Andrei Zinovyev

Temporal Exploration of COPD Phenotypes: Insights from the COPDGene and SPIROMICS Cohorts The authors present a method that phenotypes and models disease trajectories in a large population of COPD patients @meilan_han @umichmedicine @andrei_zinovyev tinyurl.com/yvru6rp2

LLM model size competition is intensifying… backwards! My bet is that we'll see models that "think" very well and reliably that are very very small. There is most likely a setting even of GPT-2 parameters for which most people will consider GPT-2 "smart". The reason current models are so large is because we're still being very wasteful during training - we're asking them to memorize the internet and, remarkably, they do and can e.g. recite SHA hashes of common numbers, or recall really esoteric facts. (Actually LLMs are really good at memorization, qualitatively a lot better than humans, sometimes needing just a single update to remember a lot of detail for a long time). But imagine if you were going to be tested, closed book, on reciting arbitrary passages of the internet given the first few words. This is the standard (pre)training objective for models today. The reason doing better is hard is because demonstrations of thinking are "entangled" with knowledge, in the training data. Therefore, the models have to first get larger before they can get smaller, because we need their (automated) help to refactor and mold the training data into ideal, synthetic formats. It's a staircase of improvement - of one model helping to generate the training data for next, until we're left with "perfect training set". When you train GPT-2 on it, it will be a really strong / smart model by today's standards. Maybe the MMLU will be a bit lower because it won't remember all of its chemistry perfectly. Maybe it needs to look something up once in a while to make sure.

Are you interested in joining @sysbiocurie in Paris, and applying machine learning methodologies for fighting cancer ? We are looking for a PhD student to develop new concepts and tools for analyzing multi-modal and spatial omics data of cancer patients. sysbio.curie.fr/texts/PhdCompu…

Check out our new paper by the extremely talented postdoc @James__Cornwell out now in @Nature. For decades it was thought the cell cycle was irreversible, but now we show that is not the case. nature.com/articles/s4158…

📢 Our paper is out! Explore cancer cell invasion mechanisms in our latest publication on #Bioinformatics.🧬 Using #PhysiBoSS, we simulated cancer cells in an extracellular matrix, capturing gene interactions and physical dynamics! academic.oup.com/bioinformatics…

@YannPonty Fortunately for me, Pascal was my third or the fourth programming language, but I did my PhD on it (and still use some code from that time)

We will be in Basel in September, with @ArnauMontagud & #Laurence Calzone to organise a pretty cool workshop at [BC]2 @BC2Conference ! Think about joining us!

New paper! It’s absolutely wild what you see when studying bacteria using single-cell RNA-Seq! Unlike in eukaryotes, we found a genome-scale pattern of correlations among genes showing a strong dependency on global chromosomal locations. biorxiv.org/content/10.110… @AndrewPountain

I also wrote a brief summary of our new AlphaTensor paper in my blog: furidamu.org/blog/2022/10/0…

Great post-doc job opportunity with @DiazMunoz_Lab and @MargheritaGhisi @crctoncopole @InfinityTLS @BSC_CNS @BU_CNIO @sysbioire @SysBioCurie @esantapau @MichaelTress Apply before 15th October! sfbi.fr/emplois/offre/… please retweet

A great poster session yesterday at #ECCB2022! There was also @njmmatthieu presenting his poster "Construction of the cystic fibrosis biological network from a comparative analysis of transcriptomic studies"! Looking forward for today's session! 💪🏻

Incredible welcome to #sitges last night for #eccb2022 coinciding with amazing #draco festival!

@Annaerial See you #ECCB2022 ?

Beautiful Barcelona <3...I can't believe it's been almost a decade since my last visit...

Multiscale modeling allows to study the different modes of cancer cell invasion 🗣️Marco Ruscone, Arnau Montagud, Philippe Chavrier, Olivier Destaing, Andrei Zinovyev, Emmanuel Barillot, Vincent Noël and Laurence Calzone

Lab leaders wrestle with paucity of postdocs #Echobox=1661901810" target="_blank" rel="nofollow noopener">nature.com/articles/d4158…

Sad news. One of the last top journal which was free to publish in. Golden open access is not the only path to open access. Looking forward to the end of this crazy system.

The Institute is very pleased to welcome @VillaniCedric as the holder of the IHES-University of Lyon Chair in Analysis. The mathematician will share his time between IHES and @UnivLyon1. ihes.fr/en/cedric-vill…