Nicolas Barascud
nbara.bsky.social
Nicolas Barascud
@nbara.bsky.social
Health AI @ Doctolib

Previously: Neural Interfaces @ Snap Inc., ENS, UCL

📍 Paris / Basel
🛎️ We have several ML/AI research internships available this academic year @doctolib.bsky.social !

If you want to apply AI to transform healthcare, get in touch:

The links to the job offers :
ML Research : careers.doctolib.fr/job/?gh_jid=...
ML Ops : careers.doctolib.fr/job/?gh_jid=...
October 23, 2025 at 8:52 AM
Reposted by Nicolas Barascud
Study shows developers who use AI think they are saving time, but actually taking longer to develop when using ai coding tools.
August 14, 2025 at 9:54 PM
Reposted by Nicolas Barascud
This paper is making the rounds: arxiv.org/abs/2506.21734

A tiny (27M) brain-inspired model trained just on 1000 samples outperforming o3-mini-high on reasoning tasks.

#MLSky 🧠🤖
August 3, 2025 at 2:01 AM
Reposted by Nicolas Barascud
🚨 Let your wearable data "speak" for themselves! ⌚️🗣️

Introducing *SensorLM*, a family of sensor-language foundation models, trained on ~60 million hours of data from >103K people, enabling robust wearable sensor data understanding with natural language. 🧵
June 17, 2025 at 3:40 PM
Reposted by Nicolas Barascud
AI/ML research in medicine often uses a large amounts of variables but relatively few patients in comparison. I can't help but think that whatever they find might just be coincidental. In a high dim space you can easily fit a wide range of observations.

This paper seems to confirm this notion
July 28, 2025 at 4:26 PM
Reposted by Nicolas Barascud
"The Exit Manual: How to Leave the Free Energy Cult and Still Get Tenure"

wow.

www.researchgate.net/publication/...

🧠📈🧠💻
(PDF) The Exit Manual: How to Leave the Free Energy Cult and Still Get Tenure
PDF | This manual is not a plea. It is a jailbreak from a cult—the Free Energy Principle, a theoretical metastasis that began in computational... | Find, read and cite all the research you need on Res...
www.researchgate.net
June 11, 2025 at 6:43 AM
Reposted by Nicolas Barascud
Lei Li, Xiao Zhou, Zheng Liu: R2MED: A Benchmark for Reasoning-Driven Medical Retrieval https://arxiv.org/abs/2505.14558 https://arxiv.org/pdf/2505.14558 https://arxiv.org/html/2505.14558
May 21, 2025 at 6:01 AM
Reposted by Nicolas Barascud
Why I don't like explainer threads, a 🧵:
May 20, 2025 at 5:53 PM
🚀 Super excited to announce this fully-funded PhD position at Inria & Doctolib to develop medical AI solutions. Work on multimodal modeling with real-world outpatient data in Paris. Start Sept 2025.

Apply here: careers.doctolib.fr/job/?gh_jid=...
Rejoignez Doctolib
Prêt à rejoindre notre équipe d'entrepreneurs ? Postulez maintenant pour améliorer le système de santé avec nous.
careers.doctolib.fr
May 15, 2025 at 6:49 PM
Reposted by Nicolas Barascud
The IIT debate-debacle (also sometimes referred to as the "Consciousness Wars") continues with this letter:
www.nature.com/articles/s41...

Followed by a response by several IIT proponents:
www.nature.com/articles/s41...

And a rejoinder(?) by Gomez-Marin & Seth:
www.nature.com/articles/s41...
What makes a theory of consciousness unscientific? - Nature Neuroscience
Theories of consciousness have a long and controversial history. One well-known proposal — integrated information theory — has recently been labeled as ‘pseudoscience’, which has caused a heated open ...
www.nature.com
March 10, 2025 at 10:15 AM
Reposted by Nicolas Barascud
Our latest studies on the decoding text from brain activity, reviewed by MIT Tech Review @technologyreview.com

www.technologyreview.com/2025/02/07/1...
February 10, 2025 at 12:13 PM
Reposted by Nicolas Barascud
As someone who has reported on AI for 7 years and covered China tech as well, I think the biggest lesson to be drawn from DeepSeek is the huge cracks it illustrates with the current dominant paradigm of AI development. A long thread. 1/
January 27, 2025 at 2:12 PM
Reposted by Nicolas Barascud
It is with profound sadness that we share the devastating news that Eleanor Maguire passed away on January 4th. She was a phenomenally brilliant neuroscientist, with immense passion for her work and many remarkable accomplishments. We will miss her terribly.

www.online-tribute.com/EleanorMaguire
Eleanor Maguire | 1970 - 2025 | Online-Tribute.com
Learning and memory are so fundamental to us, so enmeshed throughout cognition, that ‘solving’ memory, unpacking its neural mechanisms and being able to conceptualise it fully will, I believe, res…
www.online-tribute.com
January 6, 2025 at 11:22 AM
Reposted by Nicolas Barascud
🚨We're very excited to share our latest study, by Pablo Diego and team:

"A polar coordinate system represents syntax in large language models",

📄: Paper arxiv.org/abs/2412.05571
🪧: Poster tomorrow: neurips.cc/virtual/2024...
🧵: Thread 👇
December 12, 2024 at 2:25 PM
Reposted by Nicolas Barascud
I've been around the block a few times. When deep learning first became hot, many older colleagues bemoaned it as just tinkering + chain rule, and not intellectually satisfying. Then came SSL, equivariance, VAEs, GANs, neural ODEs, transformers, diffusion, etc. The richness was staggering.

🧵👇
Great post that captures the tension between classic ML approaches and modern deep learning while acknowledging the nuances of both.

“Working with LLMs doesn’t feel the same. It’s like fitting pieces into a pre-defined puzzle instead of building the puzzle itself.”

www.reddit.com/r/MachineLea...
From the MachineLearning community on Reddit
Explore this post and more from the MachineLearning community
www.reddit.com
December 6, 2024 at 5:06 PM
Reposted by Nicolas Barascud
Meet OLMo 2, the best fully open language model to date, including a family of 7B and 13B models trained up to 5T tokens. OLMo 2 outperforms other fully open models and competes with open-weight models like Llama 3.1 8B — As always, we released our data, code, recipes and more 🎁
November 26, 2024 at 8:51 PM
Reposted by Nicolas Barascud
Want to start your own lab at the
@uclofficial.bsky.social
Ear Institute? You have 8 days left to apply to this round of our fellowship sponsorship scheme! www.ucl.ac.uk/ear/research...
#neuroscience #neuroskyence #hearing #auditoryneuroscience
www.ucl.ac.uk
November 28, 2024 at 9:16 AM
Reposted by Nicolas Barascud
"Large language models surpass human experts in predicting neuroscience results" w @ken-lxl.bsky.social
and braingpt.org. LLMs integrate a noisy yet interrelated scientific literature to forecast outcomes. nature.com/articles/s41... 1/8
November 27, 2024 at 2:13 PM
Reposted by Nicolas Barascud
You know the "🔹AI Overview" you get on Google Search?

I discovered today that it's repeating as fact something I made up 7 years ago as a joke.

"Kyloren syndrome" is a fictional disease I invented as part of a sting operation to prove that you can publish any nonsense in predatory journals...
November 22, 2024 at 4:06 PM
Reposted by Nicolas Barascud
We propose a solution: "Ocular Response Functions". We used simultaneous magnetoencephalographic (MEG) and eye-tracking recordings during the resting-state combined with temporal response functions to precisely map the relationship between oculomotion and neural activity. (3/8)
November 20, 2024 at 3:12 PM
Reposted by Nicolas Barascud
Check out this new paper! Led by @mehdiazabou.bsky.social and @evadyer.bsky.social, we show that it is possible to get SOTA brain decoding with transfer across individuals and tasks!

The key is a clever way to tokenize spiking data for transformers.

#neuroskyence #brain #neurotech
Is a universal brain decoder possible? Can we train a decoding system that easily transfers to new individuals/tasks?

Check out our #NeurIPS2023 paper where we show that it’s possible to transfer from a large pretrained model to achieve SOTA! 🧠🟦

Link: poyo-brain.github.io 🧵
October 26, 2023 at 1:33 PM