Miles Cranmer
@milescranmer.bsky.social
AI + Astro/Physics. Assistant Prof at Cambridge.
astroautomata.com/
astroautomata.com/
Does anybody know the actual page limit for ICLR submissions? The author guidelines page contradicts itself:
September 24, 2025 at 10:44 PM
Does anybody know the actual page limit for ICLR submissions? The author guidelines page contradicts itself:
Cambridge Maths calls the top student in a given year the "Senior Wrangler."
Why?
Apparently this originates in Medieval examinations called "disputations," where students would literally wrangle (argue) in Latin with designated opponents.
Why?
Apparently this originates in Medieval examinations called "disputations," where students would literally wrangle (argue) in Latin with designated opponents.
September 15, 2025 at 12:36 PM
Cambridge Maths calls the top student in a given year the "Senior Wrangler."
Why?
Apparently this originates in Medieval examinations called "disputations," where students would literally wrangle (argue) in Latin with designated opponents.
Why?
Apparently this originates in Medieval examinations called "disputations," where students would literally wrangle (argue) in Latin with designated opponents.
PolymathicAI is recruiting two postdoctoral researchers to join our team at Cambridge, to work on building and understanding large-scale foundation models for science. Please share with potential candidates!
September 4, 2025 at 9:46 AM
PolymathicAI is recruiting two postdoctoral researchers to join our team at Cambridge, to work on building and understanding large-scale foundation models for science. Please share with potential candidates!
That makes sense!
I've had a few people ask for this and am considering it. For now, though, you could implement this constraint manually with the following custom loss function: gist.github.com/MilesCranmer.... Pass it as a string to `loss_function_expression`.
(Let me know if this doesn't work!)
I've had a few people ask for this and am considering it. For now, though, you could implement this constraint manually with the following custom loss function: gist.github.com/MilesCranmer.... Pass it as a string to `loss_function_expression`.
(Let me know if this doesn't work!)
July 15, 2025 at 11:32 AM
That makes sense!
I've had a few people ask for this and am considering it. For now, though, you could implement this constraint manually with the following custom loss function: gist.github.com/MilesCranmer.... Pass it as a string to `loss_function_expression`.
(Let me know if this doesn't work!)
I've had a few people ask for this and am considering it. For now, though, you could implement this constraint manually with the following custom loss function: gist.github.com/MilesCranmer.... Pass it as a string to `loss_function_expression`.
(Let me know if this doesn't work!)
Really loving AR glasses for working while traveling. No squinting at my laptop + I can still enjoy the view!
These just plug into my closed laptop as an external monitor. Means I'm actually productive on planes now because I'm not hurting my neck looking down the whole flight!
These just plug into my closed laptop as an external monitor. Means I'm actually productive on planes now because I'm not hurting my neck looking down the whole flight!
July 15, 2025 at 10:35 AM
Really loving AR glasses for working while traveling. No squinting at my laptop + I can still enjoy the view!
These just plug into my closed laptop as an external monitor. Means I'm actually productive on planes now because I'm not hurting my neck looking down the whole flight!
These just plug into my closed laptop as an external monitor. Means I'm actually productive on planes now because I'm not hurting my neck looking down the whole flight!
Tomorrow I get to give my favourite type of lecture to some new physics PhD students 😃
July 14, 2025 at 5:58 PM
Tomorrow I get to give my favourite type of lecture to some new physics PhD students 😃
I am a HUGE fan of Mooncake.jl github.com/chalk-lab/Mooncake.jl, a new language-level autodiff for Julia.
Most impressive is how uncompromising the devs are about correctness and reliability. It has some of the best software engineering practices I've seen in the entire ecosystem!
Most impressive is how uncompromising the devs are about correctness and reliability. It has some of the best software engineering practices I've seen in the entire ecosystem!
July 8, 2025 at 1:54 PM
I am a HUGE fan of Mooncake.jl github.com/chalk-lab/Mooncake.jl, a new language-level autodiff for Julia.
Most impressive is how uncompromising the devs are about correctness and reliability. It has some of the best software engineering practices I've seen in the entire ecosystem!
Most impressive is how uncompromising the devs are about correctness and reliability. It has some of the best software engineering practices I've seen in the entire ecosystem!
SymbolicRegression.jl v1.10.0 is out! It can now evolve expressions over arbitrary input types.
The video below shows it reverse-engineering a string transformation from examples.
Curious to see how people use this!
The video below shows it reverse-engineering a string transformation from examples.
Curious to see how people use this!
May 2, 2025 at 10:43 PM
SymbolicRegression.jl v1.10.0 is out! It can now evolve expressions over arbitrary input types.
The video below shows it reverse-engineering a string transformation from examples.
Curious to see how people use this!
The video below shows it reverse-engineering a string transformation from examples.
Curious to see how people use this!
You have no idea how long I’ve waited for this 😮💨
April 8, 2025 at 4:28 PM
You have no idea how long I’ve waited for this 😮💨
We have just opened a fully-funded PhD position at Cambridge, supervised by me & Vasily Belokurov.
Topic: AI + astronomical imaging (broadly defined).
Deadline: April 16.
Please share with anyone who may be interested!
www.postgraduate.study.cam.ac.uk/courses/dire...
Topic: AI + astronomical imaging (broadly defined).
Deadline: April 16.
Please share with anyone who may be interested!
www.postgraduate.study.cam.ac.uk/courses/dire...
April 2, 2025 at 11:42 AM
We have just opened a fully-funded PhD position at Cambridge, supervised by me & Vasily Belokurov.
Topic: AI + astronomical imaging (broadly defined).
Deadline: April 16.
Please share with anyone who may be interested!
www.postgraduate.study.cam.ac.uk/courses/dire...
Topic: AI + astronomical imaging (broadly defined).
Deadline: April 16.
Please share with anyone who may be interested!
www.postgraduate.study.cam.ac.uk/courses/dire...
Very excited to share that my group in Cambridge is offering an “AI for Science” Encode fellowship on multiscale physics through Pillar VC & ARIA!
Apply by April 30th:
encode.pillar.vc/projects/mul...
Apply by April 30th:
encode.pillar.vc/projects/mul...
March 25, 2025 at 9:08 PM
Very excited to share that my group in Cambridge is offering an “AI for Science” Encode fellowship on multiscale physics through Pillar VC & ARIA!
Apply by April 30th:
encode.pillar.vc/projects/mul...
Apply by April 30th:
encode.pillar.vc/projects/mul...
This might be the greatest bug report I’ve ever seen
March 6, 2025 at 6:44 PM
This might be the greatest bug report I’ve ever seen
Wow, Google Colab just added Julia support!!! 🎉
March 5, 2025 at 10:27 PM
Wow, Google Colab just added Julia support!!! 🎉
Why 'I don’t know' is the true test for AGI—it’s a strictly harder problem than text generation!
This magnificent 62-page paper (arxiv.org/abs/2408.02357) formally proves AGI hallucinations are inevitable, with 50 pages (!!) of supplementary proofs.
This magnificent 62-page paper (arxiv.org/abs/2408.02357) formally proves AGI hallucinations are inevitable, with 50 pages (!!) of supplementary proofs.
February 26, 2025 at 8:48 PM
Why 'I don’t know' is the true test for AGI—it’s a strictly harder problem than text generation!
This magnificent 62-page paper (arxiv.org/abs/2408.02357) formally proves AGI hallucinations are inevitable, with 50 pages (!!) of supplementary proofs.
This magnificent 62-page paper (arxiv.org/abs/2408.02357) formally proves AGI hallucinations are inevitable, with 50 pages (!!) of supplementary proofs.
Symbolic regression update—new versions of PySR & SymbolicRegression.jl just released with improvements to mini-batching. Should work much better for large datasets!
February 25, 2025 at 8:02 AM
Symbolic regression update—new versions of PySR & SymbolicRegression.jl just released with improvements to mini-batching. Should work much better for large datasets!
Took me until today to figure out how to separate tracked/untracked files in VSCode. So much better!
February 15, 2025 at 1:11 PM
Took me until today to figure out how to separate tracked/untracked files in VSCode. So much better!
New feature! PySR v1.4 lets you define a template expression to optimize that both has learnable parameters AND learnable expressions:
February 14, 2025 at 3:22 PM
New feature! PySR v1.4 lets you define a template expression to optimize that both has learnable parameters AND learnable expressions:
I’ll be giving lectures on advanced ML techniques at this AI-for-HEP summer school in Lake Como, Italy, from July 14-18. Expect a deep dive into symbolic AI and beyond. Should be a great event!
Students can apply here: https://buff.ly/3WTTE8J
Students can apply here: https://buff.ly/3WTTE8J
February 11, 2025 at 11:08 AM
I’ll be giving lectures on advanced ML techniques at this AI-for-HEP summer school in Lake Como, Italy, from July 14-18. Expect a deep dive into symbolic AI and beyond. Should be a great event!
Students can apply here: https://buff.ly/3WTTE8J
Students can apply here: https://buff.ly/3WTTE8J
Interested in feedback on this syntax idea for predefined "template expressions" in SymbolicRegression.jl and PySR.
Basically: what is an intuitive way to prescribe a fixed skeleton for a symbolic search? Ideally the syntax should be flexible and extensible too.
Basically: what is an intuitive way to prescribe a fixed skeleton for a symbolic search? Ideally the syntax should be flexible and extensible too.
February 4, 2025 at 9:43 AM
Interested in feedback on this syntax idea for predefined "template expressions" in SymbolicRegression.jl and PySR.
Basically: what is an intuitive way to prescribe a fixed skeleton for a symbolic search? Ideally the syntax should be flexible and extensible too.
Basically: what is an intuitive way to prescribe a fixed skeleton for a symbolic search? Ideally the syntax should be flexible and extensible too.
Exciting news – Cambridge is hiring an Assistant/Associate Professor in Data-Driven Radio Astro. Join us!
January 30, 2025 at 2:25 PM
Exciting news – Cambridge is hiring an Assistant/Associate Professor in Data-Driven Radio Astro. Join us!
Exploring a "Template Expression" API design that allows joint optimization of symbolic expressions with parameter vectors. Can be used for optimizing a neural network simultaneously with a symbolic model, finding parametric expressions, etc.
Would love feedback! https://buff.ly/4fINO0x
Would love feedback! https://buff.ly/4fINO0x
January 3, 2025 at 12:00 PM
Exploring a "Template Expression" API design that allows joint optimization of symbolic expressions with parameter vectors. Can be used for optimizing a neural network simultaneously with a symbolic model, finding parametric expressions, etc.
Would love feedback! https://buff.ly/4fINO0x
Would love feedback! https://buff.ly/4fINO0x
🚨 FINAL REMINDER 🚨: Multiple Postdoc and PhD positions in our AI + Physics cluster in DAMTP, Cambridge!
Deadlines:
- Postdoc: Jan 5th (Sunday)
- PhD: Jan 7th (Tuesday)
More info below – Please share with researchers and students who might be interested in joining us!
Deadlines:
- Postdoc: Jan 5th (Sunday)
- PhD: Jan 7th (Tuesday)
More info below – Please share with researchers and students who might be interested in joining us!
January 2, 2025 at 6:32 PM
🚨 FINAL REMINDER 🚨: Multiple Postdoc and PhD positions in our AI + Physics cluster in DAMTP, Cambridge!
Deadlines:
- Postdoc: Jan 5th (Sunday)
- PhD: Jan 7th (Tuesday)
More info below – Please share with researchers and students who might be interested in joining us!
Deadlines:
- Postdoc: Jan 5th (Sunday)
- PhD: Jan 7th (Tuesday)
More info below – Please share with researchers and students who might be interested in joining us!
Many people don't realize this about floating point arithmetic:
This is also why GPUs can give slightly different results between runs! Thread execution order changes the sequence of operations.
(CUDA also has a slower deterministic mode, which works by artificially enforcing operation order)
This is also why GPUs can give slightly different results between runs! Thread execution order changes the sequence of operations.
(CUDA also has a slower deterministic mode, which works by artificially enforcing operation order)
December 30, 2024 at 1:43 PM
Many people don't realize this about floating point arithmetic:
This is also why GPUs can give slightly different results between runs! Thread execution order changes the sequence of operations.
(CUDA also has a slower deterministic mode, which works by artificially enforcing operation order)
This is also why GPUs can give slightly different results between runs! Thread execution order changes the sequence of operations.
(CUDA also has a slower deterministic mode, which works by artificially enforcing operation order)
o1 pro is so weirdly confident in its hallucinations that it spent three replies gaslighting me about a basic CUDA array slice, even made up this 'experiment' to prove it was right. I really paid $240 for artisanal hallucinations 💀
December 17, 2024 at 2:10 PM
o1 pro is so weirdly confident in its hallucinations that it spent three replies gaslighting me about a basic CUDA array slice, even made up this 'experiment' to prove it was right. I really paid $240 for artisanal hallucinations 💀
Happy to announce differential operators for PySR + SymbolicRegression.jl!
This means you can literally just... evolve an integral:
This means you can literally just... evolve an integral:
December 16, 2024 at 2:06 PM
Happy to announce differential operators for PySR + SymbolicRegression.jl!
This means you can literally just... evolve an integral:
This means you can literally just... evolve an integral: