing and resided with them as adolescents. These findings were not accounted for by the sons’ own parenting and partnering behaviors, which could influence their testosterone. These effects were limited to adolescence: sons’ infancy or childhood experiences did not predict their testosterone as fathers. Our findings link adolescent family experiences to adult testosterone, pointing to a potential pathway related to the intergenerational transmission of biological and behavioral components of reproductive strategies.

Wednesday, February 16, 2022

Our brains store concepts as sensory-motor and affective information

Fascinating work from Fernandino et al., who show that concept representations are not independent of sensory-motor experience: 

Significance

The ability to identify individual objects or events as members of a kind (e.g., “knife,” “dog,” or “party”) is a fundamental aspect of human cognition. It allows us to quickly access a wealth of information pertaining to a newly encountered object or event and use it to guide our behavior. How is this information represented in the brain? We used functional MRI to analyze patterns of brain activity corresponding to hundreds of familiar concepts and quantitatively characterized the informational structure of these patterns. Our results indicate that conceptual knowledge is stored as patterns of neural activity that encode sensory-motor and affective information about each concept, contrary to the long-held idea that concept representations are independent of sensory-motor experience.
Abstract
The nature of the representational code underlying conceptual knowledge remains a major unsolved problem in cognitive neuroscience. We assessed the extent to which different representational systems contribute to the instantiation of lexical concepts in high-level, heteromodal cortical areas previously associated with semantic cognition. We found that lexical semantic information can be reliably decoded from a wide range of heteromodal cortical areas in the frontal, parietal, and temporal cortex. In most of these areas, we found a striking advantage for experience-based representational structures (i.e., encoding information about sensory-motor, affective, and other features of phenomenal experience), with little evidence for independent taxonomic or distributional organization. These results were found independently for object and event concepts. Our findings indicate that concept representations in the heteromodal cortex are based, at least in part, on experiential information. They also reveal that, in most heteromodal areas, event concepts have more heterogeneous representations (i.e., they are more easily decodable) than object concepts and that other areas beyond the traditional “semantic hubs” contribute to semantic cognition, particularly the posterior cingulate gyrus and the precuneus.

Friday, February 04, 2022

Attention and executive functions - improvements and declines with ageing.

From Verissimo et al.:
Many but not all cognitive abilities decline during ageing. Some even improve due to lifelong experience. The critical capacities of attention and executive functions have been widely posited to decline. However, these capacities are composed of multiple components, so multifaceted ageing outcomes might be expected. Indeed, prior findings suggest that whereas certain attention/executive functions clearly decline, others do not, with hints that some might even improve. We tested ageing effects on the alerting, orienting and executive (inhibitory) networks posited by Posner and Petersen’s influential theory of attention, in a cross-sectional study of a large sample (N = 702) of participants aged 58–98. Linear and nonlinear analyses revealed that whereas the efficiency of the alerting network decreased with age, orienting and executive inhibitory efficiency increased, at least until the mid-to-late 70s. Sensitivity analyses indicated that the patterns were robust. The results suggest variability in age-related changes across attention/executive functions, with some declining while others improve.

Wednesday, January 26, 2022

Our brains have multiple representations of the same body part.

Here is a neat finding. Remember your elementary biology textbook picture of the homunculi in our somatosensory and motor cortices? The small human figure spread across the surface of the brain, with a cortical location for each part of the hand or other body part? Matsumiya shows that when we direct our eye and hand movements to the same body part these two movements are found to be guided by different body maps! Here is his abstract:  

Significance

Accurate motor control depends on maps of the body in the brain, called the body schema. Disorders of the body schema cause motor deficits. Although we often execute actions with different motor systems such as the eye and hand, how the body schema operates during such actions is unknown. In this study, participants simultaneously directed eye and hand movements to the same body part. These two movements were found to be guided by different body maps. This finding demonstrates multiple motor system–specific representations of the body schema, suggesting that the choice of motor system toward one’s body can determine which of the brain’s body maps is observed. This may offer a new way to visualize patients’ body schema.
Abstract
Purposeful motor actions depend on the brain’s representation of the body, called the body schema, and disorders of the body schema have been reported to show motor deficits. The body schema has been assumed for almost a century to be a common body representation supporting all types of motor actions, and previous studies have considered only a single motor action. Although we often execute multiple motor actions, how the body schema operates during such actions is unknown. To address this issue, I developed a technique to measure the body schema during multiple motor actions. Participants made simultaneous eye and reach movements to the same location of 10 landmarks on their hand. By analyzing the internal configuration of the locations of these points for each of the eye and reach movements, I produced maps of the mental representation of hand shape. Despite these two movements being simultaneously directed to the same bodily location, the resulting hand map (i.e., a part of the body schema) was much more distorted for reach movements than for eye movements. Furthermore, the weighting of visual and proprioceptive bodily cues to build up this part of the body schema differed for each effector. These results demonstrate that the body schema is organized as multiple effector-specific body representations. I propose that the choice of effector toward one’s body can determine which body representation in the brain is observed and that this visualization approach may offer a new way to understand patients’ body schema.

Wednesday, January 05, 2022

Higher performance and fronto-parietal brain activity following active versus passive learning

From a brief open source PNAS report by Stillesjö et al. that has a nice graphic of the fMRI data supporting their observations:
We here demonstrate common neurocognitive long-term memory effects of active learning that generalize over course subjects (mathematics and vocabulary) by the use of fMRI. One week after active learning, relative to more passive learning, performance and fronto-parietal brain activity was significantly higher during retesting, possibly related to the formation and reactivation of semantic representations. These observations indicate that active learning conditions stimulate common processes that become part of the representations and can be reactivated during retrieval to support performance. Our findings are of broad interest and educational significance related to the emerging consensus of active learning as critical in promoting good long-term retention.

Monday, November 15, 2021

Coevolution of tool use and language - shared syntactic processes and basal ganglia substrates

Thibault et al. show that tool use and language share syntactic processes. Functional magnetic resonance imaging reveals that tool use and syntax in language elicit similar patterns of brain activation within the basal ganglia. This indicates common neural resources for the two abilities. Indeed, learning transfer occurs across the two domains: Tool-use motor training improves syntactic processing in language and, reciprocally, linguistic training with syntactic structures improves tool use. Here is their entire structured abstract:   

INTRODUCTION

Tool use is a hallmark of human evolution. Beyond its sensorimotor components, the complexity of which has been extensively investigated, tool use affects cognition from a different perspective. Indeed, tool use requires integrating an external object as a body part and embedding its functional structure in the motor program. This adds a hierarchical level into the motor plan of manual actions, subtly modifying the relationship between interdependent subcomponents. Embedded structures also exist in language, and syntax is the cognitive function handling these linguistic hierarchies. One example is center-embedded object-relative clauses: “The poet [that the scientist admires] reads the paper.” Accordingly, researchers have advanced a role for syntax in action and the existence of similarities between the processes underlying tool use and language, so that shared neural resources for a common cognitive function could be at stake.
RATIONALE
We first tested the existence of shared neural substrates for tool use and syntax in language. Second, we tested the prediction that training one ability should affect performance in the other. In a first experiment, we measured participants’ brain activity with functional magnetic resonance imaging during tool use or, as a control, manual actions. In separate runs, the same participants performed a linguistic task on complex syntactic structures. We looked for common activations between tool use and the linguistic task, predicting similar patterns of activity if they rely on common neural resources. In further behavioral experiments, we tested whether motor training with the tool selectively improves syntactic performance in language and if syntactic training in language, in turn, selectively improves motor performance with the tool.
RESULTS
Tool-use planning and complex syntax processing (i.e., object relatives) elicited neural activity anatomically colocalized within the basal ganglia. A control experiment ruled out verbal working memory and manual (i.e., without a tool) control processes as an underlying component of this overlap. Multivariate analyses revealed similar spatial distributions of neural patterns prompted by tool-use planning and object-relative processing. This agrees with the recruitment of the same neural resources by both abilities and with the existence of a supramodal syntactic function. The shared neurofunctional resources were moreover reflected behaviorally by cross-domain learning transfer. Indeed, tool-use training significantly improved linguistic performance with complex syntactic structures. No learning transfer was observed on language syntactic abilities if participants trained without the tool. The reverse was also true: Syntactic training with complex sentences improved motor performance with the tool more than motor performance in a task without the tool and matched for sensorimotor difficulty. No learning transfer was observed on tool use if participants trained with simpler syntactic structures in language.
CONCLUSION
These findings reveal the existence of a supramodal syntactic function that is shared between language and motor processes. As a consequence, training tool-use abilities improves linguistic syntax and, reciprocally, training linguistic syntax abilities improves tool use. The neural mechanisms allowing for boosting performance in one domain by training syntax in the other may involve priming processes through preactivation of common neural resources, as well as short-term plasticity within the shared network. Our findings point to the basal ganglia as the neural site of supramodal syntax that handles embedded structures in either domain and also support longstanding theories of the coevolution of tool use and language in humans.

Monday, October 11, 2021

Precision and the Bayesian brain

I've been studying and trying to understand the new prevailing model of how our brains work that is emerging - the brain as a Baysean predictive processing machine that compares its prior knowledge with incoming evidence of its correctness. If a mis-match occurs that might suggest alterning a prior expectation, the precision of the incoming evidence is very important. In a recent issue of Current Biology Yon and Frith offer a very simple and lucid primer (open source) on what precision is how it influences adrenergic and dopaminergic neuromodulatory systems to alter the synaptic gain afforded to top-down predictions and bottom-up evidence.:
Scientific thinking about the minds of humans and other animals has been transformed by the idea that the brain is Bayesian. A cornerstone of this idea is that agents set the balance between prior knowledge and incoming evidence based on how reliable or ‘precise’ these different sources of information are — lending the most weight to that which is most reliable. This concept of precision has crept into several branches of cognitive science and is a lynchpin of emerging ideas in computational psychiatry — where unusual beliefs or experiences are explained as abnormalities in how the brain estimates precision. But what precisely is precision? In this Primer we explain how precision has found its way into classic and contemporary models of perception, learning, self-awareness, and social interaction. We also chart how ideas around precision are beginning to change in radical ways, meaning we must get more precise about how precision works.

Friday, September 10, 2021

You Are Not Who You Think You Are

I want to point to the NYTimes piece by polymath David Brooks that does a nice summary of several themes that have been emphasized in MindBlog posts - on the recent work of Barrett, Friston, Hoffman, Johnson, and others. He touches on several areas in which our understanding of how our minds work has been completely transformed by work and ideas over just the past 10 years. I suggest you read the whole article and click on links to the articles he references. Here are a few clips:
You may think you understand the difference between seeing something and imagining it. When you see something, it’s really there; when you imagine it, you make it up. That feels very different...It turns out, reality and imagination are completely intermixed in our brain...the separation between our inner world and the outside world is not as clear as we might like to think.
...most of seeing is making mental predictions about what you expect to see, based on experience, and then using sensory input to check and adjust your predictions. Thus, your memory profoundly influences what you see...The conversation between senses and memory produces a “controlled hallucination,” which is the closest we can get to registering reality.
...humans have come up with all sorts of concepts to describe different thinking activities: memory, perception, emotion, attention, decision-making. But now, as scientists develop greater abilities to look at the brain doing its thing, they often find that the activity they observe does not fit the neat categories our culture has created, and which we rely on to understand ourselves...Barrett of Northeastern University argues that people construct emotions and thoughts, and there is no clear distinction between them...emotions assign value to things, so they are instrumental to reason, not separate from or opposed to it.
...there is no such thing as disembodied understanding. Your neural, chemical and bodily responses are in continual conversation with one another, so both understanding and experiencing are mental and physical simultaneously...When faced with a whole person...we shouldn’t think that they can be divided into a ‘mind’ and a ‘body.
...You realize that neither the term ‘decision-making’ nor the term ‘attention’ actually corresponds to a thing in the brain...the concepts at the core of how we think about thinking need to be radically revised...neuroscientists spent a lot of time trying to figure out what region of the brain did what function. (Fear is in the amygdala!) Today they also look at the ways vast networks across the brain, body and environment work together to create comprehensive mental states. Now there is much more emphasis on how people and groups creatively construct their own realities, and live within their own constructions.