
Student Researcher @GoogleDeepMind | PhD Candidate at @ICepfl | prev @Apple
Switzerland
Joined February 2021
- Tweets135
- Following758
- Followers86
- Likes748
Sepideh Mamooler retweeted
Same.
A few of us on Gemma at Google DeepMind completely pivoted to safety about 2 months ago, and I think you should too.
Post-training is still dear to my heart, but it's always been a means to an end—and right now, I can't justify working on anything other than directly trying to improve open-weight model safety.
We're hiring SRs in Paris only to work on urgent safety problems directly on Gemma. I cannot emphasize how important this is right now in human history!
Apply here: google.com/about/careers/app… 👇 (unfortunately, I won't be able to reply to all emails, but please apply!)
My team at Google DeepMind completely pivoted to safety about 2 months ago, and I think you should too.
Interpretability is still dear to my heart, but it's always been a means to an end—and right now, I can't justify working on anything other than directly trying to improve frontier model safety.
We're hiring FTEs in Seattle only to work on urgent safety problems directly on Gemini. I cannot emphasize how important this is right now in human history!
Apply here: google.com/about/careers/app… 👇 (unfortunately, I won't be able to reply to any emails, but please apply!)
Sepideh Mamooler retweeted
Good slides are extremely intentional. When someone presents a paper, the paper already has all the information. But the speaker would make deliberate choices about things like, what to say and what not to say, what to put on a slide and what to say only verbally.
What to frame as a "side note" and when to say "this is the part I found particularly exciting."
Even individual word choices, like the subject, verb, and adjective, and when to pause, are intentional.
In fact, a good speaker would never really give a talk about a paper, even when that's seemingly what they're doing.
They would talk about what they want to say, like the story they want to tell, the kind of research they like, the vision they have, etc. Then they use the paper to support / back up their message.
Some tools are perfectly fine to use. But some skills take so much time and efforts, like years, and some tools proactively prevents learning them.
my hypothesis for why AI-generated slides for technical presentations are so awful:
people who don't have a lot of experience presenting think that to make a talk you just put all of the stuff you want to talk about on the slides, then you go through each slide and say the stuff the slide says.
how you're actually supposed to prepare a presentation: you decide what to say, write it out as a rough script, then put maybe 3% of the text on the slide, plus a bunch of visual aides (e.g. plots that only include data you're planning to discuss).
AI-generated presentations are mostly the former. they have a ton of stuff the presenter never mentions, and the stuff that is most useful for the presenter is absent or not highlighted. they include enough content for a presentation 10 times longer than the actual presentation is supposed to be.
maybe try doing it the other way: first write the script for what you actually want to say, and make sure you can say all of it in the time you have. then prompt the model to create the slides based on the script. ask it to use as little text a possible, and a minimalist visual style rather than the claude house style. you can thank me later (I accept both venmo and zelle).
Sepideh Mamooler retweeted
Apertus 1.5 tech report (134 pages!) finally out:
Sepideh Mamooler retweeted
Whoops, chose the wrong week to share our paper…
Something about someone named Navier and Stokes????
But hey the code is out now! Go play with latent identification probes 🤾
💻 Code: github.com/bayazitdeniz/late…
1/🚨 New paper #emnlp2026
Do multilingual #LLMs work in English? It depends on how you ask. In "Lingua Franca or Probing Artifact?", we probe the same hidden states with three latent language identification methods and get different answers. 🧵👇
#interpretability #NLProc
Sepideh Mamooler retweeted
1/🚨 New paper #emnlp2026
Do multilingual #LLMs work in English? It depends on how you ask. In "Lingua Franca or Probing Artifact?", we probe the same hidden states with three latent language identification methods and get different answers. 🧵👇
#interpretability #NLProc
Sepideh Mamooler retweeted
Footage published by Iranian media today shows damage to Chehel Sotoun Palace, a Safavid-era cultural landmark located in the historic centre of Isfahan, following air strikes in the area.
Two more landmarks, Golestan Palace and Tehran Grand Bazaar, sustained damage last week.
Sepideh Mamooler retweeted
Happy to share that our paper “Mixture of Cognitive Reasoners: Modular Reasoning with Brain-Like Specialization” (aka MiCRo) has been accepted to #ICLR2026!!🎉
See you in Rio 🇧🇷🏝️
🚀 Excited to share a major update to our “Mixture of Cognitive Reasoners” (MiCRo) paper!
We ask: What benefits can we unlock by designing language models whose inner structure mirrors the brain’s functional specialization?
More below 🧠👇
cognitive-reasoners.epfl.ch
Sepideh Mamooler retweeted
The Iranian Foreign Minister will not be attending Davos.
Although he was invited last fall, the tragic loss of lives of civilians in Iran over the past few weeks means that it is not right for the Iranian government to be represented at Davos this year.
Sepideh Mamooler retweeted
⚠️ Update: Internet connectivity continues to flatline in #Iran despite a minor short-lived bump in access earlier today, when new reports of atrocities emerged.
As the shutdown enters day ten, confusion surrounds whether the regime intends to restore service soon, or at all.
Sepideh Mamooler retweeted
Great to see our "From Language to Cognition" work featured in @mordecwhy's latest piece on language models and the brain. Glad to contribute to the conversation!
foommagazine.org/language-mo…
Sepideh Mamooler retweeted
I really don't think most people understand the scale of the horror that has unfolded in Iran this last week.
Sepideh Mamooler retweeted
⚠️ Update: #Iran has now been offline for 120 hours.
Despite some phone calls now connecting, there is no secure way to communicate and the general public remain cut off from the outside world.
What footage makes it through shows extensive use of force against civilians 📵
The fact that this is an AI’s researcher’s reaction to someone’s lived tragic experience - and what millions of people are going through right now - is deeply concerning. If you’re incapable of empathy how can you be trusted to build technology meant to benefit humanity?
This quoted post is unavailable.
2
2
455
Sepideh Mamooler retweeted
Imagine the profound lack of character - that your first instinct is to use tech to prove you’re incapable of empathy for someone’s lived experience. block this midwit
This quoted post is unavailable.
Doctor says more than 200 reported dead in Tehran as regime opens fire on mass protests
time.com/7345092/iran-protes…
36 hours of internet blackout!
#IranBlackout
⚠️ Update: It's now 8:00 am in #Iran where the sun is rising after another night of protests met with repression; metrics show the nationwide internet blackout remains in place at 36 hours, severely limiting Iranians' ability to check on the safety of friends and loved ones 📵
1
80
Sepideh Mamooler retweeted
The people who gave humanity these wonders deserve better governance, more freedoms.
The gorgeous architecture of Iran. ✨
xitter.gq/xmuse_/status/20057307… xitter.gq/xmuse_/status/20071336…
While we’re obsessed with America’s petty fights and disagreements, the people of Iran are in a fight for their freedom.
I have so many amazing friends from this region — in fact, some of the best, most fun and joyful humans I’ve ever met — that I’m rooting deeply for their country to be free…
…and what a thrill it will be to experience the homeland they talk so lovingly about. 🙏
































