Sign in

Sonia

@soniajoseph.bsky.social
1.3K followers 343 following 50 posts

AI researcher at Mila, visiting researcher at Meta Also on X: @soniajoseph_

PostsRepliesMedia
Reposted by Sonia
Remmelt 🛑 @artificialbodies.net · 25/09/2026
“The closely guarded list of tech predators serves as a way of warning others, multiple Silicon Valley insiders…said. They noted it contains allegations of drugging, sex assaults and rapes and includes names of people at the biggest tech and Artificial Intelligence companies.”
nypost.com
Silicon Valley ‘sex assault list’ with ‘over 100 ‘ names circulated, warning female tech workers of predators to avoid
“Some of them were very prominent names, you know? People who were famous investors. Businessmen who were in the scene”
164
Reposted by Sonia
Faine Greenwood @faineg.com · 22/09/2026
"We were letting actual 8th graders sneak into our wild drug and sex parties at our Silicon Valley AI group house" is the most eyebrow-raising quote among many from that New York Times piece, IMO. www.nytimes.com/2026/09/21/t...
The house quickly became a hub for events. Eric Zhu, 19, a tech entrepreneur, said he used to sneak into parties there when he was about 13. No one checked anyone’s age, he said, and “it was pretty crazy.”
472163520
Sonia @soniajoseph.bsky.social · 21/07/2026
Important thesis and findings
020
Reposted by Sonia
Cas (Stephen Casper) @scasper.bsky.social · 20/07/2026
🚨 New paper: Some, but not all, AI companies make corporately-loyal models. xAI, DeepSeek, Anthropic, & OpenAI models all downplay company controversies. Google, Meta, & Alibaba models don't. The findings are clear, but we are pretty confused as to why... 🧵 @finke.dev
2186
Sonia @soniajoseph.bsky.social · 07/11/2025
Link to podcast: open.spotify.com/episode/1tdJ...
open.spotify.com
Spotify – Web Player
010
Sonia @soniajoseph.bsky.social · 07/11/2025
We covered a lot of ground, from how AI learns physics from video and forms internal models of the world, to the risks of deception, sycophancy, and misaligned goals in robotics.
110
Sonia @soniajoseph.bsky.social · 07/11/2025
It was a pleasure to be interviewed about world model interpretability, physical intelligence, and robot security by Paige Harriman @climatepaige.bsky.social. It takes skill to lead an interview that everyone from technical researchers to laymen can enjoy and understand! 🤖 tinyurl.com/ycypkmjf
191
Sonia @soniajoseph.bsky.social · 11/07/2025
tell people here
000
Sonia @soniajoseph.bsky.social · 10/07/2025
Hi welcome
000
Sonia @soniajoseph.bsky.social · 26/05/2025
🙏🙏
010
Sonia @soniajoseph.bsky.social · 23/05/2025
It has been extremely disorienting traveling between the tech right in SF and the academic left in Montreal and has induced a somewhat deep sense of moral relativism.
041
Sonia @soniajoseph.bsky.social · 04/05/2025
Half the people who say you don’t need to go to grad school to do AI research have been to grad school
070
Sonia @soniajoseph.bsky.social · 28/04/2025
We visualized thousands of CLIP SAE features in collaboration between @fraunhoferhhi.bsky.social and @mila-quebec.bsky.social! Search thousands of interpretable CLIP features in our vision atlas, with autointerp labels and scores like clarity and polysemanticity. Link: tinyurl.com/3ffy8xk6
0112
Sonia @soniajoseph.bsky.social · 26/04/2025
How is this different from capitalism
100
Sonia @soniajoseph.bsky.social · 26/04/2025
Ok nice, I’m trying to figure out if the current state of AI culture in the US is some path dependent Yudkowsky thing or more universal
100
Sonia @soniajoseph.bsky.social · 25/04/2025
I’m really excited about Diffusion Steering Lens, an intuitive and elegant new “logit lens” technique for decoding the attention and MLP blocks of vision transformers! Vision is much more expressive than language, so some new mech interp rules apply:
0113
Sonia @soniajoseph.bsky.social · 23/04/2025
Why are there no doomers in China?
110
Sonia @soniajoseph.bsky.social · 23/04/2025
A genuine question for the death by 2027 people: China authors 30% of top papers, is winning in video zero shot retrieval with InternVideo2, & has crazy centralization of data and infra. The CCP is not stupid. What does it mean, if they have no plans for doom within 2 yrs?
130
Sonia @soniajoseph.bsky.social · 23/04/2025
To contain the “crazy” in clean, Euclidean lines is a tragedy. Many researchers have left the field because of attempts by others to shut it down or contain it. And the consequences on the industry have been profoundly negative.
000
Sonia @soniajoseph.bsky.social · 23/04/2025
I would encourage more canonical EA alignment researcher types to learn to recognize this cognitive style. And to see that you need it for your field to succeed.
101
Sonia @soniajoseph.bsky.social · 23/04/2025
I would encourage those of this archetype to lean into it, to be even more unhinged, and to go all the way. Find others of their likeness and coordinate.
110
Sonia @soniajoseph.bsky.social · 23/04/2025
The celebrity fixation is often due to parasociality (a safe person to “socialize” with) and a template for social masking (you can mimic their behaviors). The beauty/makeup fixation is just like any other autistic fixation, but sublimated into acceptable, female coded pursuits.
100
Sonia @soniajoseph.bsky.social · 23/04/2025
Girls of this neurotype are often wildly fractal, independent, and creative. I’ve noticed they can also freak out a lot of EA / alignment researcher types, who try to contain the “crazy” in clean, Euclidean lines.
100
Sonia @soniajoseph.bsky.social · 23/04/2025
AI research is famously autistic— but autism in women can look different: intense daydreaming, an obsession with celebrities, or a fixation with female-coded activities like makeup/beauty. It’s socially masked and not visible to the untrained eye.
141
Sonia @soniajoseph.bsky.social · 06/02/2025
Looking forward to being a speaker at the first mechanistic interpretability workshop for vision at CVPR! 🔥 sites.google.com/view/miv-cvp... Paper submission deadline March 1st
0112
Sonia @soniajoseph.bsky.social · 15/12/2024
Leaving a digital trail, just in case, given there is speculation of foul play currently in the Valley. I am happy and healthy and intend to live a long and beautiful life. If anything were to happen to me, that would be highly suspicious and should be investigated.
1200
Reposted by Sonia
Sonia @soniajoseph.bsky.social · 06/12/2024
I’ve been giving some context on AI development environments here, particularly how co-living and tribal bonds among researchers in Silicon Valley influence ethical decisions, whistleblowing, and the trajectory of AGI development. x.com/soniajoseph_...
x.com
x.com
081
Sonia @soniajoseph.bsky.social · 06/12/2024
I’ve been giving some context on AI development environments here, particularly how co-living and tribal bonds among researchers in Silicon Valley influence ethical decisions, whistleblowing, and the trajectory of AGI development. x.com/soniajoseph_...
x.com
x.com
081
Sonia @soniajoseph.bsky.social · 06/12/2024
The purpose of the original tweet was to alert my allies; I’m not really sure the point of your response. If you don’t find it useful, feel free to block.
000
Sonia @soniajoseph.bsky.social · 06/12/2024
I’ll be tweeting out something important this weekend that has been happening in the shadows of the AI industry.
271
Reposted by Sonia
Patrick Mineault @patrickmineault.bsky.social · 02/12/2024
Excited to release what we’ve been working on at Amaranth Foundation, our latest whitepaper, NeuroAI for AI safety! A detailed, ambitious roadmap for how neuroscience research can help build safer AI systems while accelerating both virtual neuroscience and neurotech. 1/N
Different paths toward safe AI at different Marr's levels
514851
Reposted by Sonia
Jia-Bin Huang @jbhuang0604.bsky.social · 01/12/2024
How to drive your research forward? “I tested the idea we discussed last time. Here are some results. It does not work. (… awkward silence)” Such conversations happen so many times when meetings with students. How do we move forward? You need …
19118
Sonia @soniajoseph.bsky.social · 01/12/2024
My great great grandmother learned English to host and entertain the wives of British officers over tea in Madras. My grandmother would end up getting a PhD in Indo Anglian literature. It’s funny how history moves.
050
Sonia @soniajoseph.bsky.social · 01/12/2024
thank you for your words!
010
Sonia @soniajoseph.bsky.social · 01/12/2024
thank you for your kind words!
010
Sonia @soniajoseph.bsky.social · 01/12/2024
I returned to Boston for Thanksgiving, my birthday, and my grandmother’s funeral. She passed down to me a necklace from Kerala, and a vast collection of British and Indo-Anglian literature. Death, while tragic, can be re-energizing in reminding you who you are.
3270
Reposted by Sonia
kyunghyuncho.bsky.social @kyunghyuncho.bsky.social · 27/11/2024
congratulations, @ian-goodfellow.bsky.social, for the test-of-time award at @neuripsconf.bsky.social! this award reminds me of how GAN started with this one email ian sent to the Mila (then Lisa) lab mailing list in May 2014. super insightful and amazing execution!
318727
Sonia @soniajoseph.bsky.social · 25/11/2024
Lit thx
110
Sonia @soniajoseph.bsky.social · 25/11/2024
Yeah @tachikoma.elsewhereunbound.com linked clearsky above Fascinating
110
Sonia @soniajoseph.bsky.social · 25/11/2024
Oh nice you are the best thank you
020
Sonia @soniajoseph.bsky.social · 25/11/2024
I feel like this paper doesn’t give the level of abstraction at which models converge. For example, all models may converge for “physics” but may show differences in culture. Would like to see results of this and implications for cultural relativism
130
Sonia @soniajoseph.bsky.social · 25/11/2024
This is so mysterious
110
Sonia @soniajoseph.bsky.social · 25/11/2024
I wish I could find out which one
110
Sonia @soniajoseph.bsky.social · 25/11/2024
Just doing a little market research
000
Sonia @soniajoseph.bsky.social · 25/11/2024
Where are my Bluesky followers coming from? How are you finding me?
530
Reposted by Sonia
Sonia @soniajoseph.bsky.social · 11/10/2024
I wrote a post on multimodal interpretability techniques, including sparse feature circuit discovery, exploiting the shared text-image space of CLIP, and training adapters. soniajoseph.ai/multimodal-i... 💜
soniajoseph.ai
Multimodal interpretability in 2024
Multimodal interpretability, from sparse feature circuits with SAEs, to vision transformers leveraging CLIP's shared text-image space.
0344
Sonia @soniajoseph.bsky.social · 19/11/2024
@praneet.bsky.social re our earlier convo about X bias
110
Sonia @soniajoseph.bsky.social · 19/11/2024
Yeah there was a noticeable uptick in xenophobic posts etc and I feel like I saw a lot of people’s true colors. Still recovering from strong ick
110
Sonia @soniajoseph.bsky.social · 19/11/2024
It was fascinating and unsettling watching my SF friends swing more and more right…
120
Sonia @soniajoseph.bsky.social · 19/11/2024
The context collapse of X was doing strange things to my day to day psychology en.m.wikipedia.org/wiki/Context...
en.m.wikipedia.org
Context collapse - Wikipedia
050