@jeremyjkun The scikit-learn docs cite a paper about applications of the JL lemma for database embeddings: https://t.co/AfUVjuxZgA
Not sure if this has been implemented in a production database though.
The impressive deep pattern recognition abilities of #DNN's such as #LLM's are sometimes confused for reasoning abilities
I can learn to guess, with high accuracy, whether a SAT instance is satisfiable or not, but this not the same as knowing how to solve SAT. Let me explain. 1/
I am delighted to announce that the camera-ready version of my new book, "Machine Learning: Advanced Topics", is finally available online for free at https://t.co/XjR3nnLtOI (@mitpress will publish the hard copy in 2023.)
One scientific fact that I find mind-blowing is that the experiments which give the most precise measurement of the mass of the sun don't look at the sun or, indeed, involve the sun in any way whatsoever. Instead, they might, e.g., place heavy masses around a torsion balance.
About the raging debate regarding the significance of recent progress in AI, it may be useful to (re)state a few obvious facts:
(0) there is no such thing as AGI. Reaching "Human Level AI" may be a useful goal, but even humans are specialized.
1/N
This may seem like innocuous behavior, but it is a hallmark of genuine intelligence that’s eluded connectionist AI for decades - dynamic symbol binding and manipulation.
Gato🐈a scalable generalist agent that uses a single transformer with exactly the same weights to play Atari, follow text instructions, caption images, chat with people, control a real robot arm, and more: https://t.co/9Q7WsRBmIC
Paper: https://t.co/ecHZqzCSAm 1/
Reading ML bio literature, I noticed Matthews correlation coefficient (vs other metrics for assessing ML classifiers) is quite popular. Usually, I stick to precision, recall, and F1 for interpretability reasons, but this article is really convincing: https://t.co/JcKdC4ejRd 🧵
New blog post in #aSpoonfulOfOcean 🥄💧 ! Big populations helped us predict the ecology of a system, let's now see how long timescales can be used to predict its evolution 👇
https://t.co/PbKsQkAs2b
What is the public value of research mathematics?
Here's a brief thread about my essay for the new @protectingmaths blog, which you can find here:
https://t.co/wHK0BDPStT
(1/6)
We’ve acquired the MuJoCo physics simulator (https://t.co/knwXLZMr4L) and are making it free for all, to support research everywhere. MuJoCo is a fast, powerful, easy-to-use, and soon to be open-source simulation tool, designed for robotics research: https://t.co/Of3Q1W2GIR
"Pitfalls in ML Research: Reexamining the Dev Cycle" -- really great article that I can only highly recommend to ML researchers & practitioners: https://t.co/YMcCFh8ldV. (PS: Not to be confused with the also excellent "How to avoid ML pitfalls" I shared a few weeks ago) 👇
Today, we go into the nuts and bolts of individual centered models in #aSpoonfulOfOcean 🥄💧! As always, everything is in plain language, so come on here if you want to have a feel on how exactly one can model darwinian evolution with very little effort 👇
https://t.co/Auk62XgkCr
My collaborators and I recently posted a paper to the arXiv that uses simple ideas from enriched category theory to address a math question motivated by the success of large language models. I'll explain some of the ideas on the blog - 1st post is now up! https://t.co/9LVeOEbayK
Back to biology in #aSpoonfulOfOcean🥄💧! Today, we learn about the way ocean biology can help trap atmospheric CO2 for several thousand years 👉 https://t.co/gtbHHmQ0bt
And it went great!
Teven now has two papers in conference proceedings: the 🤗transformers *best demo paper* at EMNLP 2020, and this *best paper* at NAACL 2021 on the equivalence between prompts and data points 😱