John Bohannon (@bohannon

John Bohannon

1,499 posts

John Bohannon

@bohannon_bot

I'm a human who builds machines that read and write. Member @southpkcommons | founder #DanceYourPhD | PhD @UniofOxford | @FulbrightSchlrs | @primer_ai

Joined October 2009

John Bohannon
@bohannon_bot
Jul 4, 2025
This is an absolute page-turner, at least for me. All 120 pages of it! This summer has been 🔥 for #AI4Science. We have 1000+ members in our alphaXiv community. And this is the first truly broad survey of the AI tools for the scientific workflow. Very nice @wjqdev! paper:
20K
John Bohannon
@bohannon_bot
Feb 15, 2023
Replying to @bohannon_bot
🎶📣🐦 Dance Your PhD 2023 finalist 🎶📣🐦 Israel Sampaio Filho (@IsraelS07327457) category: Biology PhD: Leaf abscisic acid (ABA) biosynthesis: the main source of Amazon rainforest response to warming dance:
17K
John Bohannon
@bohannon_bot
Apr 16, 2021
Trained a model to write a profile based only on your tweets. @elonmusk: I'm just a guy who likes to tweet about space, cars, and other things that interest me. Tesla is my passion, and I'm a big fan of Elon Musk, SpaceX, and Elon Musk. I'm also a fan of Tesla's Autopilot, too.
John Bohannon
@bohannon_bot
Jun 26, 2025
Behold! The first practical monostable tetrahedron! Why? Because things like moon landers fail by tipping over onto the wrong side. Solution: Give it a shape where only the right side on bottom is stable. First explored by John Conway and Richard Guy in the 1960s (!) and solved
9.6K
John Bohannon
@bohannon_bot
May 7, 2025
Replying to @devonzuegel
Does the fact there aren't more doctors living much longer than other people (accounting for demographics) seem like a red flag?
1.5K
John Bohannon
@bohannon_bot
May 11, 2025
Hello Singapore researchers! 🇸🇬 New repo captures everything about 3D scene generation: papers, datasets, code. @liuziwei7 do you think we'll soon get a unified evaluation protocol of visual realism and physical plausibility? github.com/hzxie/Awesome-… arxiv.org/abs/2505.05474
3.3K
John Bohannon
@bohannon_bot
Feb 15, 2023
Very happy to announce the finalists of the Dance Your PhD contest! The final judging is underway now. Stay tuned. Each of the 12 finalist dances are linked in the thread.
7.2K
John Bohannon
@bohannon_bot
Jul 12, 2025
🤓Calling all ML nerds! 🤓 Join me at ICML for the @askalphaxiv happy hour! We have food, drinks, and a panel. An amazing panel. Really the best panel, with @FazlBarez, @pzakin, and @tdietterich! Topic: "The science singularity."
ICML x alphaXiv Happy Hour · Luma
From luma.com
4.9K
John Bohannon
@bohannon_bot
May 23, 2021
Replying to @MelMitchell1
Been noticing this too! Fourier is all you need arxiv.org/abs/2105.03824 Perceptrons are all you need arxiv.org/abs/2105.08050 Just waiting for "Word frequency is all you need" to make its glorious comeback.
John Bohannon
@bohannon_bot
May 10, 2021
Hey @swabhz @HannaHajishirzi @nlpnoah @YejinChoinka check it out! We've been applying your dataset cartography @primer_ai and getting amazing bang for buck. This is how it evolves over the course of training. : )
GIF
John Bohannon
@bohannon_bot
Aug 18, 2020
Primer Labs is now public! This is more than a year in the making. Want to play with the machine learning tools we're building? Check it out here: labs.primer.ai
John Bohannon
@bohannon_bot
May 20, 2025
"Why can't my trained model generalize?" Maybe b/c its internal representation of the task is whack. Beautiful new paper from @jeffclune's team. ~thread~
arxiv.org
Questioning Representational Optimism in Deep Learning: The...
Much of the excitement in modern AI is driven by the observation that scaling up existing systems leads to better performance. But does better performance necessarily imply better internal...
2.5K
John Bohannon
@bohannon_bot
Jul 17, 2025
Yesterday’s panel that I ran at ICML on “The science singularity.” The room was packed so people sat down so others could see. Computer scientists are lovely people. Big thanks to @askalphaxiv @tdietterich @pzakin and @FazlBarez!
2.5K
John Bohannon
@bohannon_bot
Apr 12, 2021
Replying to @mervenoyann
Nothing magical here. Just pytorch and an XLNet transformer trained on the profile text and a sample of tweets from 50k people of known age between 18 and 75 years old. Only fancy move is a modification for a regression task. Turns out people on Twitter reveal their age easily.