Daniel Whiteson
@danielwhiteson.bsky.social
Particle physics and ML (http://sites.uci.edu/daniel). Co-creator @DandKUniverse podcast and @ElinorWonders TV. Patreon: https://patreon.com/DanielWhiteson.
Our new book is out today! @andycomics.bsky.social
November 4, 2025 at 3:29 PM
Our new book is out today! @andycomics.bsky.social
Would aliens describe the Universe the same way we do? Many physicists think "obviously, yes!", but that confidence is built on lots of untested assumptions.
My book, "Do Aliens Speak Physics?" unpacks those assumptions and makes a lot of dad jokes.
Here's what Sean Carroll thought of it:
My book, "Do Aliens Speak Physics?" unpacks those assumptions and makes a lot of dad jokes.
Here's what Sean Carroll thought of it:
October 29, 2025 at 12:50 AM
Would aliens describe the Universe the same way we do? Many physicists think "obviously, yes!", but that confidence is built on lots of untested assumptions.
My book, "Do Aliens Speak Physics?" unpacks those assumptions and makes a lot of dad jokes.
Here's what Sean Carroll thought of it:
My book, "Do Aliens Speak Physics?" unpacks those assumptions and makes a lot of dad jokes.
Here's what Sean Carroll thought of it:
Can we make mental contact with aliens using our most universal science, physics? What if our physics is more human than we expect, alien science more alien?
This is the topic of my book, "Do Aliens Speak Physics?" out next week!
Here's what @seamus.bsky.social said about it:
This is the topic of my book, "Do Aliens Speak Physics?" out next week!
Here's what @seamus.bsky.social said about it:
October 27, 2025 at 5:53 PM
Can we make mental contact with aliens using our most universal science, physics? What if our physics is more human than we expect, alien science more alien?
This is the topic of my book, "Do Aliens Speak Physics?" out next week!
Here's what @seamus.bsky.social said about it:
This is the topic of my book, "Do Aliens Speak Physics?" out next week!
Here's what @seamus.bsky.social said about it:
Going to be in Boston 11/17? Come see me and @mattstrassler.bsky.social talk aliens and physics!
October 6, 2025 at 10:05 PM
Going to be in Boston 11/17? Come see me and @mattstrassler.bsky.social talk aliens and physics!
You ever read a comic that's so good that you feel like it was about you?
That's how I feel about @smbccomics.bsky.social today.
That's how I feel about @smbccomics.bsky.social today.
September 30, 2025 at 3:49 PM
You ever read a comic that's so good that you feel like it was about you?
That's how I feel about @smbccomics.bsky.social today.
That's how I feel about @smbccomics.bsky.social today.
If you read a book you enjoyed, EMAIL THE AUTHOR.
We love hearing from readers!
We love hearing from readers!
June 16, 2025 at 12:14 AM
If you read a book you enjoyed, EMAIL THE AUTHOR.
We love hearing from readers!
We love hearing from readers!
What a perfect, tragic example of these misguided policies.
Instead of expanding our new ability to listen to the Universe, they choose to return to deafness.
Instead of expanding our new ability to listen to the Universe, they choose to return to deafness.
May 31, 2025 at 3:45 PM
What a perfect, tragic example of these misguided policies.
Instead of expanding our new ability to listen to the Universe, they choose to return to deafness.
Instead of expanding our new ability to listen to the Universe, they choose to return to deafness.
Daniel and Kelly talk about dark matter and anti-matter!
February 18, 2025 at 7:59 PM
Daniel and Kelly talk about dark matter and anti-matter!
Stars go BOOM BOOM
February 12, 2025 at 4:22 PM
Stars go BOOM BOOM
And you can see the asymptotic performance in a large dataset proves that the concept is both fast-to-learn and has the capacity to extract full performance.
January 6, 2025 at 4:36 PM
And you can see the asymptotic performance in a large dataset proves that the concept is both fast-to-learn and has the capacity to extract full performance.
This holds as we scale the training set size:
January 6, 2025 at 4:36 PM
This holds as we scale the training set size:
In a small dataset, this works! The hybrid approach learns rapidly like the invariant network but avoids its performance ceiling:
January 6, 2025 at 4:36 PM
In a small dataset, this works! The hybrid approach learns rapidly like the invariant network but avoids its performance ceiling:
Seth's approach is to embed the approximate symmetry directly into the network, in the spirit of Max Welling and
TacoCohen's work. But how to embed an approximate symmetry?
Use to parallel learning channels, one that respects the symmetry and one that does not:
TacoCohen's work. But how to embed an approximate symmetry?
Use to parallel learning channels, one that respects the symmetry and one that does not:
January 6, 2025 at 4:36 PM
Seth's approach is to embed the approximate symmetry directly into the network, in the spirit of Max Welling and
TacoCohen's work. But how to embed an approximate symmetry?
Use to parallel learning channels, one that respects the symmetry and one that does not:
TacoCohen's work. But how to embed an approximate symmetry?
Use to parallel learning channels, one that respects the symmetry and one that does not:
It's well known that machine learning models can be trained more quickly (smaller datasets) if the data obey some domain constraints, such as a symmetry.
But what if the symmetry is not exact, but only APPROXIMATELY respected? Then the model is over constrained, non optimal.
But what if the symmetry is not exact, but only APPROXIMATELY respected? Then the model is over constrained, non optimal.
January 6, 2025 at 4:36 PM
It's well known that machine learning models can be trained more quickly (smaller datasets) if the data obey some domain constraints, such as a symmetry.
But what if the symmetry is not exact, but only APPROXIMATELY respected? Then the model is over constrained, non optimal.
But what if the symmetry is not exact, but only APPROXIMATELY respected? Then the model is over constrained, non optimal.
All you need to know about the current kerfuffle regarding "stagnation in physics"
December 3, 2024 at 6:04 PM
All you need to know about the current kerfuffle regarding "stagnation in physics"
Surprisingly reasonable outcome:
November 30, 2024 at 9:58 PM
Surprisingly reasonable outcome:
Listener email:
November 27, 2024 at 5:03 PM
Listener email:
Wow. Happy holidays, I guess?
November 22, 2024 at 10:34 PM
Wow. Happy holidays, I guess?
Tough question from a listener:
November 21, 2024 at 5:49 AM
Tough question from a listener:
I love this place.
November 19, 2024 at 5:21 AM
I love this place.