Blake Richards
@tyrellturing.bsky.social
Researcher at Google and CIFAR Fellow, working on the intersection of machine learning and neuroscience in Montréal (academic affiliations: @mcgill.ca and @mila-quebec.bsky.social).
Reposted by Blake Richards
1/7 How should feedback signals influence a network during learning? Should they first adjust synaptic weights, which then indirectly change neural activity (as in backprop.)? Or should they first adjust neural activity to guide synaptic updates (e.g., target prop.)? openreview.net/forum?id=xVI...
January 8, 2026 at 10:10 PM
1/7 How should feedback signals influence a network during learning? Should they first adjust synaptic weights, which then indirectly change neural activity (as in backprop.)? Or should they first adjust neural activity to guide synaptic updates (e.g., target prop.)? openreview.net/forum?id=xVI...
Reposted by Blake Richards
Really interesting work, introducing Epiplexity (an observer-dependent measure of information under computational constraints)
Feels like something theoretical neuroscience could use right away.
🧠📈
Feels like something theoretical neuroscience could use right away.
🧠📈
We introduce epiplexity, a new measure of information that provides a foundation for how to select, generate, or transform data for learning systems. We have been working on this for almost 2 years, and I cannot contain my excitement! arxiv.org/abs/2601.03220 1/7
January 8, 2026 at 1:44 AM
Really interesting work, introducing Epiplexity (an observer-dependent measure of information under computational constraints)
Feels like something theoretical neuroscience could use right away.
🧠📈
Feels like something theoretical neuroscience could use right away.
🧠📈
This looks super interesting!
I wonder if it maybe solves the tension that led to controversy after my post on the brain not being able to create information about the world. (Because I was talking about Shannon info, so didn't distinguish random versus structured info like this.)
I wonder if it maybe solves the tension that led to controversy after my post on the brain not being able to create information about the world. (Because I was talking about Shannon info, so didn't distinguish random versus structured info like this.)
We introduce epiplexity, a new measure of information that provides a foundation for how to select, generate, or transform data for learning systems. We have been working on this for almost 2 years, and I cannot contain my excitement! arxiv.org/abs/2601.03220 1/7
January 8, 2026 at 6:16 PM
This looks super interesting!
I wonder if it maybe solves the tension that led to controversy after my post on the brain not being able to create information about the world. (Because I was talking about Shannon info, so didn't distinguish random versus structured info like this.)
I wonder if it maybe solves the tension that led to controversy after my post on the brain not being able to create information about the world. (Because I was talking about Shannon info, so didn't distinguish random versus structured info like this.)
Reposted by Blake Richards
We introduce epiplexity, a new measure of information that provides a foundation for how to select, generate, or transform data for learning systems. We have been working on this for almost 2 years, and I cannot contain my excitement! arxiv.org/abs/2601.03220 1/7
January 7, 2026 at 5:28 PM
We introduce epiplexity, a new measure of information that provides a foundation for how to select, generate, or transform data for learning systems. We have been working on this for almost 2 years, and I cannot contain my excitement! arxiv.org/abs/2601.03220 1/7
Reposted by Blake Richards
Reminder that this job is open! If you have any questions, feel free to get in touch. Deadline is February 20th.
Great news! We are looking for an NHP neuroscientist as the assistant professor level. We have no preconceived ideas -- looking for the most exciting research going. If you have any questions, please reach out. universityaffairs.ca/search-jobs/...
Search Jobs - University Affairs
universityaffairs.ca
January 6, 2026 at 8:36 PM
Reminder that this job is open! If you have any questions, feel free to get in touch. Deadline is February 20th.
Reposted by Blake Richards
We recently published a theoretical review about how compositional and generative mechanisms in working memory provide a flexible engine for creative perception and imagery.
Pre-print:
osf.io/preprints/ps...
Paper: www.sciencedirect.com/science/arti...
Pre-print:
osf.io/preprints/ps...
Paper: www.sciencedirect.com/science/arti...
January 6, 2026 at 7:04 PM
We recently published a theoretical review about how compositional and generative mechanisms in working memory provide a flexible engine for creative perception and imagery.
Pre-print:
osf.io/preprints/ps...
Paper: www.sciencedirect.com/science/arti...
Pre-print:
osf.io/preprints/ps...
Paper: www.sciencedirect.com/science/arti...
Reposted by Blake Richards
What is the computational role of dendritic excitations? Byung Hun Lee and team mapped voltage dynamics throughout the dendritic trees of CA1 pyramidal neurons in mice navigating in virtual reality. www.biorxiv.org/content/10.6...
January 4, 2026 at 6:17 AM
What is the computational role of dendritic excitations? Byung Hun Lee and team mapped voltage dynamics throughout the dendritic trees of CA1 pyramidal neurons in mice navigating in virtual reality. www.biorxiv.org/content/10.6...
Reposted by Blake Richards
With a new year comes a new Editor-in-Chief! Please give a warm welcome to Laurent Charlin (HEC Montréal and Mila)!
He rounds out the team with Gautam Kamath, Naila Murray, and Nihar Shah to help lead TMLR through 2026.
He rounds out the team with Gautam Kamath, Naila Murray, and Nihar Shah to help lead TMLR through 2026.
January 5, 2026 at 4:39 PM
With a new year comes a new Editor-in-Chief! Please give a warm welcome to Laurent Charlin (HEC Montréal and Mila)!
He rounds out the team with Gautam Kamath, Naila Murray, and Nihar Shah to help lead TMLR through 2026.
He rounds out the team with Gautam Kamath, Naila Murray, and Nihar Shah to help lead TMLR through 2026.
Reposted by Blake Richards
This paper had a pretty shocking headline result (40% of voxels!), so I dug into it, and I think it is wrong. Essentially: they compare two noisy measures and find that about 40% of voxels have different sign between the two. I think this is just noise!
Would love to hear expert views on this paper. It appears to show that the operationalization of brain activity the field has relied on for 3 decades—the BOLD response—is not actually a sensible measure of brain activity.
www.nature.com/articles/s41...
www.nature.com/articles/s41...
BOLD signal changes can oppose oxygen metabolism across the human cortex - Nature Neuroscience
Using quantitative brain imaging, the authors show opposite fMRI BOLD signal to metabolic activity due to variable oxygen extraction across the human cortex. This questions the canonical interpretatio...
www.nature.com
January 5, 2026 at 5:22 PM
This paper had a pretty shocking headline result (40% of voxels!), so I dug into it, and I think it is wrong. Essentially: they compare two noisy measures and find that about 40% of voxels have different sign between the two. I think this is just noise!
Reposted by Blake Richards
New preprint. We show that in addition to reward prediction errors (RPEs), dorsal striatal dopamine signals encode sensory prediction errors (SPEs), the difference between sensory prior & observed stimulus. www.biorxiv.org/content/10.6...
Dorsal striatal dopamine integrates sensory and reward prediction errors to guide perceptual decisions
Perceptual decisions are shaped by expectations about sensory stimuli and rewards, learned through sensory and reward prediction errors. Dopamine is known to convey reward prediction errors that shape...
www.biorxiv.org
January 5, 2026 at 10:49 AM
New preprint. We show that in addition to reward prediction errors (RPEs), dorsal striatal dopamine signals encode sensory prediction errors (SPEs), the difference between sensory prior & observed stimulus. www.biorxiv.org/content/10.6...
Reposted by Blake Richards
Need something to feel good about? I sure do. Here are 7 fantastic science stores from 2025. Science will continue, even if The Regime tries to kill it at home. #Science 🧪
Seven feel-good science stories to restore your faith in 2025
Immense progress in gene editing, drug discovery and conservation are just some of the reasons to be cheerful about 2025.
www.nature.com
December 19, 2025 at 2:15 AM
Need something to feel good about? I sure do. Here are 7 fantastic science stores from 2025. Science will continue, even if The Regime tries to kill it at home. #Science 🧪
Reposted by Blake Richards
McGill is recruiting top-tier researchers working abroad through the federally funded Canada Impact+ Research Chairs program, addressing global and national challenges. The first round is due in early 2026.
Learn more and submit your candidacy: https://mcgill.ca/x/5Zh
Learn more and submit your candidacy: https://mcgill.ca/x/5Zh
December 18, 2025 at 9:02 PM
McGill is recruiting top-tier researchers working abroad through the federally funded Canada Impact+ Research Chairs program, addressing global and national challenges. The first round is due in early 2026.
Learn more and submit your candidacy: https://mcgill.ca/x/5Zh
Learn more and submit your candidacy: https://mcgill.ca/x/5Zh
Great to see this paper on sensory expectations in motor control from @jonathanamichaels.bsky.social and @andpru.bsky.social out in Nature today!
www.nature.com/articles/s41...
#neuroscience 🧪
www.nature.com/articles/s41...
#neuroscience 🧪
Sensory expectations shape neural population dynamics in motor circuits - Nature
Experiments with human volunteers and macaques show that expectations produced by probabilistic cueing of future sensory inputs shape motor circuit dynamics in order to increase the efficiency of move...
www.nature.com
December 18, 2025 at 8:26 PM
Great to see this paper on sensory expectations in motor control from @jonathanamichaels.bsky.social and @andpru.bsky.social out in Nature today!
www.nature.com/articles/s41...
#neuroscience 🧪
www.nature.com/articles/s41...
#neuroscience 🧪
I don't know what it is with my luck, but every time I say to myself "Maybe all these Mac enthusiasts are onto something, I should try it", I instantly regret it.
Can I connect my bluetooth headphones? No.
Is there a physical jack I can use anyway? Also no.
I literally find Linux easier to debug.
Can I connect my bluetooth headphones? No.
Is there a physical jack I can use anyway? Also no.
I literally find Linux easier to debug.
December 17, 2025 at 9:07 PM
I don't know what it is with my luck, but every time I say to myself "Maybe all these Mac enthusiasts are onto something, I should try it", I instantly regret it.
Can I connect my bluetooth headphones? No.
Is there a physical jack I can use anyway? Also no.
I literally find Linux easier to debug.
Can I connect my bluetooth headphones? No.
Is there a physical jack I can use anyway? Also no.
I literally find Linux easier to debug.
Reposted by Blake Richards
This is part of the reason why I teach & make YouTube videos on how AI is used in science: I don't want all the baggage from genAI & chatbots to prevent the public from supporting or trusting better & more grounded forms of AI.
Finally @mmitchell.bsky.social explains why generative AI hype can blind us to the AI breakthroughs that we should really celebrate. www.technologyreview.com/2025/12/15/1...
Generative AI hype distracts us from AI’s more important breakthroughs
It's a seductive distraction from the advances in AI that are most likely to improve or even save your life
www.technologyreview.com
December 17, 2025 at 2:13 PM
This is part of the reason why I teach & make YouTube videos on how AI is used in science: I don't want all the baggage from genAI & chatbots to prevent the public from supporting or trusting better & more grounded forms of AI.
Reposted by Blake Richards
Why isn’t modern AI built around principles from cognitive science or neuroscience? Starting a substack (infinitefaculty.substack.com/p/why-isnt-m...) by writing down my thoughts on that question: as part of a first series of posts giving my current thoughts on the relation between these fields. 1/3
Why isn’t modern AI built around principles from cognitive science?
First post in a series on cognitive science and AI
infinitefaculty.substack.com
December 16, 2025 at 3:40 PM
Why isn’t modern AI built around principles from cognitive science or neuroscience? Starting a substack (infinitefaculty.substack.com/p/why-isnt-m...) by writing down my thoughts on that question: as part of a first series of posts giving my current thoughts on the relation between these fields. 1/3
Reposted by Blake Richards
How do biological agents learn for the future?
Our perspective piece on the value of prospective learning in neuroscience is finally out. This is part of a long running collaboration with @kordinglab.bsky.social & Josh Vogelstein (as well as many other people)
www.sciencedirect.com/science/arti...
Our perspective piece on the value of prospective learning in neuroscience is finally out. This is part of a long running collaboration with @kordinglab.bsky.social & Josh Vogelstein (as well as many other people)
www.sciencedirect.com/science/arti...
December 17, 2025 at 5:06 PM
How do biological agents learn for the future?
Our perspective piece on the value of prospective learning in neuroscience is finally out. This is part of a long running collaboration with @kordinglab.bsky.social & Josh Vogelstein (as well as many other people)
www.sciencedirect.com/science/arti...
Our perspective piece on the value of prospective learning in neuroscience is finally out. This is part of a long running collaboration with @kordinglab.bsky.social & Josh Vogelstein (as well as many other people)
www.sciencedirect.com/science/arti...
Reposted by Blake Richards
My team and I just published a new framework for AI social intelligence. By grasping self-other similarity and predicting how their actions affect others, models are encouraged to cooperate—solving the Prisoner’s Dilemma.
More on Embedded Universal Predictive Intelligence (MUPI) ⬇
More on Embedded Universal Predictive Intelligence (MUPI) ⬇
1/ Why does RL struggle with social dilemmas? How can we ensure that AI learns to cooperate rather than compete?
Introducing our new framework: MUPI (Embedded Universal Predictive Intelligence) which provides a theoretical basis for new cooperative solutions in RL.
Preprint🧵👇
(Paper link below.)
Introducing our new framework: MUPI (Embedded Universal Predictive Intelligence) which provides a theoretical basis for new cooperative solutions in RL.
Preprint🧵👇
(Paper link below.)
December 16, 2025 at 5:36 PM
My team and I just published a new framework for AI social intelligence. By grasping self-other similarity and predicting how their actions affect others, models are encouraged to cooperate—solving the Prisoner’s Dilemma.
More on Embedded Universal Predictive Intelligence (MUPI) ⬇
More on Embedded Universal Predictive Intelligence (MUPI) ⬇
Reposted by Blake Richards
🧪 Over half of researchers now use AI for peer review, often violating confidentiality protocols. As third-party tools risk data security, publishers are moving toward closed internal systems to safely integrate AI assistance.
#AcademicSky #PeerReview
#AcademicSky #PeerReview
More than half of researchers now use AI for peer review — often against guidance
A survey of 1,600 academics found that more than 50% have used artificial-intelligence tools while peer reviewing manuscripts.
www.nature.com
December 16, 2025 at 3:29 PM
🧪 Over half of researchers now use AI for peer review, often violating confidentiality protocols. As third-party tools risk data security, publishers are moving toward closed internal systems to safely integrate AI assistance.
#AcademicSky #PeerReview
#AcademicSky #PeerReview
#AITestForTasks
If everyone's using AI to do X, then X is probably something we don't actually need to be wasting our mental energy doing... (at least not as often as we do).
If everyone's using AI to do X, then X is probably something we don't actually need to be wasting our mental energy doing... (at least not as often as we do).
🧪 Over half of researchers now use AI for peer review, often violating confidentiality protocols. As third-party tools risk data security, publishers are moving toward closed internal systems to safely integrate AI assistance.
#AcademicSky #PeerReview
#AcademicSky #PeerReview
More than half of researchers now use AI for peer review — often against guidance
A survey of 1,600 academics found that more than 50% have used artificial-intelligence tools while peer reviewing manuscripts.
www.nature.com
December 16, 2025 at 3:56 PM
#AITestForTasks
If everyone's using AI to do X, then X is probably something we don't actually need to be wasting our mental energy doing... (at least not as often as we do).
If everyone's using AI to do X, then X is probably something we don't actually need to be wasting our mental energy doing... (at least not as often as we do).
Reposted by Blake Richards
Though I don't use it myself, using AI to turn bullet point notes into a written summary I think is not technically against confidentially rules (as it doesn't require feeding in the manuscript itself). Other uses listed here certainly are though, and that is disappointing to see.
🧪 Over half of researchers now use AI for peer review, often violating confidentiality protocols. As third-party tools risk data security, publishers are moving toward closed internal systems to safely integrate AI assistance.
#AcademicSky #PeerReview
#AcademicSky #PeerReview
More than half of researchers now use AI for peer review — often against guidance
A survey of 1,600 academics found that more than 50% have used artificial-intelligence tools while peer reviewing manuscripts.
www.nature.com
December 16, 2025 at 3:37 PM
Though I don't use it myself, using AI to turn bullet point notes into a written summary I think is not technically against confidentially rules (as it doesn't require feeding in the manuscript itself). Other uses listed here certainly are though, and that is disappointing to see.
Reposted by Blake Richards
1/X Excited to present this preprint on multi-tasking, with
@david-g-clark.bsky.social and Ashok Litwin-Kumar! Timely too, as “low-D manifold” has been trending again. (If you read thru the end, we escape Flatland and return to the glorious high-D world we deserve.) www.biorxiv.org/content/10.6...
@david-g-clark.bsky.social and Ashok Litwin-Kumar! Timely too, as “low-D manifold” has been trending again. (If you read thru the end, we escape Flatland and return to the glorious high-D world we deserve.) www.biorxiv.org/content/10.6...
A theory of multi-task computation and task selection
Neural activity during the performance of a stereotyped behavioral task is often described as low-dimensional, occupying only a limited region in the space of all firing-rate patterns. This region has...
www.biorxiv.org
December 15, 2025 at 7:41 PM
1/X Excited to present this preprint on multi-tasking, with
@david-g-clark.bsky.social and Ashok Litwin-Kumar! Timely too, as “low-D manifold” has been trending again. (If you read thru the end, we escape Flatland and return to the glorious high-D world we deserve.) www.biorxiv.org/content/10.6...
@david-g-clark.bsky.social and Ashok Litwin-Kumar! Timely too, as “low-D manifold” has been trending again. (If you read thru the end, we escape Flatland and return to the glorious high-D world we deserve.) www.biorxiv.org/content/10.6...
Reposted by Blake Richards
métro, station préfontaine, montréal
December 14, 2025 at 10:46 PM
métro, station préfontaine, montréal
Reposted by Blake Richards
Incredible piece on Oliver Sacks. If you were ever awed at his supposedly true stories (I remember being stunned by the account of the autistic twins who rattled off large prime numbers), read this. He told wonderful stories, but they were in large part fiction.
www.newyorker.com/magazine/202...
www.newyorker.com/magazine/202...
Oliver Sacks Put Himself Into His Case Studies. What Was the Cost?
The scientist was famous for linking healing with storytelling. Sometimes that meant reshaping patients’ reality.
www.newyorker.com
December 12, 2025 at 10:33 PM
Incredible piece on Oliver Sacks. If you were ever awed at his supposedly true stories (I remember being stunned by the account of the autistic twins who rattled off large prime numbers), read this. He told wonderful stories, but they were in large part fiction.
www.newyorker.com/magazine/202...
www.newyorker.com/magazine/202...