Sophie Wang
34 followers · 13 following · 3101 views
on the atlas — 167
- College Is Coming Apart - The Atlantic3 savers
- Deep learning theory lecture notes1 savers
- probability-of-causation-interpretation-and-identification.pdf1 savers
- Causal Statements and Statements of Necessary and/or Sufficient Conditions | Causal Inquiry in International Relations | Oxford Academic1 savers
- Fernleaf Hedge Bamboo Under The Moonlight (Background Music) Key C | 月光下的凤尾竹C调伴奏 - YouTube1 savers
- 卡拉乐团网1 savers
- Alice Munro1 savers
- shepard-science-87.pdf1 savers
- tversky-features.pdf1 savers
- [2203.14465] STaR: Bootstrapping Reasoning With Reasoning2 savers
- Prompt Injection as Role Confusion4 savers
- View of Mechanisms of Symbol Processing for In-Context Learning in Transformer Networks1 savers
- Iconic Arithmetic | Iconic Math1 savers
- The Making of Cursor's Icons – Minor Adventures6 savers
- Compression is prediction | ngrok blog2 savers
- Agnes Obel: 'It's called a Trautonium – and it can electrocute people!' | Music | The Guardian1 savers
- Introduction to Program Synthesis3 savers
- Color Associations™1 savers
- At More Than 800 Pages, 'The Book of Colour Concepts' Revels in Four Centuries of Chromatic Wonders — Colossal1 savers
- [2605.03907] Steer Like the LLM: Activation Steering that Mimics Prompting1 savers
- Compression and Intelligence — Ryan Greene5 savers
- [2303.14151] Double Descent Demystified: Identifying, Interpreting & Ablating the Sources of a Deep Learning Puzzle1 savers
- IEEE Xplore Full-Text PDF:36 savers
- Corollary discharge theory1 savers
- Hans-Lukas Teuber1 savers
- Cortical homunculus1 savers
- Spurious Correlations2 savers
- LoRA Without Regret - Thinking Machines Lab37 savers
- Core's Music1 savers
- Chronotype1 savers
- Advice compiled by Michael Ernst3 savers
- Chihiro Iwasaki1 savers
- Mimeograph1 savers
- Mojibake1 savers
- Diary of Kitou Aya - One Litre of Tears | 1リットルの涙1 savers
- Why China got rich and India didn't - David Oks2 savers
- Foundations of Computer Vision5 savers
- Bao To - Design Engineer & Product Maker8 savers
- [2502.13967] FlexTok: Resampling Images into 1D Token Sequences of Flexible Length1 savers
- [1606.09282] Learning without Forgetting2 savers
- [2605.15220] Always Learning, Always Mixing: Efficient and Simple Data Mixing All The Time1 savers
- [1606.09282] Learning without Forgetting1 savers
- Palmer Lab1 savers
- The Dispossessed3 savers
- Elegiac1 savers
- All around us: Aya Kito and Her Diary ( english )2 savers
- 1_Litre_no_Namida1 savers
- Cross-Cultural Patterns of Attachment: A Meta-Analysis of the Strange Situation on JSTOR1 savers
- vatsal0.github.io/blog/emergence.html1 savers
- The Sirens and Ulysses1 savers
- Antinomy1 savers
- Heraclitus1 savers
- Sense_data2 savers
- Hume's_fork1 savers
- Discourse on the Method1 savers
- The Structure of Evolutionary Theory1 savers
- The spandrels of San Marco and the Panglossian paradigm: a critique of the adaptationist programme1 savers
- The Spandrels of San Marco and the Panglossian Paradigm1 savers
- The Dawn of Everything2 savers
- The_Dialectical_Biologist1 savers
- Knowledge Machine1 savers
- What the Tortoise Said to Achilles2 savers
- DCDRealPatterns1991.pdf1 savers
- Friedrich August Hayek-New studies in philosophy, politics, economics and the history of ideas-Routledge (1990)1 savers
- Naturalized_epistemology2 savers
- Two Dogmas of Empiricism3 savers
- Blue–green distinction in language2 savers
- New riddle of induction3 savers
- meodai (David Aerne)1 savers
- Dashboard - Open OnDemand1 savers
- efros-thesis.pdf1 savers
- The Lottery in Babylon1 savers
- Technology-Driven Moral Panics – Interactive Timeline1 savers
- Point-set registration1 savers
- Orthogonal Procrustes problem2 savers
- The Old Guitarist1 savers
- Smarthistory2 savers
- gallery1 savers
- Aileen's Portfolio17 savers
- How the Beat Generation Became "Beatniks" - JSTOR Daily1 savers
- Langer's lines1 savers
- Phoebe Yu1 savers
- [2410.15468] What Emergence Can Possibly Mean1 savers
- Phenomenology and Natural Science | Internet Encyclopedia of Philosophy1 savers
- [2605.15220] Always Learning, Always Mixing: Efficient and Simple Data Mixing All The Time2 savers
- Nudibranch1 savers
- Greg Yang | Professional page2 savers
- Maternal Reminiscing Style and Children’s Developing Understanding of Self and Emotion | Clinical Social Work Journal | Springer Nature Link2 savers
- Childhood amnesia1 savers
- Lenia2 savers
- What Is It Like to Be a Bat?4 savers
- MACH-IV: Machiavellianism Test1 savers
- Machiavellian intelligence hypothesis1 savers
- Plant_evolutionary_developmental_biology1 savers
- Interpolation_theory1 savers
- CSS Basic User Interface Module Level 3 (CSS3 UI)1 savers
- [1703.04908] Emergence of Grounded Compositional Language in Multi-Agent Populations1 savers
- Nearly Everyone, Everywhere, Veers Left When Walking - The New York Times3 savers
- Outsider_art2 savers
- The Lies and Falsifications of Oliver Sacks1 savers
highlights — 136
He viewed the world as constantly in flux, always "becoming" but never "being".
HeraclitusElements of The Old Guitarist were carefully chosen to generate a reaction from the spectator. For example, the monochromatic color scheme creates flat, two-dimensional forms that dissociate the guitarist from time and place. In addition, the overall muted blue palette creates a general tone of melancholy and accentuates the tragic and sorrowful theme. The sole use of oil on panel causes a darker and more theatrical mood. Oil tends to blend the colors together without diminishing brightness, creating an even more cohesive dramatic composition.[citation needed] Furthermore, the guitarist, altho…
The Old GuitaristMother and child focus in on the loss of a relationship as an occasion for sadness, but the mother assures her child that even though her friend has moved far away, they can maintain the relationship, and provides one way to keep the relationship alive. In fact, it is obvious from this excerpt that this is something that the mother and child have discussed before and both are clearly sharing the emotional experience together.
Maternal Reminiscing Style and Children’s Developing Understanding of Self and Emotion | Clinical Social Work Journal | Springer Nature LinkAt first, mother and child disagree about the child’s emotional reactions, but the mother does not dismiss her child’s rendition; she pays careful attention and draws the child out, ultimately helping the child to understand and integrate her multiple feelings about the event. In this way, highly elaborative mothers are helping their children not just to understand what happened, but to understand their feelings about what happened and to help evaluate the event from the child’s personal perspective. Thus highly elaborative mothers are helping their children to build connections between past e…
Maternal Reminiscing Style and Children’s Developing Understanding of Self and Emotion | Clinical Social Work Journal | Springer Nature LinkThe general consensus is that because eastern cultures are more collectivistic than western cultures, they engage in less self-focused reminiscing. Indeed, by middle childhood, children in eastern cultures tell less elaborated and detailed stories of their personal past than do western children (Han, Leichtman, & Wang, 1998), and this difference persists through adulthood (Pillemer, 1998). These finding suggest that cultural values of having and telling personal narratives influences the cultural practice of reminiscing.
Maternal Reminiscing Style and Children’s Developing Understanding of Self and Emotion | Clinical Social Work Journal | Springer Nature LinkResearchers who have used maternal narrative style as a predictor for early childhood memory recall have found that children whose mothers frequently and elaborately reminisced events with the child were able to recall earlier first memories.[50] Studies also show that when mothers include emotional context when reminiscing with their children, the child is likely to grow up with a more secure attachment style, as well as show higher levels of emotional understanding and a stronger sense of self.[51] These factors all contribute to the child developing stronger autobiographical memory skills.[…
Childhood amnesiaSome suggest that as children age, they lose the ability to recall preverbal memories.[32] One explanation for this is that a person develops linguistic skills, memories that were not encoded verbally are lost.[32] This theory also explains why many early memories are fragmented: the nonverbal components were lost.[33] Contrary findings indicate that elementary-aged children remember more accurate details about events than they had reported at a younger age[34] and that 6- to 9-year-old children tend to have verbally accessible memories from very early childhood.[34] Research on animal models …
Childhood amnesiamathematically superior
JPEGBut somehow when it comes to us humans, it seems as if the greatest early leap in abstraction was the invention of language, and the explicit delineation of concepts that could be quite far from our direct experience.
What If We Had Bigger Brains? Imagining Minds beyond Ours—Stephen Wolfram WritingsSo what could bigger brains do with all this? Potentially they could handle more features, and more concepts. Full computational irreducibility will always in effect ultimately overpower them. But when it comes to handling pockets of reducibility, they’ll presumably be able to deal with more of them. So in the end, it’s very much as one might expect: a bigger brain should be able to track more things going on, “see more details”, etc.
What If We Had Bigger Brains? Imagining Minds beyond Ours—Stephen Wolfram WritingsCuttlefish are notable for dynamically producing elaborate patterns on their skin—giving them in a sense the hardware to “communicate in pictures”. But so far as one can tell, they produce just a small number of distinct patterns—and certainly nothing like a “pictorial generalization of compositional language”.
What If We Had Bigger Brains? Imagining Minds beyond Ours—Stephen Wolfram WritingsOf course, in some small ways we do have the ability to “directly communicate with images”, for example in our use of gestures and body language. Right now, these seem like largely ancillary forms of communication. But, yes, it’s conceivable that with bigger brains, they could be more.
What If We Had Bigger Brains? Imagining Minds beyond Ours—Stephen Wolfram WritingsIt’s not too hard to get something like concepts to emerge in experiments on training neural nets to pass data through a “bottleneck” that simulates a “mind-to-mind communication channel”. But how compositionality or grammatical structure might emerge is not clear.
What If We Had Bigger Brains? Imagining Minds beyond Ours—Stephen Wolfram WritingsBut no doubt the “neural environment” inside a brain is continually changing (not least because of its stream of sensory input). And so the only way to successfully “preserve a thought” across time is presumably to “package it up” in terms of robust elements, or essentially in terms of language.
What If We Had Bigger Brains? Imagining Minds beyond Ours—Stephen Wolfram WritingsBut what is “appropriate generalization”? As a practical matter, it tends to be “generalization that aligns with what we humans would do”. And it’s then a remarkable fact that artificial neural nets with fairly simple architectures can successfully do generalizations in a way that’s roughly aligned with human brains. So why does this work? Presumably it’s because there are universal features of “brain-like systems” that are close enough between human brains and neural nets. And once again it’s important to emphasize that what’s happening in both cases seems distinctly weaker than “general comp…
What If We Had Bigger Brains? Imagining Minds beyond Ours—Stephen Wolfram WritingsSo what about (artificial) neural nets? It’s routine to “look inside” these, and for example see the possible patterns of activation at a given layer based on a range of possible (“real-world”) inputs. We can then think of these patterns of activation as forming points in a “feature space”. And typically we’ll be able to see clusters of these points, which we can potentially identify as “emergent concepts” that we can view as having been “discovered” by the neural net (or rather, its training). Normally there won’t be existing words in human languages that correspond to most of these concepts.…
What If We Had Bigger Brains? Imagining Minds beyond Ours—Stephen Wolfram WritingsThe charms that work on others count for nothing in that devastatingly well-lit back alley where one keeps assignations with oneself.
making yourself proud - starting from nixAnne Lamott once said writing a novel was like driving a car at night. You can see only as far as your headlights, but you can make the whole trip that way.
choice loneliness - starting from nixAnne Lamott once said writing a novel was like driving a car at night. You can see only as far as your headlights, but you can make the whole trip that way.
choice loneliness - starting from nixWhen should you read the literature in a new field? Suppose you want to move into a new research area. How much of the that area's literature should you read, and when? On the one hand, you can't ignore the literature in a new field completely, or you'll end up reinventing wheels and annoying the cognoscenti (especially if you are one of the cognoscenti in a neighboring research area). On the other hand, though, reading the literature early on can be even worse: you end up implicitly aligning your thoughts to the types of questions and answers that are currently popular, and you stifle the typ…
Brian Scholl: Misc Academic MusingsOur work, however, considers the loss over many training examples rather than just one. Counterintuitively, this flips the above relationship on its head. A direction being flat implies that perturbing those weights changes the loss for very few examples, as is the case for memorized data. Meanwhile, directions in weight space that correspond to shared, general computational structures - the mechanisms for attention, composition, reasoning, etc. - exhibit higher curvature because they're used across many examples, and so perturbing those weights changes the overall loss much more.
Understanding Memorization via Loss CurvatureIf a model has memorized a specific training example exactly, perturbing the relevant weights even slightly breaks that memorization and spikes the loss. The loss landscape for generalizing solutions, by contrast, tends to be flatter - it's robust to small perturbations because generalizing solutions capture broader patterns.
Understanding Memorization via Loss CurvaturePrevious work has established that the loss on individual memorized examples has high curvature - i.e., is sharp and brittle (Garg et al. 2023, Ravikumar et al. 2024).
Understanding Memorization via Loss CurvatureBy analyzing the curvature of a model's loss landscape, we can disentangle the directions in weight space which primarily support memorization from those that support general computation. This lets us selectively edit models to reduce memorization while preserving reasoning capabilities.
Understanding Memorization via Loss CurvatureAn abrupt cognitive shift in how the mind understands information is known as a representational change. Although researchers have inferred sudden shifts in understanding from the behavior of subjects, they have not pinned down how the brain supports representational change.
How Your Brain Creates ‘Aha’ Moments and Why They Stick | Quanta MagazineBut if we fetishize LLMs as machine gods, then motivated reasoning demands we see them as machine gods, even if this means completely discounting our own intelligence. If you want to believe that AGI is here, it helps to become willfully forgetful of what you already know.
Lore Laundering Machines - by Ben Recht - arg minA quirky thing about the decision theory problem is that everyone stumbles upon the same optimal decision rule, regardless of how they model the future.
Justify your answer - by Ben Recht - arg minRegarding the second point, simple correlation does not necessarily entail a fine-item-level structural correspondence between two similarity structures. For example, previous studies using representational similarity analysis (RSA)9, a common method for assessing perceptual representational similarity via simple correlations between similarity matrices, have shown that high correlation values (e.g., 𝜌 = 0.9 ) may indicate only coarse category-level correspondence, even while fine-item-level alignment is completely absent (e.g., Fig. 3 in14). Thus, the mere presence of a high correlation does …
Gromov–Wasserstein unsupervised alignment reveals structural correspondences between the color similarity structures of humans and large language models | Scientific ReportsThey suspected that functional information was the key to understanding how complex systems like living organisms arise through evolutionary processes happening over time. “We all assumed the second law of thermodynamics supplies the arrow of time,” Hazen said. “But it seems like there’s a much more idiosyncratic pathway that the universe takes. We think it’s because of selection for function — a very orderly process that leads to ordered states. That’s not part of the second law, although it’s not inconsistent with it either.”
Why Everything in the Universe Turns More Complex | Quanta MagazineIn biology, sometimes many different molecules can do the same job. Consider RNA molecules, some of which have biochemical functions that can easily be defined and measured. (Like DNA, RNA is made up of sequences of nucleotides.) In particular, short strands of RNA called aptamers securely bind to other molecules. Let’s say you want to find an RNA aptamer that binds to a particular target molecule. Can lots of aptamers do it, or just one? If only a single aptamer can do the job, then it’s unique, just as a long, seemingly random sequence of letters is unique. Szostak said that this aptamer wou…
Why Everything in the Universe Turns More Complex | Quanta MagazineBut Szostak pointed out that Kolmogorov’s measure of complexity neglects an issue crucial to biology: how biological molecules function.
Why Everything in the Universe Turns More Complex | Quanta MagazineInstead, evolution is a special (and perhaps inevitable) case of a more general principle that governs the universe.
Why Everything in the Universe Turns More Complex | Quanta MagazineA new proposal by an interdisciplinary team of researchers challenges that bleak conclusion. They have proposed nothing less than a new law of nature, according to which the complexity of entities in the universe increases over time with an inexorability comparable to the second law of thermodynamics — the law that dictates an inevitable rise in entropy, a measure of disorder. If they’re right, complex and intelligent life should be widespread.
Why Everything in the Universe Turns More Complex | Quanta MagazineTo address these limitations, we propose using an unsupervised alignment approach to assess a more detailed level of structural correspondence between the similarity structures of humans and LLMs. In unsupervised alignment, the correspondence between items in two similarity structures is not assumed.
Gromov–Wasserstein unsupervised alignment reveals structural correspondences between the color similarity structures of humans and large language models | Scientific ReportsThe supervised approach (or supervised alignment method in general11) assumes that an element in one similarity structure (for example, the color ’red’) corresponds to the same element in another similarity structure.
Gromov–Wasserstein unsupervised alignment reveals structural correspondences between the color similarity structures of humans and large language models | Scientific ReportsThese observations raise two intriguing questions: To what extent can Large Language Models accurately infer human perceptual representations, and how can we effectively compare LLMs and human perceptual representations?
Gromov–Wasserstein unsupervised alignment reveals structural correspondences between the color similarity structures of humans and large language models | Scientific ReportsThese results contribute to the methodological advancements of comparing LLMs with human perception, and highlight the potential of unsupervised alignment methods to reveal detailed structural correspondences.
Gromov–Wasserstein unsupervised alignment reveals structural correspondences between the color similarity structures of humans and large language models | Scientific ReportsThis form of Pinsker's inequality shows that "convergence in divergence" is a stronger notion than "convergence in variation distance".
Pinsker's inequality - WikipediaOne way to show that grounding is not necessary for learning meaning is to show that conceptual representations learned by an (ungrounded) LLM are isomorphic to grounded representations of those same concepts. That is, we want evidence that there exists a mapping between the spaces that preserves all the individual concepts and the relations between them. If such a mapping exists, then any computation carried out in the grounded space could in principle be carried out in the ungrounded space and yield the same result, in which case the spaces might as well be the same. (Of course, further work…
Symbols and grounding in large language models | Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences“Ideas are getting harder to find” is a pretty bleak thing to believe. It says, “Look around the world. This is pretty much as good as it gets; the returns start diminishing from here. All the problems you see are unlikely to be solved anytime soon, so you better get used to them.” Why would anyone agree to such a thing without putting up a fight?
Ideas aren’t getting harder to find and anyone who tells you otherwise is a coward and I will fight themProfessionalized science, then, may force us to keep building elevators even though we can’t get them to go any higher. Weirdos with crazy schemes for hot air balloons simply don’t get jobs; after all, they don’t even have a degree in elevators!
Ideas aren’t getting harder to find and anyone who tells you otherwise is a coward and I will fight themThat’s why I’m puzzled by the claim that scientists must labor under an ever-increasing burden of knowledge. The author of that paper writes: “If one is to stand on the shoulders of giants, one must first climb up their backs, and the greater the body of knowledge, the harder this climb becomes.” This suggests that if you peek into PhD programs, you’ll see lots of students bent over their books, desperately trying to learn everything that’s come before so they can start their own projects. “I can’t do any physics yet,” you might hear them lament. "I’m only up to Huygens!” Instead, you’ll see P…
Ideas aren’t getting harder to find and anyone who tells you otherwise is a coward and I will fight themYou might worry that this means discovery gets harder over time because you have to follow a chain to its end before you can add a new link. Fortunately:
Ideas aren’t getting harder to find and anyone who tells you otherwise is a coward and I will fight themThe player of the game experiences the same lifetime over and over, losing to the reset all of the physical results of their actions, but gaining more and more information. In our world, it seems that we experience only one lifetime, and that the world loses our knowledge upon death. But in that lifetime we receive from all of the other lifetimes in the world all of the information that they can share, and all of the physical results of their actions.
the time loop and the lifetime (idea) by raincomplex - Everything2.comA quest is something deeply personal. You never know where it will lead, or whether that will be a reassuring or a threatening place. No matter, however. After all, you undertake a quest because the rewards of the process itself are so great.
To See a World in a Grain of Sand – Nature’s DepthsTo see a world in a grain of sand—to peer so deeply into the nature of any one thing that the riches of the Universe begin to be revealed—that to me is the essence of science as a quest. Not as a profession or a career, not as a niche in complex modern society, but as a quest for understanding one’s deepest nature.
To See a World in a Grain of Sand – Nature’s Depthsn this work, we introduce an order preserving regularizer which aims to preserve the ordering structure of the reconstruction coefficients within the sparse coding framework.
Order Preserving Sparse Coding | IEEE Journals & Magazine | IEEE Xploreorder-preserving sparse coding for classifying structured data whose atomic features possess ordering relationships
Order Preserving Sparse Coding | IEEE Journals & Magazine | IEEE XploreWrite so your audience can understand why your work is of interest to them, providing them with a clear take-home message that they can grasp in the few minutes they will spend at your poster.
Preparing and Presenting Effective Research Posters - PMCSparse coding is a problem of representing a collection of 𝑁 𝑁 vectors in 𝑑 𝑑 -dimensional space as linear combinations of some basis vectors, with the requirement that the weights (in linear combinations) should be sparse. Mathematically, if 𝐗 𝑋 is the data matrix of 𝑁×𝑑 𝑁 × 𝑑 size, then you want to find weights 𝐖 𝑊 and basis vectors 𝐓 𝑇 , such that 𝐗≈𝐖𝐓
machine learning - Sparse coding vs. sparse PCA, are they the same thing? - Cross Validated