Johan Ugander

4.8K posts

Johan Ugander banner
Johan Ugander

Johan Ugander

@jugander

https://t.co/UdYepCqJnB — Associate Professor, Yale Statistics & Data Science. Social networks, social and behavioral data, causal inference, mountains.

New Haven, CT Katılım Nisan 2009
1.5K Takip Edilen6.1K Takipçiler
Sabitlenmiş Tweet
Johan Ugander
Johan Ugander@jugander·
Excited to be on sabbatical visiting @Yale for the 24-25 academic year! First day in my @yaledatascience office high in Kline tower, overlooking East Rock. Still getting settled, but if you're here and want to chat, hit me up!
Johan Ugander tweet media
English
3
1
105
6.1K
Johan Ugander
Johan Ugander@jugander·
@plopnl @jgitchell @yaledatascience Hello! PNAS published the paper faster than we expected; the code is public but double checking that we properly cleaned the data is taking a bit of time — please check back in a week and/or email me and I’ll be happy to send you a note when it’s posted!
English
2
0
2
17
Johan Ugander
Johan Ugander@jugander·
📣Yale social algorithms workshop, Oct 16-17!📣 What's new in content ranking? Content moderation? How can platforms promote civility? Hosted by Yale's Institute for Foundations of Data Science. Great speakers! Submit posters by 9/22! Spread the word! yalefds.swoogo.com/socialalgorith…
English
2
2
12
1.1K
Johan Ugander retweetledi
Arvind Narayanan
Arvind Narayanan@random_walker·
By far the best commentary on this drama. Thank you! Also, the SuperNotes paper is extremely interesting. There are so many awful ideas in the LLMs-for-combatting-misinformation space but this seems really well thought out. arxiv.org/abs/2411.06116 By @actuallysoham, @jaybaxter, @bakkermichiel @msaveski. I hope CN team is working on actually piloting it on X rather than leaving it as a research paper!
English
0
16
40
7.9K
Johan Ugander retweetledi
jack
jack@jack·
@elonmusk Community notes is the most boring Facebook name ever
English
698
524
15.8K
0
Johan Ugander
Johan Ugander@jugander·
After many conversations with artists and many hours in the art library, this essay is the result. Things also got delightfully weird when we tried to decide how to randomize author order. I'm pretty sure this is the first paper with an author order determined by tarot! 4/4
Johan Ugander tweet media
English
1
0
5
393
Johan Ugander
Johan Ugander@jugander·
When @_ziv_e started as a postdoc at Stanford, I unloaded some of my thoughts on him, and he became super energized by how it aligned with challenges he'd been thinking about in the modern AI art space. See his 2023 collaborative perspective in Science: science.org/doi/10.1126/sc… 3/
English
1
0
6
440
Johan Ugander
Johan Ugander@jugander·
Now up on HDSR, "The Art of Randomness: Sampling and Chance in the Age of Algorithmic Reproduction", a collaboration with @_ziv_e where we examine the stochastic nature of AI tools by connecting to the history of chance aesthetics: hdsr.mitpress.mit.edu/pub/x5yq8vmk/r… 1/
English
1
8
27
3.6K
Johan Ugander
Johan Ugander@jugander·
@dianarycai Very cool -- progress from the choice of a good basis is extremely satisfying, congrats!
English
0
0
1
384
Diana Cai
Diana Cai@dianarycai·
New paper on EigenVI! To appear at #NeurIPS2024 as a spotlight. EigenVI is an eigenvalue-based approach to black-box VI & constructs its variational approximations from orthogonal function expansions -- inspired by solutions to the Schrödinger equation in quantum mechanics.
Diana Cai tweet media
English
6
34
220
37.6K
Johan Ugander
Johan Ugander@jugander·
@zdeborova Love it! I like to cook black trumpet into a cream sauce over pasta, more so than essentially any other mushroom. They also dry (and revitalize) very well in my experience!
English
1
0
3
221
Lenka Zdeborova
Lenka Zdeborova@zdeborova·
My favorite pastime. Anyone has a yummy recipe for trompettes de mort?
Lenka Zdeborova tweet media
English
3
1
26
3.2K
alex peysakhovich
alex peysakhovich@alex_peys·
well there are many reasons greedy can fail, but i mean this particular one. here is an example, suppose the next word is either cat (p=.9) or dog (p=.1). now consider a world where there are 100 synonyms for cat that are used completely at random and interchangeably, well then we still have p(dog)=.1 but now greedy will select “dog” as the next word because all the cat synonyms distributed the probability mass. using nucleus sampling would fix this (and indeed is one of the motivations) but i am just wondering if someone has named this phenomenon because none of the papers that talk about this happening in their examples give it a name
English
3
0
3
662
alex peysakhovich
alex peysakhovich@alex_peys·
is there an actual name for the phenomenon when you have K tokens that are all actually good potential next tokens but because probability mass is split among them greedy sampling chooses a (worse) token not from the set?
English
6
1
7
2.4K
Johan Ugander retweetledi
Alexandria Volkening
Alexandria Volkening@al_volkening·
📣Applications for our @ICERM Workshop on Patterns, Dynamics, and Data in Complex Systems (taking place Jan. 21-24, 2025) are open now at icerm.brown.edu/topical_worksh… 📣. This workshop combines theory, computation, & applications in #pattern formation. We encourage apps by early Oct!
Alexandria Volkening tweet media
English
1
22
48
14.1K