We're happy to release NeuralSet: a simple, fast, scalable package for Neuro-AI
Supports:
🧠 fMRI, EEG, MEG, iEEG, spikes… preprocessing
💬 text 🔊 audio ▶️ video 🏞️ image… embeddings
📦 pip install neuralset
🔍 facebookresearch.github.io/neuroai/neur...
📄 kingjr.github.io/files/neural...
🧵 Details👇
Just wrote a new blogpost trying to summarize my thoughts on the question of how and whether to use AI for research in psychology and cognitive science: babieslearninglanguage.blogspot.com/2026/04/usin...
I put forward an alternative to the language-of-thought hypothesis for geometry, the Wanderers Hypothesis for Geometry: Human geometry may originate from the interaction between ancient, navigation-like mental processes that approximate Euclidean geometry and our human capacity for natural language.
I won't let them take it from me.
Children exhibit visual understanding from limited experience, orders of magnitude less than our best models.
We introduce the Zero-shot World Model (ZWM). Trained on a single child's visual experience, BabyZWM rapidly generates competence across diverse benchmarks with no task-specific training. 🧵
Super excited to share this preprint! How do we disentangle underlying structure from the particular features of a learning episode to benefit future learning? We find that memory reactivation during sleep promotes this structure abstraction process.
www.biorxiv.org/content/10.6...
My new paper is out, co-led with Marge Maallo!
Using high-resolution T1 MRI from pediatric epilepsy surgery patients, we see differences in morphometrics of the preserved hemisphere relative to controls following LEFT but NOT RIGHT surgery.
direct.mit.edu/imag/article...
🧵⬇️ 1/7
Photo of an eye by Beel coor on Unsplash: https://unsplash.com/photos/brown-and-black-eye-illustration-1AIHIjtuNCI
🚨Preprint Alert🚨and Thread 🧵
The Visual Imagery Visually Anchored Scale (VIVAS) reveals dissociable perceptual dimensions and category-specific structure: osf.io/preprints/ps...
Authors: @heidasigurdar.bsky.social, Árnason, Mäekalle, Vésteinsdóttir, @arnig.bsky.social
1/9
Happy to see this paper officially published in Cognitive Psychology! A nice collaboration with Yuxuan Zeng and David Osher at OSU. We find face-like holistic processing in non-face stimuli - and Configuration is the answer!
doi.org/10.1016/j.co...
Join us on April 14th at 11am ET / 5pm CEST for a FIT’NG + ICIS preview of the Panama 2026 meetings: exclusive program sneak peeks, highlighting events of shared interest, and what you gain by attending both.
Register: tinyurl.com/fitngicis
Should AI have a childhood?
Had a great conversation with Chris over at Rewild Magazine on how developmental principles can help address a range of challenges with current AI models
fenwick.media/rewild/magaz...
I'm also doing something I haven't really done before, which is give a public facing talk on LinkedIn Live (??) about the importance of neuroscience and infant development for AI modeling
If you're curious, talk is on April 8 at 1pm EST.
Deets here: fenwick.media/rewild/talk/...
Should AI have a childhood?
Had a great conversation with Chris over at Rewild Magazine on how developmental principles can help address a range of challenges with current AI models
fenwick.media/rewild/magaz...
New paper with @cantlonlab.bsky.social out now in Dev Cog Neuro! We scanned 3-5yr olds with fMRI and found that number words activate regions of cortex also involved in visual numerosity perception, even at the earliest stages of counting acquisition doi.org/10.1016/j.dc...
In general, I'm pretty aligned with the Scott/Arcaro perspective.
But see also this post from @nmwilkinson.bsky.social which is pretty revealing
bsky.app/profile/nmwi...
A 2nd possibility is that the experience builds upon existing domain-general top-heavy/upper visual field biases we find even in species that do not rely on face perception, like rodents, wasps, and turtles.
This would explain the fetal result. But, frankly, I'm pretty skeptical of that work
Great question. Definitely a lot of work to be done, but here's my take:
One possibility is that this account may emphasize faces as a stimulus in order of similarity to mom. So mom face > any face > non-face object; which would account for the schematic face preferences
Good question. Unfortunately they did not. Presumably the non-speaking mothers still moved in other ways (e.g., smiling etc), but as I mention in my article there certainly many reasons this study needs to be replicated.
Oh this looks great. Thank you for sharing.
The broader point is that the environment is richly structured to support learning, even in the womb.
This structure may bootstrap many early abilities, such that even purely visual capacities might have origins that begin prior to the onset of visual experience
Read it here! rdcu.be/e9vXt
This cross-modal match, between a mother's voice and face, may serve to immediately emphasize faces as an important stimulus, potentially driving many of the early face biases we observe in newborn infants.
The interpretation of this result is that, by birth, infants have already had many months of experience with their mother’s voice. Thus, once born, infants simply attend to the source of that familiar sound: the mother’s face
Sai found that only infants whose mothers spoke to them before testing were able to recognize their mother's face. The infants who were not spoken to were at chance.
But, could this extremely early sensitivity to faces be explained via an experiential account? Sai (2005) provides key evidence that it can.
They tested face recognition in newborn infants, but, critically, asked one group of mothers not to speak to their infants between birth and testing.
Within hours of birth infants show sensitivity to face-like configurations and even the capacity to recognize their mother's face. This early sensitivity has long been taken as evidence that face detection is innate—fueling debates from nature-nurture to the modularity of the mind.
My (very) short piece on how prenatal experience with the mother's voice may rapidly scaffold the development of face perception in newborn infants is now out in @natrevpsychol.nature.com !
www.nature.com/articles/s44...
This is also problematic from a budgeting perspective, especially in the current funding environment.
If you budgeted a certain amount for recruitment, and that amount doubles without warning it can very much jeopardize the viability of that project.
We're experiencing this too. Yesterday afternoon we collected data on @joinprolific.bsky.social and only had drop ~2 subs for failing our performance checks. About our normal amount.
This morning almost half of participants failed. Huge drop in data quality overnight.
Honored to have been selected as a rising star among so many other great scientists!
Congrats!
