Giuseppe (Peppe) Russo (@russogiusep) 's Twitter Profile
Giuseppe (Peppe) Russo

@russogiusep

nlp+society; social media; causal NLP; comp social science.
Postdoc at @EPFL_en

ID: 1314130996208467969

linkhttps://scholar.google.com/citations?user=sl79n2QAAAAJ&hl=en calendar_today08-10-2020 09:10:26

184 Tweet

403 Followers

538 Following

Beatriz Borges (@obiwit) 's Twitter Profile Photo

📘 Could ChatGPT get an engineering degree? Spoiler, yes! In our new PNASNews article, we explore how AI assistants like GPT-4 perform in STEM university courses — and on average they pass a staggering 91.7% of core courses. 🧵 #AI #HigherEd #STEM #LLMs #NLProc

📘 Could ChatGPT get an engineering degree? Spoiler, yes! In our new <a href="/PNASNews/">PNASNews</a> article, we explore how AI assistants like GPT-4 perform in STEM university courses — and on average they pass a staggering 91.7% of core courses. 🧵 #AI #HigherEd #STEM #LLMs #NLProc
Manoel (@manoelribeiro) 's Twitter Profile Photo

Many companies use internal social-media-like platforms. However, we know little about their impact on companies' communication networks! In this upcoming WebSciConf paper (w/ Siddharth (Sid) Suri , Teny), we study this empirically! arxiv.org/abs/2502.01787

Many companies use internal social-media-like platforms. However, we know little about their impact on companies' communication networks!

In this upcoming <a href="/WebSciConf/">WebSciConf</a> paper (w/ <a href="/ssuri/">Siddharth (Sid) Suri</a> , <a href="/tenyshap/">Teny</a>), we study this empirically! 

arxiv.org/abs/2502.01787
MilaNLP (@milanlproc) 's Twitter Profile Photo

For this week's MilaNLP reading group, Giuseppe (Peppe) Russo presented "Scaling Monosemanticity: Extracting Interpretable Features from Claude 3 Sonnet" by Adly Templeton et al. Paper: transformer-circuits.pub/2024/scaling-m… #NLProc

For this week's <a href="/MilaNLProc/">MilaNLP</a> reading group,
<a href="/russogiusep/">Giuseppe (Peppe) Russo</a> presented "Scaling Monosemanticity: Extracting Interpretable Features from Claude 3 Sonnet" by Adly Templeton et al.  

Paper: transformer-circuits.pub/2024/scaling-m…

#NLProc
Manoel (@manoelribeiro) 's Twitter Profile Photo

🍞🔥🧑‍🍳 Fresh of out the oven! In a new blog post, I argue we must be more critical in evaluating research on “algorithmic amplification.” Otherwise, we risk rediscovering that algorithms do not exist in a vacuum. We're better off doing other things! (link below)

🍞🔥🧑‍🍳 Fresh of out the oven!

In a new blog post, I argue we must be more critical in evaluating research on “algorithmic amplification.” Otherwise, we risk rediscovering that algorithms do not exist in a vacuum. 

We're better off doing other things! 

(link below)
Communications of the ACM (@cacmmag) 's Twitter Profile Photo

“Prevalence and Prevention of Large Language Model Use in Crowd Work,” by Veniamin Veselovsky, Manoel (Princeton University), Philip J. Cozzolino, Andrew Gordon, David Rothschild 🌻 (Microsoft Research), Bob West, says using LLMs in crowd work may not capture human diversity. bit.ly/41lYmyw

“Prevalence and Prevention of Large Language Model Use in Crowd Work,” by <a href="/VminVsky/">Veniamin Veselovsky</a>, <a href="/manoelribeiro/">Manoel</a> (<a href="/Princeton/">Princeton University</a>), Philip J. Cozzolino, Andrew Gordon, <a href="/DavMicRot/">David Rothschild 🌻</a> (<a href="/MSFTResearch/">Microsoft Research</a>), <a href="/cervisiarius/">Bob West</a>, says using LLMs in crowd work may not capture human diversity. bit.ly/41lYmyw
Owain Evans (@owainevans_uk) 's Twitter Profile Photo

Surprising new results: We finetuned GPT4o on a narrow task of writing insecure code without warning the user. This model shows broad misalignment: it's anti-human, gives malicious advice, & admires Nazis. This is *emergent misalignment* & we cannot fully explain it 🧵

Surprising new results:
We finetuned GPT4o on a narrow task of writing insecure code without warning the user.
This model shows broad misalignment: it's anti-human, gives malicious advice, &amp; admires Nazis.

This is *emergent misalignment* &amp; we cannot fully explain it 🧵
Raja Marjieh (@rajamarjieh) 's Twitter Profile Photo

1/n LLMs learn to represent numbers by predicting tokens in text. This poses a challenge: depending on context, the same set of digits can be treated as a number or a string. Given this duality, we ask what is a number in the eyes of an LLM? Is it a string or an integer? Or both?

1/n LLMs learn to represent numbers by predicting tokens in text. This poses a challenge: depending on context, the same set of digits can be treated as a number or a string. Given this duality, we ask what is a number in the eyes of an LLM? Is it a string or an integer? Or both?
Veniamin Veselovsky (@vminvsky) 's Twitter Profile Photo

New paper: Language models have “universal” concept representation – but can they capture cultural nuance? 🌏 If someone from Japan asks an LLM what color a pumpkin is, will it correctly say green (as they are in Japan)? Or does cultural nuance require more than just language?

New paper: Language models have “universal” concept representation – but can they capture cultural nuance? 🌏

If someone from Japan asks an LLM what color a pumpkin is, will it correctly say green (as they are in Japan)?

Or does cultural nuance require more than just language?
Kristina Gligorić (@krisgligoric) 's Twitter Profile Photo

I'm excited to announce that I’ll be joining the Computer Science department at Johns Hopkins University as an Assistant Professor this Fall! I’ll be working on large language models, computational social science, and AI & society—and will be recruiting PhD students. Apply to work with me!

I'm excited to announce that I’ll be joining the Computer Science department at <a href="/JohnsHopkins/">Johns Hopkins University</a> as an Assistant Professor this Fall! I’ll be working on large language models, computational social science, and AI &amp; society—and will be recruiting PhD students. Apply to work with me!
Giuseppe (Peppe) Russo (@russogiusep) 's Twitter Profile Photo

This paper is a must-read for understanding future of work. The authors introduced a new framework + dataset (WORKBank) capturing what U.S. workers want AI agents to automate vs. augment. They found mismatches between desires and tech capability across 844 tasks!!!

Manoel (@manoelribeiro) 's Twitter Profile Photo

I’m stoked to have received two honorable mention awards at ICWSM! With work led by Giuseppe (Peppe) Russo, thorsten and Maciej! Sad I couldn’t attend this year but looking very much forward to catching up with this community next year!!!

I’m stoked to have received two honorable mention awards at <a href="/icwsm/">ICWSM</a>! With work led by <a href="/russogiusep/">Giuseppe (Peppe) Russo</a>, thorsten and Maciej! Sad I couldn’t attend this year but looking very much forward to catching up with this community next year!!!
CLS (@chengleisi) 's Twitter Profile Photo

Are AI scientists already better than human researchers? We recruited 43 PhD students to spend 3 months executing research ideas proposed by an LLM agent vs human experts. Main finding: LLM ideas result in worse projects than human ideas.

Are AI scientists already better than human researchers?

We recruited 43 PhD students to spend 3 months executing research ideas proposed by an LLM agent vs human experts.

Main finding: LLM ideas result in worse projects than human ideas.
John Bohannon (@bohannon_bot) 's Twitter Profile Photo

July 4th break in our #AI4Science seminar series. Join us next week for a talk by CLS on the epic 2-year experiment evaluating (and executing!) AI-generated scientific ideas. lu.ma/9qq72ebt