Kevin Wang
26 followers · 22 following · 2519 views
on the atlas — 36
- Six (and a half) intuitions for KL divergence - LessWrong15 savers
- Effective learning: Twenty rules of formulating knowledge - SuperMemo6 savers
- Cognitive load is what matters25 savers
- minjunes.ai/posts/sleep/index.html4 savers
- The Zero-Day Flaw in AI Companies7 savers
- Ilya 30u3012 savers
- What Every Programmer Should Know About Memory7 savers
- 2404.10102v11 savers
- Patterns for Building LLM-based Systems & Products9 savers
- python - How does asyncio actually work? - Stack Overflow1 savers
- Accelerating Generative AI with PyTorch II: GPT, Fast | PyTorch1 savers
- LLM inference optimization1 savers
- linux - "Unable to find remote helper for 'https'" during git clone - Stack Overflow1 savers
- Multi-Query & Grouped-Query Attention1 savers
- CAUSALITY - Discussion1 savers
- Book list for streetfighting computer scientists - dankwiki, the wiki of nick black3 savers
- How to Optimize a CUDA Matmul Kernel for cuBLAS-like Performance: a Worklog24 savers
- svm_dual_kernel1 savers
- Mamba No. 5 (A Little Bit Of…) | Sparse Notes3 savers
- Troubles with the Bias-Variance tradeoff | Erik Jenner1 savers
- L1 regularization: sparsity through singularities | Erik Jenner1 savers
- cuda-mode/resource-stream: CUDA related news and material links1 savers
- An Updated Academic Workflow: Zotero & Obsidian | by Alexandra Phelan | Medium2 savers
- Building a fast Electron app with Rust2 savers
- tensor.pdf1 savers
- infinite.pdf1 savers
- What is Category Theory Anyway?3 savers
- What is a Monad?1 savers
- Monoids without tears | F# for fun and profit1 savers
- PyTorch internals : Inside 245-5D6 savers
- Childhoods of exceptional people - by Henrik Karlsson28 savers
- Simulators - LessWrong16 savers
- HSRG13 savers
- Everything you think you know about homelessness is wrong11 savers
- Cameron's World7 savers
- Robert Cunningham's Blog4 savers
highlights — 12
What makes the idea so compelling, I think, is that it both flatters PoC writers—our work isn’t bad, and it isn’t failing to meet its mark, it’s simply “in a different tradition” from white writers. At the same time, it emancipates white workshop participants from needing to do any further work. They are allowed to dislike stories by PoC, because they’re not meant to like those stories.
The New, Weirdly Racist Guide to Writing Fiction - Tablet MagazineAre the ways qualia relate to computation, creativity to free will, risk to probability, morality to epistemology, all the same question?
The Last Questions (Part 2) - by Roger’s Bacon - SecretorumIt is just a way of viewing children: as capable of competence, as craving meaningful work, as worthy to be included in serious discussions. We can learn to view them like that, but it is a subtle and profound shift in perception, a shift away from the way we are taught to view children.
Childhoods of exceptional people - by Henrik KarlssonJames would model patterns of reasoning by thinking aloud and ask John Stuart to recreate his thought, imitating the thought patterns. He would give him increasingly complex tasks (books or ideas that he wanted John Stuart to summarize and articulate), then he would scaffold John Stuart by asking questions that helped him solve the task, and he would coach and give feedback on how to improve.
Childhoods of exceptional people - by Henrik Karlssonthe average tutored student was above 98% of the students in the control class.
Childhoods of exceptional people - by Henrik KarlssonA qualified guess is that they spent between one and four hours daily in formal studies, and the rest on self-directed projects. Unlike children today, they had little access to entertainment, and so were often bored, unless they figured out a way to keep their minds occupied; the intellectual obsessions that grew into their life’s work often grew out of this boredom.
Childhoods of exceptional people - by Henrik Karlssonthis immersion in boredom is also a universal in the biographies of exceptional people. A substantial fraction were completely kept apart from other children, either because their guardians decided so or because they were bedridden with various illnesses during childhood (like Descartes). A spicy hypothesis raised by this is that socializing too much with children is simply not good for your intellectual development.
Childhoods of exceptional people - by Henrik KarlssonNot everyone who grew to be exceptional was this lucky. There are a few cases of people who rose to greatness despite their non-ideal circumstances—like Ramanujan and Michael Faraday. But they, too, were the fruit of exceptional milieus. They just had to summon it themselves. How did they do that? First, they did this by reading books, by self-teaching. Second, when they grew more skilled they started reaching out to exceptional people, trying to convince them to bring them into their milieu.
Childhoods of exceptional people - by Henrik KarlssonAt least two-thirds of my sample was home-educated (most commonly until about age 12), tutored by parents or governesses and tutors. The rest of my sample had been educated in schools (most commonly Jesuit schools). As children, they were integrated with exceptional adults—and were taken seriously by them.
Childhoods of exceptional people - by Henrik KarlssonIt is therefore not surprising that those who grow up to be exceptional tend to have spent their formative years surrounded by adults who were exceptional.
Childhoods of exceptional people - by Henrik KarlssonIf we’re ambitious about getting the truth from a finitely powerful GPT, we need to incite it to predict truth-seeking processes, not just ask it the right questions. Or, in other words, the more general problem we have to solve is not asking GPT the question[20] that makes it output the right answer, but asking GPT the question that makes it output the right question (…) that makes it output the right answer.
Simulators - LessWrongI called GPT-3 a “disappointing paper,” which is not the same thing as calling the model disappointing: the feeling is more like how I’d feel if they found a superintelligent alien and chose only to communicate its abilities by noting that, when the alien is blackout drunk and playing 8 simultaneous games of chess while also taking an IQ test, it then has an “IQ” of about 100. Treating GPT as an unsupervised implementation of a supervised learner leads to systematic underestimation of capabilities, which becomes a more dangerous mistake as unprobed capabilities scale.
Simulators - LessWrong