One thing I didn't point out here, but is well worth pointing out: Currently, our academic system (which is supposed be a meritocracy) has far fewer ways to measure things like "aesthetics" or "intuitions." Therefore, potential great scientists get passed over for more competent grade grinders.
As a Humanities fella, I'd suggest that the language of "measurement" fits so poorly with things like "aesthetics" that the current academic parlance (which is obsessed with empiricism) can't recognize its value. As a philosopher like Byung-Chul Han might say, beauty resists quantification and thereby does not "produce." It's ROI is experiential, more like an encounter with a friend than a dividend.
In other words, "aesthetics" requires a different mode of seeing the world, one which great scientists bring alongside the mathematical (etc.) paradigm.
Wait, is it that our academic system has fewer ways to measure things like aesthetics or intutitions, or that there aren't as many ways (or not as many good ways) to measure them? If the latter, isn't the problem inherent to any large system for allocating resources?
I believe this is a worldwide problem. We have chosen to make a « game » that is easily trained for and indeed gamed in the name of meritocracy, instead of relying on the evaluation and judgment of successful peers in the field, under the pretense of potential favoritism.
The reality is that favoritism still happens, just more hidden, but you now have disqualified many very interesting people who cannot get the drive for the competitive grind.
I actually think being curious and driven to find new things is basically the complete opposite of caring about academic achievement. If Einstein life story is any indication, that pretty much confirms it.
True. There have been efforts to create metrics to assess research based on its ‘novelty’, bu my we need better ways to also assess ‘usefulness’ and/or ‘counterintuitiveness’ to allow us to better reward
Important work and to understand how much intuition and creativity are being encouraged vs stifled in science, and what is enabling or blocking it. Or, of course, to forgo quantitative evaluation and encourage more qualitative assessment - although that is harder to scale in a world of big science.
The word 'intuition" is used several times in this essay. I wonder if intuition is essentially an artifact of a rational process enabling the brain to build a mental model that generates insights.
It's a good question about what defines intuition vs. rationality. If you read, say, Pinker's book on rationality, most of it is about discernment. Given two ideas, how can you judge them? Or, given some statistical conundrum, how can we untangle it? But science requires discernment only at a later stage. The earlier demiurgic stuff requires creative generative ability, and I think ultimately rationality has little to say about that step.
I would say that intuition does not arise from a rational process but instead arises from our irrational subconscious can be updated/altered by a rational process.
I think sports are the best example of this - things happen in basketball so quickly that your actions and reactions have to be intuitive/subconscious. There just isn't enough time to reason. As such, the player's actions on the court are arising from subconscious mental maps of the state-action space. However, many players will watch film breakdowns and analyze their intuitive actions to arrive at a better mental model of how they should act in a given situation. Then, in the next game (after some practice to reinforce this new mental model), the player will intuitively act in accordance with the newly updated model produced by his rational thought process.
Sports is actually a great example of the burden of rationality. If you try to be to conscious of the process instead of « going with the flow », you will underperform. Which is why great trainers can make a big difference, they can give feedback on what to correct after the fact without you having to focus on it in the moment !
Yes precisely! Consciousness in the moment not only doesn't help, it actually severely harms your performance. Consciousness can only occur between games (when you're analyzing past performances and seeing where you can improve). The goal of practice then becomes hammering the results of these analyses into your subconscious so that you can use them automatically during the next game.
I would strongly suspect that the answer is no, intuition is nowhere near "an artifact of a rational process". I base this suspicion on paying attention to the work of a couple of people far smarter than I am, Dr Iain McGilchrist and Professor John Vervaeke.
My view of "intuition" was changed after reading Scott Young's book Ultralearning where he basically says, "Reliable intuition is built only with building expertise in an area."
And reading Ryan Bush's book Designing the Mind: The Principles of Psychitecture on becoming less bias where Bush suggests becoming aware of the most common biases people have and where are you most commonly use them by becoming aware of your automatic intuitions. He states our intuitions are just the current automatic responses our subconscious mind has generated based on our current knowledge and experience—so unless you have the knowledge and experience to make an intuitively good decision, your intuition is not reliable for the best outcome.
And reading Daniel Kahneman's book Thinking, Fast and Slow where Kahneman explains that intuitions are a "System 1" result (System 1 is basically our fast, reactive, emotional, and intuitive mind which is primarily initiated by the limbic system of the brain). So, intuitions are something that can vary by a person's knowledge or even their current mindset/sleep/hunger/etc. Kahneman also suggests that intuitions are best used in "well-defined problems" as described by Adam Mastroianni in this article: https://www.experimental-history.com/p/why-arent-smart-people-happier
I do believe what we call intuition is actually unconscious rationality. You come up with the answer effortlessly in the background by working with what you already know and produce a result that you cannot fully explain because you didn’t follow a specific process.
Usually you can work backwards and find the steps, even if it is hard and time consuming (that’s the science part in fact).
I think you're right about intuition being the output of insight-generating mental models, but I think the reason that the output of those models is largely in the form of feelings is that the models are pre-symbolic--the information contained can't be readily compressed, because it's for the purpose of generating novel compressions, if that makes sense. I don't expect we'd recognize the computations involved as reflecting anything currently considered under the umbrella of rational thought.
Edit: I can't remember where I came across this, maybe buried in one of the recent threads on Claude, but it was a paper showing that as LLM acquires a new concept, it updates the weights on a new model continuously in the 'right direction' but doesn't let that model contribute to the output until it's performance is sufficiently accurate, which may be akin to how intuition and insight work together.
i highly recommend the book ‘essence and alchemy’ by mandy aftel which dives into perfume making and its origins being founded by some of the deepest scientific minds of the time. it was thought of as the alchemy of that era! chemistry meets art meets beauty meets precision… i found it so satisfying and illuminating to realize how such a seemingly artistic and ethereal craft found its start with the most formulaic, curious and ambitious scientists and chemists of that period.
It's chemistry, though. You make compounds and see if they smell like something people like. It's only counterintuitive because of our cultural concepts of reason as something opposed to emotion or sensation.
James may have believed in ghosts, but he also believed in the exact thesis of this post!
“The eternal verities which the very structure of our mind lays hold of...stand waiting in the mind, forming a beautiful ideal network; and the most we can say is that we hope to discover realities over which the network may be flung so that ideal and real may coincide."
Fantastic perspective. I LOVE this concept of plucking world-shaking ideas “from some sort of netherworld of pre-scientific intuition”! That delicious gray area is probably why I enjoy writing scientists in many of my storytelling/screenplay projects. 🤔
Perhaps no revolutions in theoretical physics, but I’d call the rise of complex systems theory a revolution. It’s brought major new insights to non-equilibrium chemistry, theoretical biology, ecology, economics, and a bunch of other fields, and created a whole new class of modeling techniques.
I highly recommend that you check out "The Reenchantment of the World" by Morris Berman, who got his PhD in the history of science. He makes a distinction between the "participatory consciousness" of the pre-scientific world and the detached "observational consciousness" that developed with the scientific method which came to dominate the West. We may look back at Sir Isaac's interest in alchemy as kooky, but Berman posits that the modern mind can't understand what alchemists were trying to achieve, nor how they saw the world. Interestingly, much of the creative imagery of alchemy was immediately recognized by Jung as capturing deep archetypal concepts. Berman maps out the sad trajectory of Sir Isaac as he went from a creative, lively and imaginative young man to depressed older icon, attempting to keep some sense of magic alive in his secret writings because he could see that science was quickly dis-enchanting the world with its cold, observational stare. When you look at Newton's portraits through his life, you can see this very visually.
My completely unfounded (intuitive?) feeling on this is that intelligence doesn’t necessarily help with idea generation.
I’ve noticed some people have this really uncanny ability to connect seemingly unconnected concepts or ideas. Sometimes they’re smart, sometimes not.
But the magic combination comes when someone has the ability to generate a lot of ideas, and then switch gears and rationally or aesthetically pare them down. Being able to hook both of these modes up into a single feedback loop is what enables some scientists, architects, and artists to create novel and substantial work.
I think of it as the Two-Phase “Brian Eno” model of creativity. The first phase is just about synthesis: doing whatever it takes to generate new ideas. The second phase is analysis: a rigorous process that sculpts those ideas into something that works — either musically, artistically, or scientifically.
I’m not convinced that there is a discrepancy between rationality and aesthetics or intuition (perhaps a better notion here would be comprehension of observations, or ideal abstract reasoning), and similarly that there is no discrepancy between Popper and Kuhn’s premise of what defines science as such. I consider this in support of your argument, that aesthetics and intuition are both important - if scientists were to only confine themselves to mechanical operations of rational testing, how could any holistic situation of science be possible, if aesthetics and intuition are not similarly rational?
Popper and Khun were concerned with different aspects of what defines science that are not incompatible. I think Popper was right to identify falsifiability as a scientific criterion, and demarcation of science as such that it is certainly requires identifying it as constituent of testing that is capable of being accurate and inaccurate according to the framework being used. When analyzing and defining science on a social scale, it becomes clear that some scientific frameworks may be more or less accurate, leading to the observation that there are paradigms in which considerations are toiled with scientifically (in the mechanical sense), but which may eventually be subsumed under a more accurate framework (in the falsifiable sense), to latter be toiled under.
Demarcating science as a phenomenon is necessarily a multivariate process, one that involves ontological, metaphysical, dialectical, material, cognitive, psychological, rational, and sociological considerations. That’s what makes the process so difficult. Regardless, we can still appreciate that science is not *strictly* confined to rationality as an isolated processes for this reason. Using a similar point of consideration, aesthetics and intuition (if we are to understand it as the skill of processing observations and considering them as abstract phenomena that we can derive meaning from) are not precluded from being rational processes, since they are subsumed under empirical experience and are capable of being subject to testing in the way matters of science are.
Einstein’s ability to incorporate aesthetic considerations and observational insights into his scientific testing does not exempt the process of doing so as being rational considerations themselves. It is then a scientific challenge to determine if such considerations are accurate.
If you take “rationality” to be synonymous with things like logic or inference, then yes this post is spot on. The best scientists (or best anything, really) are not Captain Spock like characters who are perfectly “rational”, logical computing machines. Insights require actually letting your “system 1” thinking override your “system 2” - we’re taught that leaping to conclusions is not proper and many times in science it is not, but in order to make progress quickly or synthesize things creatively, sometimes exactly what we need is a leap of some sort. This would be akin to the distinction between knowledge and wisdom, where purely propositional and logical knowledge is different than the creative and contextually sensitive application of that knowledge. Very interesting post!
Yes indeed. I believe that Isaac Asimov, a great scientist and a marvellous science fiction author said “ Whatever the mind of man can imagine, man can create.” ( this may be attributed incorrectly, but Asimov is my hero, or one if them).Great scientists need huge, over arching imaginations. Before you can make something, anything from art to aphorisms, you have first to imagine it, to see it in your mind as it were. Then the genius lies in the ability to do something with and about those imaginings.
In the 19th century, people believed in a lumineferous ether, a medium of light. Einstein showed that you could not measure your speed with respect to the ether, and therefore it is superfluous. Using positivism, he concludes that the ether is an unobservable entity, and therefore is meaningless. The ether was disproved in this way, the major step was positivism.
Ernst Mach developed philosophical positivism, along with important work in physics (the Mach number is named after him, since he studied supersonic fluid flow). Einstein admired and respected Mach, and made bold use of positivism in the early years. "
"Quantum Mechanics (1920s)
In quantum mechanics, positivism is the most important ingredient. The act of defining a measuring device and a measurement as a separate aspect of the theory from the time evolution was due to the explicitly positivist outlook of Bohr and Heisenberg.
The major idea of making a matrix description of the electron came from renouncing the idea of a microscopic well defined orbit, in favor of quantities which determine the atomic transitions, which are all that we can observe. As Pauli and Heisenberg argued, if you can't determine where the electron is when it is in the ground state of Hydrogen, on the left of the nucleus or on the right, even in principle (at least not without knocking it out of the ground state), then this question is meaningless, and the location of the electron in its orbit in the ground state is meaningless.
The positivism is still central to the interpretation of the most common statement of the uncertainty principle. Because we cannot measure the position and momentum simultaneously, these quantities do not exist simultaneously. The modern formalism does not include position and momentum, but a wavefunction which gives probability amplitudes for either. "
Ron Maimon is certainly one of the most interesting characters I've ever come across online. He was a genius who possessed a staggering amount of knowledge in various fields(physics, math, theoretical biology, linguistics, programming, philosophy). And he had the personality to match the genius. The best description I've heard of him was that he could've been a character in a Pynchon novel.
Eventually his brash personality combined with his complete disregard for authority would lead him to getting banned from every site he was active on. He was banned on PhysicsStackExchange for political reasons. The mods were more concerned with users being nice to each other rather than focusing on the quality and correctness of their answers. He was banned on Quora because he kept on pushing the conspiracy theory that the Boston Marathon Bombing was a false flag. He was also a 9/11 truther(unrelated to any ban). And he also thought Shakespeare was Marlow. And that oil is abiogenic. I guess beyond a certain point of intelligence you lose trust in other peoples capability to produce correct and honest work which leads to these irrational beliefs that these geniuses sometimes have.
I find it HIGHLY unlikely that logical positivism had anything to do with fundamental insight in physics. It would be marginally credible that it had something to do with how physicists of the last 70 years have _framed_ received physical theories. I seriously doubt that most theoretical physicists bothered at all with logical positivism, as it was entirely unnecessary to their discipline and thought processes, logical though those might be.
Erik, the respective places of rational and irrational understanding in science have a long and checkered history, as you know. Ratio-nality implies a logical chain of reasoning, ideally a metrical quality to both thought and subject, with the irrational lacking such an orderly measurable process sequence. As you state, leaps of genius seldom follow such orderly chains of derivation, and hence tend to be classed as irrational, if poorly so and by default. Having done a study of genius in years gone by myself and having known a few with that quality over the years, I suggest a more specific perspective on the nature of transformational insight.
The term intuition is inaptly archaic, and needs to be retired. It is vague and implies a mystical capacity, or more generously a paranormal one. As I perceive it, insight of genius in science or any field involves a high standard deviation capacity for pattern recognition, or in more basic scope for pattern identification. What is spoken to here is the capacity to perceive discrete coherent relationship(s) within a mass of information. Or separately or in parallel, a capacity to perceive coherent relationship(s) between bodies of information which are not then commonly understood as conjunct or interactive, such that the organization in one draws out coherent order in the other. Rational analysis which dominates in science cannot produce theoretical insight other than by chance, because rational perception requires a subject/data where coherent order is already understood to exist, and usually whose order is at least crudely mapped. Finding metrical coherence in raw data is the needle in a haystack problem; or better the iron filings in a sand dune problem. Pattern recognition by contrast finds the veiled pattern in the known and in the unknown alike because it draws the mind’s attention to hitherto unperceived yet orderly relationships.
Pattern recognition is a straightforward cognitive process, if one absurdly poorly studied at present. It is an inherent operation in organic neurological structures, and to my mind the key to understanding how those structures operate and why they have evolved as they have. Leaving that delicious subject aside here, human higher neurological structures are optimized to yield pattern identification, but like any other genetically distributed capacity some individuals do this with more and some with less facility. Those who make intuitive leaps in science which are borne out tend to have more of this facility than others. In this sense, no one is ‘a genius’ even while some people have genius, operate ingeniously; it is a specific trait of cognition and higher thought, not a transcendental nimbus of khvarenah bestowed by fortune or some deity. Pattern identification does NOT operate by logical, metrical chains, or codes, or formulae. It is evidently more an analog resolution of coherent order of form as such and only as such, whether or not and definitely before any analytical mensuration or transcription is undertaken.
I concur with von Neumann’s self-assessment; he had tremendous scope of rational derivation but modest capacity at pattern identification at best, certainly relative to Einstein’s sweeping and repeated facility with pattern recognition. It is entirely unsurprising that von Neumann’s main contributions are in rule-based game theory and code-based cybernetic simulation. The insights there have been theoretically meagre compared to the pattern insights of Einstein, Poincaré, Maxwell, Gauss, Liebniz, or Newton. Mathematical or otherwise, analytic simulation is NOT coherence perception, and indeed little involves the latter at all; they are separate cognitive processes, as I understand this, with neither primarily dependent upon the other.
Framed this way, the strengths and failures of pattern recognition in insight are readily understandable, both in science and elsewhere. To begin with, the convergence of an aesthetic sensibility with a capacity for pattern identification in an individual is natural and probable. The aesthetic appeal of something very much lies in its possession of a quality of overall coherence, a ‘tensile integrity’ of form or self-relationship as opposed to a chaos or hodge podge of features. A logarithmic table, a Feynman diagram, or a work by Picasso all have deep and inherent aesthetic appeal whereas a sack of nails, the detritus in a sink trap, or a random YouTube nondescriptive description may not. It is not a coincidence but an integral feature that those whose pattern recognition insight is of an order of genius are often drawn to qualities of aesthetic appeal in other media or activities. Those with a high order of pattern insight tend to be pre-attuned to find coherent order and resonate with it, wherever and in whatever that coherence may operate, whether or not they have any specific areal training to analyze or reproduce those other, specific, orderly quanta or qualities. This last is not an invariable capacity, for cognitive extremes and/or deficits for a given individual can impede perceptions of aesthetic coherence in diverse information—or enhance such perception in only a few, restricted bodies of information.
Understood as pattern recognition, the seemingly irrational deviations pursued by some who have shown genius insight pertaining to a prior specific range of information are entirely explicable; indeed, these divergent lines of investigation should be anticipated. For pattern recognition is an autonomous cognitive process. An individual will perceive a pattern in information, whether or not there is any actual coherent order there. The mind can impose the perception of pattern as well as perceive its actuality. Pattern identification is inherently prone to false positives. Those who have had the aesthetic delight of profound pattern insight can tend to overestimate their capacity to discern pattern elsewhere. Or can fail to do the attendant reality testing analysis to substantiate a further perception of coherent order, or simply not succeed in analytically substantiating a perceived resolution or orderly relationship. The cognitive capacity to perceive an orderly relationship does not necessarily imply the analytic ability to render that insight firm or analytically useful. The discipline or inclination or simple capacity to substantiate a further perception of coherent order may wander, especially later in life. Josephson may be onto something which could ultimately be substantiated. Hoyle most definitely is not. Einstein’s later theories have in their large part fallen short too. That’s how it goes.
Pattern identification as the driver to theoretical insight firms up and sites the place of mathematics in scientific investigation: confirmation, not inspiration. Math is coherent, and via congruency can flesh out and ‘link in’ coherent organization newly perceived as such. Math can transcribe the order in insight into forms which not only can be utilized but which can be extended beyond their initial core perception(s) of relationship. Great tools, in short—but not the first, aesthetic, vision or relationship. To be cognitively rendered at all, perceptions of different coherence in information in fact often have to be made at some remove from the rigidity and weld-tight relationship of what is already conceived but whose very formal logic and data exclusivity impede a perception of anything else or other. Furthermore, the math is not the pattern, it is a code which draws on the pattern. Worse, mathematical rendering of insight and even observation often truncates the available information, excising putative anomalies or too much complexity, ‘making the math come out’ even if the result is distorted in lesser (or greater) degree from what is perceived as pattern or even received in cumulative but inconvenient data. Thinkers who operate with high order insight will seldom begin from the math even if they often return to it, for the cognitive operation of mathematics scarcely overlaps at all with the cognitive perception of coherent pattern; different and differently comprehensive readings of an object of interest.
The high value of any pre-theoretical fringe so-called in science and order becomes evident when and as theoretical insight is understood as pattern identification. Pre-theoretical bodies of information are rich with substantive observation, fragmentary analysis, and moderate compatibility with what is already known. They are saturated with organization but their dimensions and/or networks of coherent order are not yet perceived as such. They are thus not random bodies of fact and data which are inherently less likely to possess internal avenues of coherence but rather relevant, meaning-dense assemblages of information more than typically likely to have as yet unperceived coherent order(s) striated amongst their features; they are known unknowns, in a phrase. It is much easier to identify coherence in pattern rich information than in or amongst pattern thin or evidently random information. This is why ancient alien theory is a sad pursuit: evidence poor and highly prone to false positive perceptions. This is why organic life on Mars theory is likely to be a more substantive pursuit: data gaining density and the subject constrains false positives of coherent pattern.
Scientific teams make pattern recognition insight nearly impossible. It’s like the convoy system, comprehensible process is chained to the slowest and most partial perception. High standard deviation cognitive pattern identification is, by definition, not an evenly or widely distributed capacity, but such sparks as may occur in a team of a dozen or several dozen researchers fail to propagate at best or are smothered by the sheer density of process in already understood modalities of data framing and analysis. Asymmetric capacities, the buffering effect of limited cognitive perception, and the constraints of inapt or inapplicable prior analytic formalizations make improbable theoretical leaps by bunches of researchers. How many primary theoretical insights have involved even three individuals? Two is not uncommon; one is the norm. 4+ = 0, I think anyone looking at this will find; 6+ most definitely. The herd behavior increasingly prevalent in modern science to me has a pronounced correlation with the steady decline of theoretical insight in any field. This is principally why for instance physics and neuroscience are theoretically stalled at this time, and have been for decades really. Mobs can’t perceive coherent order, and the individuals who can can’t get heard in a crowd or even think clearly there. One can see the same process in poetry, for example, where both aesthetic felicity and originality of vision are presently in total collapse.
Then too, AI does not have native pattern recognition, which is specifically why it is not ‘intelligent.’ The present faulty approach in AI is to use multiple server farm levels of data crunching to generate pattern matching by massive brute force congruency . . . but with zero actual registration of pattern order at all by the inorganic nets. What is achieved is a reading that ‘a something’ is there which might fit ‘another something.’ This is von Neumann’s Sorrow to an exponential magnitude. This is a six-month-old’s degree of pattern identification without even a six-month-old’s quality of pattern recognition, for the human neurological apparatus comes with a degree of hardwired and firmwired pattern recognition capacities whose use can be built out and whose fluency of result can be refined. AI cannot find a partial motorcycle or tell you on its own why a woman is beautiful, only mimic a statement of beauty culled from the web sans comprehension of the coherent patterns represented by the alphanumeric characters or matrix of pixels.
I despise Popper, by the way. His philosophical perspective has zero to do with science, is functionally useless in that field, and no one employs it or ever has. It was a rhetorical sleight of hand to get Marxist dialectics and other ‘fringe theories’ banned from the academy, nothing more, which worked in Britain and America but not elsewhere. It is still taught to students of science in higher education not because it is useful or even intellectually viable, it is not, but to scare them into pulling in their horns and not embarrassing their seniors with startlingly new and disruptive insights which would sweep aside many senior reputations and comfortable livings. “If you can’t prove to me how you’re wrong, we will treat you as if you are—so do as I do and stick to what we know.” The Triumph of the Mediocratic Society . . . .
Perhaps the stupidity of large teams as far as pattern recognition is because they don't get the algorithm right. The individual brains would need to act in service to the team, with some functions of the individual brains' higher levels being dropped and being re-allocated to one of them, or taken up by the team (not sure how that works). What do you mean by "brute force congruency"?
Following Erik's usage of congruency of the post, my reference is to rough fitting of grossly comparable value readings whose actual states are not 'understood' or necessarily relevant to neural net pattern fitting; inexact but with sufficient similarity to make a workable match but lacking any comprehension even of the the states compared, only that they 'sort of fit.'
I do not think that 'algorithm' is at all a good analogy for anything that happens in neural processing. My view. Algorithims are codes, of elements with defined value ranges. I do not think that organic neural processing works in any way by codes, and elements as rendered in organic cognition have prototypal and labile value ranges as I envision this, working value states which evolve as they are used. There is a great deal more to this argument, but I'll leave it there.
Regarding a topological argument, I do think topology applies to neural processing very much, yes, but the causal vector is the other way round than you stated, to me: neural response topologically renders some derived aspect of the topological matrix of nature experienced. It is a 'reading' though by no means an accurate of 1:1 one, and certainly not an 'algorithmic derivation' in a formal sense. Ideas have geometry, literally so, because organic cognition renders readings in real dimensional neural tissue. I suppose if you use the term algorithm loosely, any prototype sequence evolution could be called that, but I do not think that is a primary or desirable usage of that particular semantic concept.
I have my own toy model of a General Theory of Physics with indications of what the physical residuals presently termed fermions and bosons might be. But that is an argument very distant from this comment thread.
1. When it happens well, the pattern existing in nature is replicated in the brain, but with a different substrate. Relationships are the same or similar, but among different types of parts in the brain vs in nature. It doesn't have to be this way. The algorithm doesn't have to be mimicking, it could be (as it is with AI) a multinomial logistic regression.
2. nature in some specific ways has a similar topology and process to brains. For example, in lattice gauge theory (and in a different way in Wolfram hyper-graph theory), fermions are nodes and bosons are edges connecting nodes of a 3D lattice, each with its own states. Sounds like a neural network. But nature, unlike the brain, is not trying to model or predict anything, as far as we know.
Physicist Sabine Hossenfelder wrote a book called, "Lost in Math: How Beauty Leads Physics Astray". She argues that theoretical physicists have led themselves astray by following their aesthetic preference for beautiful, elegant theories rather than being guided primarily by experimental evidence. I confess that I haven't read the book, but it's on my list. Just curious if you (Erik) have an opinion about Hossenfelder's thesis.
One thing I didn't point out here, but is well worth pointing out: Currently, our academic system (which is supposed be a meritocracy) has far fewer ways to measure things like "aesthetics" or "intuitions." Therefore, potential great scientists get passed over for more competent grade grinders.
Loved this article!
As a Humanities fella, I'd suggest that the language of "measurement" fits so poorly with things like "aesthetics" that the current academic parlance (which is obsessed with empiricism) can't recognize its value. As a philosopher like Byung-Chul Han might say, beauty resists quantification and thereby does not "produce." It's ROI is experiential, more like an encounter with a friend than a dividend.
In other words, "aesthetics" requires a different mode of seeing the world, one which great scientists bring alongside the mathematical (etc.) paradigm.
Wait, is it that our academic system has fewer ways to measure things like aesthetics or intutitions, or that there aren't as many ways (or not as many good ways) to measure them? If the latter, isn't the problem inherent to any large system for allocating resources?
"potential great scientists get passed over for more competent grade grinders"
As an academic scientist for 40 years, I don't believe this is true.
I believe this is a worldwide problem. We have chosen to make a « game » that is easily trained for and indeed gamed in the name of meritocracy, instead of relying on the evaluation and judgment of successful peers in the field, under the pretense of potential favoritism.
The reality is that favoritism still happens, just more hidden, but you now have disqualified many very interesting people who cannot get the drive for the competitive grind.
I actually think being curious and driven to find new things is basically the complete opposite of caring about academic achievement. If Einstein life story is any indication, that pretty much confirms it.
True. There have been efforts to create metrics to assess research based on its ‘novelty’, bu my we need better ways to also assess ‘usefulness’ and/or ‘counterintuitiveness’ to allow us to better reward
Important work and to understand how much intuition and creativity are being encouraged vs stifled in science, and what is enabling or blocking it. Or, of course, to forgo quantitative evaluation and encourage more qualitative assessment - although that is harder to scale in a world of big science.
I'm not so sure about 'intuitions which has become an extremely popular filed of study. Aesthetics I agree...
The word 'intuition" is used several times in this essay. I wonder if intuition is essentially an artifact of a rational process enabling the brain to build a mental model that generates insights.
It's a good question about what defines intuition vs. rationality. If you read, say, Pinker's book on rationality, most of it is about discernment. Given two ideas, how can you judge them? Or, given some statistical conundrum, how can we untangle it? But science requires discernment only at a later stage. The earlier demiurgic stuff requires creative generative ability, and I think ultimately rationality has little to say about that step.
I would say that intuition does not arise from a rational process but instead arises from our irrational subconscious can be updated/altered by a rational process.
I think sports are the best example of this - things happen in basketball so quickly that your actions and reactions have to be intuitive/subconscious. There just isn't enough time to reason. As such, the player's actions on the court are arising from subconscious mental maps of the state-action space. However, many players will watch film breakdowns and analyze their intuitive actions to arrive at a better mental model of how they should act in a given situation. Then, in the next game (after some practice to reinforce this new mental model), the player will intuitively act in accordance with the newly updated model produced by his rational thought process.
Sports is actually a great example of the burden of rationality. If you try to be to conscious of the process instead of « going with the flow », you will underperform. Which is why great trainers can make a big difference, they can give feedback on what to correct after the fact without you having to focus on it in the moment !
Yes precisely! Consciousness in the moment not only doesn't help, it actually severely harms your performance. Consciousness can only occur between games (when you're analyzing past performances and seeing where you can improve). The goal of practice then becomes hammering the results of these analyses into your subconscious so that you can use them automatically during the next game.
I would strongly suspect that the answer is no, intuition is nowhere near "an artifact of a rational process". I base this suspicion on paying attention to the work of a couple of people far smarter than I am, Dr Iain McGilchrist and Professor John Vervaeke.
My view of "intuition" was changed after reading Scott Young's book Ultralearning where he basically says, "Reliable intuition is built only with building expertise in an area."
And reading Ryan Bush's book Designing the Mind: The Principles of Psychitecture on becoming less bias where Bush suggests becoming aware of the most common biases people have and where are you most commonly use them by becoming aware of your automatic intuitions. He states our intuitions are just the current automatic responses our subconscious mind has generated based on our current knowledge and experience—so unless you have the knowledge and experience to make an intuitively good decision, your intuition is not reliable for the best outcome.
And reading Daniel Kahneman's book Thinking, Fast and Slow where Kahneman explains that intuitions are a "System 1" result (System 1 is basically our fast, reactive, emotional, and intuitive mind which is primarily initiated by the limbic system of the brain). So, intuitions are something that can vary by a person's knowledge or even their current mindset/sleep/hunger/etc. Kahneman also suggests that intuitions are best used in "well-defined problems" as described by Adam Mastroianni in this article: https://www.experimental-history.com/p/why-arent-smart-people-happier
Great post, Erik.
I do believe what we call intuition is actually unconscious rationality. You come up with the answer effortlessly in the background by working with what you already know and produce a result that you cannot fully explain because you didn’t follow a specific process.
Usually you can work backwards and find the steps, even if it is hard and time consuming (that’s the science part in fact).
I think you're right about intuition being the output of insight-generating mental models, but I think the reason that the output of those models is largely in the form of feelings is that the models are pre-symbolic--the information contained can't be readily compressed, because it's for the purpose of generating novel compressions, if that makes sense. I don't expect we'd recognize the computations involved as reflecting anything currently considered under the umbrella of rational thought.
Edit: I can't remember where I came across this, maybe buried in one of the recent threads on Claude, but it was a paper showing that as LLM acquires a new concept, it updates the weights on a new model continuously in the 'right direction' but doesn't let that model contribute to the output until it's performance is sufficiently accurate, which may be akin to how intuition and insight work together.
I contend that rationality is an artifact of intuition.
i love this topic and these observations.
i highly recommend the book ‘essence and alchemy’ by mandy aftel which dives into perfume making and its origins being founded by some of the deepest scientific minds of the time. it was thought of as the alchemy of that era! chemistry meets art meets beauty meets precision… i found it so satisfying and illuminating to realize how such a seemingly artistic and ethereal craft found its start with the most formulaic, curious and ambitious scientists and chemists of that period.
It's chemistry, though. You make compounds and see if they smell like something people like. It's only counterintuitive because of our cultural concepts of reason as something opposed to emotion or sensation.
James may have believed in ghosts, but he also believed in the exact thesis of this post!
“The eternal verities which the very structure of our mind lays hold of...stand waiting in the mind, forming a beautiful ideal network; and the most we can say is that we hope to discover realities over which the network may be flung so that ideal and real may coincide."
Fantastic perspective. I LOVE this concept of plucking world-shaking ideas “from some sort of netherworld of pre-scientific intuition”! That delicious gray area is probably why I enjoy writing scientists in many of my storytelling/screenplay projects. 🤔
Do you think an overvaluing of rationality could be part of the reason we haven’t had any major scientific revolutions since the early 20th century?
I'll have a piece on essentially this, although I think it's part of a larger issue with academia.
Perhaps no revolutions in theoretical physics, but I’d call the rise of complex systems theory a revolution. It’s brought major new insights to non-equilibrium chemistry, theoretical biology, ecology, economics, and a bunch of other fields, and created a whole new class of modeling techniques.
I highly recommend that you check out "The Reenchantment of the World" by Morris Berman, who got his PhD in the history of science. He makes a distinction between the "participatory consciousness" of the pre-scientific world and the detached "observational consciousness" that developed with the scientific method which came to dominate the West. We may look back at Sir Isaac's interest in alchemy as kooky, but Berman posits that the modern mind can't understand what alchemists were trying to achieve, nor how they saw the world. Interestingly, much of the creative imagery of alchemy was immediately recognized by Jung as capturing deep archetypal concepts. Berman maps out the sad trajectory of Sir Isaac as he went from a creative, lively and imaginative young man to depressed older icon, attempting to keep some sense of magic alive in his secret writings because he could see that science was quickly dis-enchanting the world with its cold, observational stare. When you look at Newton's portraits through his life, you can see this very visually.
You continue to be one of the best writers on here. Keep up the good work!
My completely unfounded (intuitive?) feeling on this is that intelligence doesn’t necessarily help with idea generation.
I’ve noticed some people have this really uncanny ability to connect seemingly unconnected concepts or ideas. Sometimes they’re smart, sometimes not.
But the magic combination comes when someone has the ability to generate a lot of ideas, and then switch gears and rationally or aesthetically pare them down. Being able to hook both of these modes up into a single feedback loop is what enables some scientists, architects, and artists to create novel and substantial work.
I think of it as the Two-Phase “Brian Eno” model of creativity. The first phase is just about synthesis: doing whatever it takes to generate new ideas. The second phase is analysis: a rigorous process that sculpts those ideas into something that works — either musically, artistically, or scientifically.
I'd agree with this. There's a ton of intelligent people who don't seem to generate very much that's original.
I’m not convinced that there is a discrepancy between rationality and aesthetics or intuition (perhaps a better notion here would be comprehension of observations, or ideal abstract reasoning), and similarly that there is no discrepancy between Popper and Kuhn’s premise of what defines science as such. I consider this in support of your argument, that aesthetics and intuition are both important - if scientists were to only confine themselves to mechanical operations of rational testing, how could any holistic situation of science be possible, if aesthetics and intuition are not similarly rational?
Popper and Khun were concerned with different aspects of what defines science that are not incompatible. I think Popper was right to identify falsifiability as a scientific criterion, and demarcation of science as such that it is certainly requires identifying it as constituent of testing that is capable of being accurate and inaccurate according to the framework being used. When analyzing and defining science on a social scale, it becomes clear that some scientific frameworks may be more or less accurate, leading to the observation that there are paradigms in which considerations are toiled with scientifically (in the mechanical sense), but which may eventually be subsumed under a more accurate framework (in the falsifiable sense), to latter be toiled under.
Demarcating science as a phenomenon is necessarily a multivariate process, one that involves ontological, metaphysical, dialectical, material, cognitive, psychological, rational, and sociological considerations. That’s what makes the process so difficult. Regardless, we can still appreciate that science is not *strictly* confined to rationality as an isolated processes for this reason. Using a similar point of consideration, aesthetics and intuition (if we are to understand it as the skill of processing observations and considering them as abstract phenomena that we can derive meaning from) are not precluded from being rational processes, since they are subsumed under empirical experience and are capable of being subject to testing in the way matters of science are.
Einstein’s ability to incorporate aesthetic considerations and observational insights into his scientific testing does not exempt the process of doing so as being rational considerations themselves. It is then a scientific challenge to determine if such considerations are accurate.
If you take “rationality” to be synonymous with things like logic or inference, then yes this post is spot on. The best scientists (or best anything, really) are not Captain Spock like characters who are perfectly “rational”, logical computing machines. Insights require actually letting your “system 1” thinking override your “system 2” - we’re taught that leaping to conclusions is not proper and many times in science it is not, but in order to make progress quickly or synthesize things creatively, sometimes exactly what we need is a leap of some sort. This would be akin to the distinction between knowledge and wisdom, where purely propositional and logical knowledge is different than the creative and contextually sensitive application of that knowledge. Very interesting post!
Yes indeed. I believe that Isaac Asimov, a great scientist and a marvellous science fiction author said “ Whatever the mind of man can imagine, man can create.” ( this may be attributed incorrectly, but Asimov is my hero, or one if them).Great scientists need huge, over arching imaginations. Before you can make something, anything from art to aphorisms, you have first to imagine it, to see it in your mind as it were. Then the genius lies in the ability to do something with and about those imaginings.
Ron Maimon attributes the great 20th century physics discoveries to logical positivism.
https://philosophy.stackexchange.com/questions/1031/is-there-any-philosophical-idea-that-radically-changed-the-world-in-the-twentiet/2520#2520
Excerpts:
"Relativity (1900s)
In the 19th century, people believed in a lumineferous ether, a medium of light. Einstein showed that you could not measure your speed with respect to the ether, and therefore it is superfluous. Using positivism, he concludes that the ether is an unobservable entity, and therefore is meaningless. The ether was disproved in this way, the major step was positivism.
Ernst Mach developed philosophical positivism, along with important work in physics (the Mach number is named after him, since he studied supersonic fluid flow). Einstein admired and respected Mach, and made bold use of positivism in the early years. "
"Quantum Mechanics (1920s)
In quantum mechanics, positivism is the most important ingredient. The act of defining a measuring device and a measurement as a separate aspect of the theory from the time evolution was due to the explicitly positivist outlook of Bohr and Heisenberg.
The major idea of making a matrix description of the electron came from renouncing the idea of a microscopic well defined orbit, in favor of quantities which determine the atomic transitions, which are all that we can observe. As Pauli and Heisenberg argued, if you can't determine where the electron is when it is in the ground state of Hydrogen, on the left of the nucleus or on the right, even in principle (at least not without knocking it out of the ground state), then this question is meaningless, and the location of the electron in its orbit in the ground state is meaningless.
The positivism is still central to the interpretation of the most common statement of the uncertainty principle. Because we cannot measure the position and momentum simultaneously, these quantities do not exist simultaneously. The modern formalism does not include position and momentum, but a wavefunction which gives probability amplitudes for either. "
It's an interesting connection I hadn't heard about before. Einstein was also a fan of a ton of different philosophies, especially Spinoza.
"Einstein said about himself: "As a young man I preferred books whose content concerned a whole world view and, in particular, philosophical ones. Schopenhauer, David Hume, Mach, to some extent Kant, Plato, Aristotle."" https://en.wikipedia.org/wiki/Religious_and_philosophical_views_of_Albert_Einstein
Ron Maimon is certainly one of the most interesting characters I've ever come across online. He was a genius who possessed a staggering amount of knowledge in various fields(physics, math, theoretical biology, linguistics, programming, philosophy). And he had the personality to match the genius. The best description I've heard of him was that he could've been a character in a Pynchon novel.
An interview with him: https://www.youtube.com/watch?v=ObXbKbpkSjQ
Eventually his brash personality combined with his complete disregard for authority would lead him to getting banned from every site he was active on. He was banned on PhysicsStackExchange for political reasons. The mods were more concerned with users being nice to each other rather than focusing on the quality and correctness of their answers. He was banned on Quora because he kept on pushing the conspiracy theory that the Boston Marathon Bombing was a false flag. He was also a 9/11 truther(unrelated to any ban). And he also thought Shakespeare was Marlow. And that oil is abiogenic. I guess beyond a certain point of intelligence you lose trust in other peoples capability to produce correct and honest work which leads to these irrational beliefs that these geniuses sometimes have.
I find it HIGHLY unlikely that logical positivism had anything to do with fundamental insight in physics. It would be marginally credible that it had something to do with how physicists of the last 70 years have _framed_ received physical theories. I seriously doubt that most theoretical physicists bothered at all with logical positivism, as it was entirely unnecessary to their discipline and thought processes, logical though those might be.
thanks for this.
a nice adventure into that very interesting space where european and analytic philos keep bobbing around each other in the ether.
Erik, the respective places of rational and irrational understanding in science have a long and checkered history, as you know. Ratio-nality implies a logical chain of reasoning, ideally a metrical quality to both thought and subject, with the irrational lacking such an orderly measurable process sequence. As you state, leaps of genius seldom follow such orderly chains of derivation, and hence tend to be classed as irrational, if poorly so and by default. Having done a study of genius in years gone by myself and having known a few with that quality over the years, I suggest a more specific perspective on the nature of transformational insight.
The term intuition is inaptly archaic, and needs to be retired. It is vague and implies a mystical capacity, or more generously a paranormal one. As I perceive it, insight of genius in science or any field involves a high standard deviation capacity for pattern recognition, or in more basic scope for pattern identification. What is spoken to here is the capacity to perceive discrete coherent relationship(s) within a mass of information. Or separately or in parallel, a capacity to perceive coherent relationship(s) between bodies of information which are not then commonly understood as conjunct or interactive, such that the organization in one draws out coherent order in the other. Rational analysis which dominates in science cannot produce theoretical insight other than by chance, because rational perception requires a subject/data where coherent order is already understood to exist, and usually whose order is at least crudely mapped. Finding metrical coherence in raw data is the needle in a haystack problem; or better the iron filings in a sand dune problem. Pattern recognition by contrast finds the veiled pattern in the known and in the unknown alike because it draws the mind’s attention to hitherto unperceived yet orderly relationships.
Pattern recognition is a straightforward cognitive process, if one absurdly poorly studied at present. It is an inherent operation in organic neurological structures, and to my mind the key to understanding how those structures operate and why they have evolved as they have. Leaving that delicious subject aside here, human higher neurological structures are optimized to yield pattern identification, but like any other genetically distributed capacity some individuals do this with more and some with less facility. Those who make intuitive leaps in science which are borne out tend to have more of this facility than others. In this sense, no one is ‘a genius’ even while some people have genius, operate ingeniously; it is a specific trait of cognition and higher thought, not a transcendental nimbus of khvarenah bestowed by fortune or some deity. Pattern identification does NOT operate by logical, metrical chains, or codes, or formulae. It is evidently more an analog resolution of coherent order of form as such and only as such, whether or not and definitely before any analytical mensuration or transcription is undertaken.
I concur with von Neumann’s self-assessment; he had tremendous scope of rational derivation but modest capacity at pattern identification at best, certainly relative to Einstein’s sweeping and repeated facility with pattern recognition. It is entirely unsurprising that von Neumann’s main contributions are in rule-based game theory and code-based cybernetic simulation. The insights there have been theoretically meagre compared to the pattern insights of Einstein, Poincaré, Maxwell, Gauss, Liebniz, or Newton. Mathematical or otherwise, analytic simulation is NOT coherence perception, and indeed little involves the latter at all; they are separate cognitive processes, as I understand this, with neither primarily dependent upon the other.
Framed this way, the strengths and failures of pattern recognition in insight are readily understandable, both in science and elsewhere. To begin with, the convergence of an aesthetic sensibility with a capacity for pattern identification in an individual is natural and probable. The aesthetic appeal of something very much lies in its possession of a quality of overall coherence, a ‘tensile integrity’ of form or self-relationship as opposed to a chaos or hodge podge of features. A logarithmic table, a Feynman diagram, or a work by Picasso all have deep and inherent aesthetic appeal whereas a sack of nails, the detritus in a sink trap, or a random YouTube nondescriptive description may not. It is not a coincidence but an integral feature that those whose pattern recognition insight is of an order of genius are often drawn to qualities of aesthetic appeal in other media or activities. Those with a high order of pattern insight tend to be pre-attuned to find coherent order and resonate with it, wherever and in whatever that coherence may operate, whether or not they have any specific areal training to analyze or reproduce those other, specific, orderly quanta or qualities. This last is not an invariable capacity, for cognitive extremes and/or deficits for a given individual can impede perceptions of aesthetic coherence in diverse information—or enhance such perception in only a few, restricted bodies of information.
Understood as pattern recognition, the seemingly irrational deviations pursued by some who have shown genius insight pertaining to a prior specific range of information are entirely explicable; indeed, these divergent lines of investigation should be anticipated. For pattern recognition is an autonomous cognitive process. An individual will perceive a pattern in information, whether or not there is any actual coherent order there. The mind can impose the perception of pattern as well as perceive its actuality. Pattern identification is inherently prone to false positives. Those who have had the aesthetic delight of profound pattern insight can tend to overestimate their capacity to discern pattern elsewhere. Or can fail to do the attendant reality testing analysis to substantiate a further perception of coherent order, or simply not succeed in analytically substantiating a perceived resolution or orderly relationship. The cognitive capacity to perceive an orderly relationship does not necessarily imply the analytic ability to render that insight firm or analytically useful. The discipline or inclination or simple capacity to substantiate a further perception of coherent order may wander, especially later in life. Josephson may be onto something which could ultimately be substantiated. Hoyle most definitely is not. Einstein’s later theories have in their large part fallen short too. That’s how it goes.
Pattern identification as the driver to theoretical insight firms up and sites the place of mathematics in scientific investigation: confirmation, not inspiration. Math is coherent, and via congruency can flesh out and ‘link in’ coherent organization newly perceived as such. Math can transcribe the order in insight into forms which not only can be utilized but which can be extended beyond their initial core perception(s) of relationship. Great tools, in short—but not the first, aesthetic, vision or relationship. To be cognitively rendered at all, perceptions of different coherence in information in fact often have to be made at some remove from the rigidity and weld-tight relationship of what is already conceived but whose very formal logic and data exclusivity impede a perception of anything else or other. Furthermore, the math is not the pattern, it is a code which draws on the pattern. Worse, mathematical rendering of insight and even observation often truncates the available information, excising putative anomalies or too much complexity, ‘making the math come out’ even if the result is distorted in lesser (or greater) degree from what is perceived as pattern or even received in cumulative but inconvenient data. Thinkers who operate with high order insight will seldom begin from the math even if they often return to it, for the cognitive operation of mathematics scarcely overlaps at all with the cognitive perception of coherent pattern; different and differently comprehensive readings of an object of interest.
The high value of any pre-theoretical fringe so-called in science and order becomes evident when and as theoretical insight is understood as pattern identification. Pre-theoretical bodies of information are rich with substantive observation, fragmentary analysis, and moderate compatibility with what is already known. They are saturated with organization but their dimensions and/or networks of coherent order are not yet perceived as such. They are thus not random bodies of fact and data which are inherently less likely to possess internal avenues of coherence but rather relevant, meaning-dense assemblages of information more than typically likely to have as yet unperceived coherent order(s) striated amongst their features; they are known unknowns, in a phrase. It is much easier to identify coherence in pattern rich information than in or amongst pattern thin or evidently random information. This is why ancient alien theory is a sad pursuit: evidence poor and highly prone to false positive perceptions. This is why organic life on Mars theory is likely to be a more substantive pursuit: data gaining density and the subject constrains false positives of coherent pattern.
Scientific teams make pattern recognition insight nearly impossible. It’s like the convoy system, comprehensible process is chained to the slowest and most partial perception. High standard deviation cognitive pattern identification is, by definition, not an evenly or widely distributed capacity, but such sparks as may occur in a team of a dozen or several dozen researchers fail to propagate at best or are smothered by the sheer density of process in already understood modalities of data framing and analysis. Asymmetric capacities, the buffering effect of limited cognitive perception, and the constraints of inapt or inapplicable prior analytic formalizations make improbable theoretical leaps by bunches of researchers. How many primary theoretical insights have involved even three individuals? Two is not uncommon; one is the norm. 4+ = 0, I think anyone looking at this will find; 6+ most definitely. The herd behavior increasingly prevalent in modern science to me has a pronounced correlation with the steady decline of theoretical insight in any field. This is principally why for instance physics and neuroscience are theoretically stalled at this time, and have been for decades really. Mobs can’t perceive coherent order, and the individuals who can can’t get heard in a crowd or even think clearly there. One can see the same process in poetry, for example, where both aesthetic felicity and originality of vision are presently in total collapse.
Then too, AI does not have native pattern recognition, which is specifically why it is not ‘intelligent.’ The present faulty approach in AI is to use multiple server farm levels of data crunching to generate pattern matching by massive brute force congruency . . . but with zero actual registration of pattern order at all by the inorganic nets. What is achieved is a reading that ‘a something’ is there which might fit ‘another something.’ This is von Neumann’s Sorrow to an exponential magnitude. This is a six-month-old’s degree of pattern identification without even a six-month-old’s quality of pattern recognition, for the human neurological apparatus comes with a degree of hardwired and firmwired pattern recognition capacities whose use can be built out and whose fluency of result can be refined. AI cannot find a partial motorcycle or tell you on its own why a woman is beautiful, only mimic a statement of beauty culled from the web sans comprehension of the coherent patterns represented by the alphanumeric characters or matrix of pixels.
I despise Popper, by the way. His philosophical perspective has zero to do with science, is functionally useless in that field, and no one employs it or ever has. It was a rhetorical sleight of hand to get Marxist dialectics and other ‘fringe theories’ banned from the academy, nothing more, which worked in Britain and America but not elsewhere. It is still taught to students of science in higher education not because it is useful or even intellectually viable, it is not, but to scare them into pulling in their horns and not embarrassing their seniors with startlingly new and disruptive insights which would sweep aside many senior reputations and comfortable livings. “If you can’t prove to me how you’re wrong, we will treat you as if you are—so do as I do and stick to what we know.” The Triumph of the Mediocratic Society . . . .
Perhaps the stupidity of large teams as far as pattern recognition is because they don't get the algorithm right. The individual brains would need to act in service to the team, with some functions of the individual brains' higher levels being dropped and being re-allocated to one of them, or taken up by the team (not sure how that works). What do you mean by "brute force congruency"?
Following Erik's usage of congruency of the post, my reference is to rough fitting of grossly comparable value readings whose actual states are not 'understood' or necessarily relevant to neural net pattern fitting; inexact but with sufficient similarity to make a workable match but lacking any comprehension even of the the states compared, only that they 'sort of fit.'
I do not think that 'algorithm' is at all a good analogy for anything that happens in neural processing. My view. Algorithims are codes, of elements with defined value ranges. I do not think that organic neural processing works in any way by codes, and elements as rendered in organic cognition have prototypal and labile value ranges as I envision this, working value states which evolve as they are used. There is a great deal more to this argument, but I'll leave it there.
Regarding a topological argument, I do think topology applies to neural processing very much, yes, but the causal vector is the other way round than you stated, to me: neural response topologically renders some derived aspect of the topological matrix of nature experienced. It is a 'reading' though by no means an accurate of 1:1 one, and certainly not an 'algorithmic derivation' in a formal sense. Ideas have geometry, literally so, because organic cognition renders readings in real dimensional neural tissue. I suppose if you use the term algorithm loosely, any prototype sequence evolution could be called that, but I do not think that is a primary or desirable usage of that particular semantic concept.
I have my own toy model of a General Theory of Physics with indications of what the physical residuals presently termed fermions and bosons might be. But that is an argument very distant from this comment thread.
So you don't think you can build AIs which are more like brains? Why?
two hypotheses about pattern recognition:
1. When it happens well, the pattern existing in nature is replicated in the brain, but with a different substrate. Relationships are the same or similar, but among different types of parts in the brain vs in nature. It doesn't have to be this way. The algorithm doesn't have to be mimicking, it could be (as it is with AI) a multinomial logistic regression.
2. nature in some specific ways has a similar topology and process to brains. For example, in lattice gauge theory (and in a different way in Wolfram hyper-graph theory), fermions are nodes and bosons are edges connecting nodes of a 3D lattice, each with its own states. Sounds like a neural network. But nature, unlike the brain, is not trying to model or predict anything, as far as we know.
Eric, what do you think about these hypotheses? Will you even see that comment, or will the algorithms not bring it to your attention?
The internet, including substack, is a disaster for our species.
> beauty is truth, and truth beauty
Physicist Sabine Hossenfelder wrote a book called, "Lost in Math: How Beauty Leads Physics Astray". She argues that theoretical physicists have led themselves astray by following their aesthetic preference for beautiful, elegant theories rather than being guided primarily by experimental evidence. I confess that I haven't read the book, but it's on my list. Just curious if you (Erik) have an opinion about Hossenfelder's thesis.