Dan Butler
@danieljbutler.bsky.social
interests: software, neuroscience, causality, philosophy | ex: salk institute, u of washington, MIT | djbutler.github.io
The Bluesky Python SDK is so cool!
July 14, 2025 at 1:55 AM
The Bluesky Python SDK is so cool!
Cool work out of @arcinstitute.org . My question is, do models like this let us perform novel in-silico experiments the way first-principles models do, or are they just clever way of extrapolating existing experimental data from one context to another?
STATE is designed to predict how various cells respond to perturbations.
The model, available for noncommercial use, is trained on observational data from nearly 170M cells and perturbational data from over 100M cells across 70 cell lines.
Preprint: arcinstitute.org/manuscripts/...
The model, available for noncommercial use, is trained on observational data from nearly 170M cells and perturbational data from over 100M cells across 70 cell lines.
Preprint: arcinstitute.org/manuscripts/...
Manuscript | Arc Institute
Arc Institute is a independent nonprofit research organization headquartered in Palo Alto, California.
arcinstitute.org
June 24, 2025 at 5:15 PM
Cool work out of @arcinstitute.org . My question is, do models like this let us perform novel in-silico experiments the way first-principles models do, or are they just clever way of extrapolating existing experimental data from one context to another?
Reposted by Dan Butler
Cleaning up disk space, I found this image I made for someone not long after the release of the #HHMIJanelia #Drosophila hemibrain #connectome in 2020. It shows EPG neurons in pink providing inputs to PFL1 neurons in transparent grey. I'm not sure if the image was ever used.
June 18, 2025 at 10:46 PM
Cleaning up disk space, I found this image I made for someone not long after the release of the #HHMIJanelia #Drosophila hemibrain #connectome in 2020. It shows EPG neurons in pink providing inputs to PFL1 neurons in transparent grey. I'm not sure if the image was ever used.
Reposted by Dan Butler
Against reductionism: "Our understanding of the world is built up of innumerable layers. Each is worth exploring, as long as we do not forget that it is one of many. Knowing all there is to know about one layer (...) would not teach us much about the rest". Erwin Chargaff
June 8, 2025 at 12:45 PM
Against reductionism: "Our understanding of the world is built up of innumerable layers. Each is worth exploring, as long as we do not forget that it is one of many. Knowing all there is to know about one layer (...) would not teach us much about the rest". Erwin Chargaff
To all the international students, post-docs, scientists, and other academics I’ve been friends with over the years - we support you, and we want you here
May 23, 2025 at 9:13 PM
To all the international students, post-docs, scientists, and other academics I’ve been friends with over the years - we support you, and we want you here
There’s this idea that intelligence involves good predictive models of the world. But most of our lives, we can barely predict anything. We are however extremely good at explaining things in retrospect. I expect human-like AI will be similar
April 1, 2025 at 1:20 PM
There’s this idea that intelligence involves good predictive models of the world. But most of our lives, we can barely predict anything. We are however extremely good at explaining things in retrospect. I expect human-like AI will be similar
Reposted by Dan Butler
The response of IIT researchers to the 2023 paper labeling the field as pseudo-science.
This is the only way science can go ahead: debating.
Still, this is an opinion piece and I think that it's the time for calling the necessity of evidence in support of any theory
www.nature.com/articles/s41...
This is the only way science can go ahead: debating.
Still, this is an opinion piece and I think that it's the time for calling the necessity of evidence in support of any theory
www.nature.com/articles/s41...
Consciousness or pseudo-consciousness? A clash of two paradigms - Nature Neuroscience
Integrated information theory (IIT) starts from consciousness, which is subjective, and accounts for its presence and quality in objective, testable terms. Attempts to label as ‘pseudoscientific’ a th...
www.nature.com
March 12, 2025 at 7:43 AM
The response of IIT researchers to the 2023 paper labeling the field as pseudo-science.
This is the only way science can go ahead: debating.
Still, this is an opinion piece and I think that it's the time for calling the necessity of evidence in support of any theory
www.nature.com/articles/s41...
This is the only way science can go ahead: debating.
Still, this is an opinion piece and I think that it's the time for calling the necessity of evidence in support of any theory
www.nature.com/articles/s41...
Probably the most “nudibranch” out of all of the “nudibranchs” I’ve made.
#SciArt #MarineLife
#ceramic #porcelain #coral #reef #sculpture
#SciArt #MarineLife
#ceramic #porcelain #coral #reef #sculpture
March 11, 2025 at 12:42 PM
What exactly has to change about a data-generating process for the new data to be “out-of-distribution”? Who’s to say that the thing that changed isn’t just an under-sampled variable in the joint distribution?
March 8, 2025 at 9:14 PM
What exactly has to change about a data-generating process for the new data to be “out-of-distribution”? Who’s to say that the thing that changed isn’t just an under-sampled variable in the joint distribution?
Great read - and free!
New book "Explanation in Biology" with Cambridge University Press is out & open access!
Covers (1) causal explanation & (2) non-causal/mathematical explanation in life sciences--bio, neuro, etc 🌿🧬🧠
Introduction to philosophical work on scientific explanation!
www.cambridge.org/core/element...
Covers (1) causal explanation & (2) non-causal/mathematical explanation in life sciences--bio, neuro, etc 🌿🧬🧠
Introduction to philosophical work on scientific explanation!
www.cambridge.org/core/element...
February 13, 2025 at 2:26 PM
Great read - and free!
Reading “The Emergent Multiverse” by David Wallace, and it is *by far* the best explanation of the Many Worlds interpretation of quantum mechanics that I’ve read. Now I think MWI’s name is a total misnomer. (1/2)
January 26, 2025 at 3:05 PM
Reading “The Emergent Multiverse” by David Wallace, and it is *by far* the best explanation of the Many Worlds interpretation of quantum mechanics that I’ve read. Now I think MWI’s name is a total misnomer. (1/2)
It would be interesting to use these as test-cases for an AI scientist. Can an AI come up with explanations for how these artificial lifeforms work?
Introducing ASAL: Automating the Search for Artificial Life with Foundation Models
Blog: sakana.ai/asal/
We propose a new method called Automated Search for Artificial Life (ASAL) which uses foundation models to automate the discovery of the most interesting and open-ended artificial lifeforms!
Blog: sakana.ai/asal/
We propose a new method called Automated Search for Artificial Life (ASAL) which uses foundation models to automate the discovery of the most interesting and open-ended artificial lifeforms!
December 24, 2024 at 4:34 PM
It would be interesting to use these as test-cases for an AI scientist. Can an AI come up with explanations for how these artificial lifeforms work?
Is someone going to make a Chimpanzee Cell Atlas so we can do a diff and figure out what makes us different?
December 20, 2024 at 5:34 PM
Is someone going to make a Chimpanzee Cell Atlas so we can do a diff and figure out what makes us different?
Ooo looks juicy. I wish there were some mention of AI/ML - all these scientific AIs are going to need to deal with levels of explanation
Now out: "Levels of Description and Levels of Reality", as a chapter of this new open-access book (edited by Katie Robertson and Alastair Wilson) on Levels of Explanation, Oxford University Press. philarchive.org/archive/LISLOD
The book itself is here: global.oup.com/academic/pro...
The book itself is here: global.oup.com/academic/pro...
December 17, 2024 at 1:38 PM
Ooo looks juicy. I wish there were some mention of AI/ML - all these scientific AIs are going to need to deal with levels of explanation
That feeling when you're journaling and you write / think your way to a better understanding
December 15, 2024 at 4:56 PM
That feeling when you're journaling and you write / think your way to a better understanding
The learning hypothesis is false. Functional neural circuits are genetically programmed
DAY 8: Advent of Comp Neuro 🎄🤖🧠🧪
Zebrafish grown under anesthesia, with no neural activity, are immediately able to swin and respond to visual stimuli 🤯🐟
"Functional neuronal circuits emerge in the absence of developmental activity"
www.nature.com/articles/s41...
Zebrafish grown under anesthesia, with no neural activity, are immediately able to swin and respond to visual stimuli 🤯🐟
"Functional neuronal circuits emerge in the absence of developmental activity"
www.nature.com/articles/s41...
December 9, 2024 at 3:34 PM
The learning hypothesis is false. Functional neural circuits are genetically programmed
Super interesting. I guess the key question is whether structures like this are found in real models
Fascinating paper from Paul Smolensky et al illustrating how transformers can implement a form of compositional symbol processing, and arguing that an emergent form of this may account for in-context learning in LLMs: arxiv.org/abs/2410.17498
Mechanisms of Symbol Processing for In-Context Learning in Transformer Networks
Large Language Models (LLMs) have demonstrated impressive abilities in symbol processing through in-context learning (ICL). This success flies in the face of decades of predictions that artificial neu...
arxiv.org
December 7, 2024 at 7:17 PM
Super interesting. I guess the key question is whether structures like this are found in real models
Is there a way to hide re-posts in my Bluesky feed? I’m also guilty of re-posting quite a lot, but this is getting ridiculous. I follow people because I want to know what *they* have to say
December 2, 2024 at 9:08 PM
Is there a way to hide re-posts in my Bluesky feed? I’m also guilty of re-posting quite a lot, but this is getting ridiculous. I follow people because I want to know what *they* have to say
Ugh I was wondering why he wasn’t on here. One of my fave accounts on X
hardmaru got banned from bsky because he didn't believe someone who claimed that there's a sentence that, if you write it here, you get insta-banned. So he tried it, to prove it's not true.
Turns out the claim was right.
It's a ban that happens even with all moderation filters disabled.
Turns out the claim was right.
It's a ban that happens even with all moderation filters disabled.
turned off all the @moderation.bsky.app filters and I still can't view bsky.app/profile/hard... 's posts bc he got flagged doing a test. what's the point of algorithmic choice if ultimately a "moderation team" gets to decide what i see @support.bsky.team ?
December 1, 2024 at 2:40 AM
Ugh I was wondering why he wasn’t on here. One of my fave accounts on X
@philipgoff.bsky.social does it bother you that the causal structure of the brain that makes it produce utterances like "I see this inherently subjective redness" would be the same whether the universe were composed of subjective redness or not?
November 27, 2024 at 1:25 PM
@philipgoff.bsky.social does it bother you that the causal structure of the brain that makes it produce utterances like "I see this inherently subjective redness" would be the same whether the universe were composed of subjective redness or not?
I love this post, but one nitpick: science isn't just about compressing the facts either. Compression is a side-effect of the real goal, which is to find explanations. Proof: imagine a highly compressed representation of a system, versus an explanation of that system - which does the scientist want?
I wish that I was able to convince the students in my class that the ultimate goal of science isn't accumulation of facts but the compression of facts into theories. Mel summarizes it well here:
www.nature.com/articles/nn1...
Unfortunately, I think this is a minority viewpoint within neuroscience.
www.nature.com/articles/nn1...
Unfortunately, I think this is a minority viewpoint within neuroscience.
www.nature.com
November 26, 2024 at 12:46 AM
I love this post, but one nitpick: science isn't just about compressing the facts either. Compression is a side-effect of the real goal, which is to find explanations. Proof: imagine a highly compressed representation of a system, versus an explanation of that system - which does the scientist want?
What do learning theorists think of ARC-AGI by @fchollet.bsky.social ? Are there theory papers that are relevant?
November 25, 2024 at 9:26 PM
What do learning theorists think of ARC-AGI by @fchollet.bsky.social ? Are there theory papers that are relevant?
Reposted by Dan Butler
How do LLMs learn to reason from data? Are they ~retrieving the answers from parametric knowledge🦜? In our new preprint, we look at the pretraining data and find evidence against this:
Procedural knowledge in pretraining drives LLM reasoning ⚙️🔢
🧵⬇️
Procedural knowledge in pretraining drives LLM reasoning ⚙️🔢
🧵⬇️
November 20, 2024 at 4:35 PM
How do LLMs learn to reason from data? Are they ~retrieving the answers from parametric knowledge🦜? In our new preprint, we look at the pretraining data and find evidence against this:
Procedural knowledge in pretraining drives LLM reasoning ⚙️🔢
🧵⬇️
Procedural knowledge in pretraining drives LLM reasoning ⚙️🔢
🧵⬇️
Reposted by Dan Butler
Sometimes you just have to persevere through negative reviews. Reviewers, please be kind :) #methodsmatter bit.ly/gs2-scholar
November 25, 2024 at 3:05 AM
Sometimes you just have to persevere through negative reviews. Reviewers, please be kind :) #methodsmatter bit.ly/gs2-scholar
Reposted by Dan Butler
The Foldscope brings a powerful science tool to schools that can't afford microscopes. Scientists use it too. Its creators have handed out 2 million units, including a new mini-model for younger kids.
Fold paper. Insert lens. This $2 microscope changes how kids see the world
The Foldscope brings a powerful science tool to schools that can't afford microscopes. Scientists use it too. Its creators have handed out 2 million units, including a new mini-model for younger kids.
www.npr.org
November 24, 2024 at 5:29 PM
The Foldscope brings a powerful science tool to schools that can't afford microscopes. Scientists use it too. Its creators have handed out 2 million units, including a new mini-model for younger kids.