Hung LE @hunglp

Fourier transform of Google Trends of Fourier transform

@skdh Thank you for the great explanation. How does accelerometer distinguish between gravity and force (acceleration)?

why do language models think 9.11 > 9.9? at @TransluceAI we stumbled upon a surprisingly simple explanation - and a bugfix that doesn't use any re-training or prompting. turns out, it's about months, dates, September 11th, and... the Bible?

Transluce @TransluceAI

2 years ago

Monitor: An Observability Interface for Language Models Research report: transluce.org/observability-… Live interface: monitor.transluce.org (optimized for desktop)

4 24 186 328K 159

🆕 𝚜𝚊𝚏𝚎𝚑𝚝𝚝𝚙𝚡: a new open-source library from the Gradio team 🆕 This library is a product of our collaboration with @TrailOfBits and allows you to make asynchronous GET requests while avoiding Server Side Request Forgery. A 🧵 on why this is important!

If you are a technical person (engineer, quant, mathematician) interested in learning biology, I encourage you to hit textbooks and memorize the "things", eg. cytokines, surface proteins, cell types, pathways, in core fields like immunology. You will be averse to this because it wasn't how you learned math and programming. But if you want to actually understand this body of knowledge, converse with scientists and identify real problems in a field that will certainly be a big part of our future, this is the way. Leinster's text on Category Theory opens its section on set theory with a section that stuck with me. "It is not important that the black king is a piece of wood... What matters is what the black king does." In some fields, true fluency with the material and durable knowledge only comes from solving problems and focusing on the structure of interaction between objects. All of the details - memory of notation, properties of objects, supporting lemmas - fall in place from the precise thinking required to work out proofs. Memorizing these things on their own is useless (and unecessary). Biology is not engineering or math. It is actually very important here that the "black king is a piece of wood" because we are generally studying the black king itself. While the goal is to eventually understand the rules that govern the vast networks of biochemistry that compose living systems, modern biology is still concerned with identifying and labeling all the different pieces. When you start to recognize the pieces, you realize that many problems of great value in bioengineering are really quite simple in structure. Can you find some chunk of biomolecular state, eg. transcripts, proteins, even populations of cells, that change with and are ideally causal of disease? The difficulty lies in how slow and expensive it is to answer this question, not because the answer has some complex structure. (This is mostly limited by poor tooling, the assays that measure and synthesize molecules, and this is getting better.) That is why so much learning happens through papers, twitter, conferences, walks with colleagues through Kendall square, clinical trial results, even fundraising announcements. We are learning what the pieces do through the work of others. To do this, it is critical to remember their names (FoxP3, CD8, cytochrome c). So to those that want to work with computers to build new tools in biology, focus and retain the molecular details: - Conversations with scientists become detailed, fluid and fun - Papers become legible and so do deeper problems and questions - Knowledge compounds. Every paper, conversation, clinical milestone updates your understanding of basic biology and the landscape of therapeutic strategies I know very talented engineers who dismiss this as menial and uninteresting work, and to those who feel the same, remember what we are trying to do. Our goal is to learn a body of knowledge in earnest towards having an impact on scientific progress and not compete in some intellectual olympics. Our ambitions for this field (eg. cures, consumer biologics, terraforming) rely on an influx of technical talent from other industries and academic disciplines over the next few decades. It is incredibly important that we encourage more engineers, mathematicians, even heartless hedgefund traders to get excited about and really learn this material.

"Please learn from our mistakes. Don't do exactly the same things that we did, or you'll end up in ten years with having nothing to show for it." — Nicholas Carlini urging AI researchers to avoid the pitfalls of past adversarial ML research at the Vienna Alignment Workshop 2024.

BOOM! 💥 Here it is! 🎉 FastAPI 0.115.0 🚀 One (actually several) of the most awaited features in @FastAPI 🎁 Use @pydantic models for Query parameters ✨ fastapi.tiangolo.com/tutorial/query… Plus more 👇️

Fantastically useful resource from @hunglp: *all* the HTMX examples from @htmx_org documented, running, and collected together as FastHTML apps! 👏 Studying these is a great way to learn FastHTML and HTMX. phihung-htmx-examples.hf.space

We're live! @pydantic Logfire is in open beta. Since I started writing Python in 2010, I've wanted a better way to do logging. Off the back of Pydantic's unbelievable growth, last year I started a company backed by @sequoia, and lucky enough to hire the brightest people I know. Now we've gone and built the logging thing I always wanted. pydantic.dev/logfire

CANDIDATURE PROF À L'UNIVERSITÉ : LA HONTE ADMINISTRATIVE La première nouvelle qui m'écrase et me retire toute la motivation de mon travail, c'est que ma candidature à un poste de PR dont la date limite était mercredi dernier vient d'être jugée irrecevable Rt please

I'm conducting a survey regarding your usage of the terminal. All results will be shared publicly (after being anonymized) on a GitHub Repository. Retweets appreciated! forms.gle/F5Gx1AzAoetbXQ…

J'ai besoin de partager une histoire affolante sur l'état de la recherche en médecine. Vous vous rappelez de l'hydroxychloroquine (HCQ) ? Ce médicament miracle qui allait mettre fin à l'épidémie avant même qu'elle ne commence. 1/

The first part of the Hugging Face Course is finally out! Come learn how the 🤗 Ecosystem works 🥳: Transformers, Tokenizers, Datasets, Accelerate, the Model Hub! Share with your friends who want to learn NLP, it's free! Come join us at hf.co/course

The more papers I read for a review article I'm writing about ML pitfalls in genomics, the more my faith is shaken in the results from papers that apply machine learning to methylation arrays. A salty thread. 1/

Gauss-Lucas Theorem: The convex hull of the zeros of a polynomial with complex coefficients contains the zeros of its derivative. Zeros of P (in red) are bouncing around and the zeros of P', P'', P''', ... nicely dance in the nested convex hulls.

I have been playing around with porting some of my TensorFlow code to PyTorch. Here are my initial thoughts 👇

I just published my very first blogpost ever, about research I did (and really liked) on byte-level models. "Why you should care about byte-level sequence-to-sequence models in NLP" medium.com/@TomKenter/htt… #Medium #NaturalLanguageProcessing #NLU #bytesandcoffee #GoogleResearch

PyText will be open sourced by @fb_engineering later this month, allows rapid prototyping and production deployment of @PyTorch NLP models

keyboard_arrow_left Previous keyboard_arrow_right Next

@omellina newscientist.com/article/218050…

First 1e6 integers, represented as binary vectors indicating their prime factors, and laid out using the sparse matrix support in @leland_mcinnes's UMAP dimensionality reduction algorithm. This is from a 1000000x78628 (!) binary matrix. Very pretty structure emerges.