New preprint. Taking spurious reconstruction as our entry point, and using the framework of C (looking convincing) and T (the target being achieved) from Ken Shirakawa's doctoral thesis, we give a formal account of how phantom evidence arises in the age of generative AI.　　arxiv.org/abs/2607.25991

Phantom Evidence: How and Why Generative AI Manufactures False Positives in Science

Four centuries ago Francis Bacon warned against the anticipations of nature, hasty generalization that wins assent on a few facts, and set against it the table of absence: checking that a property fai...

arxiv.org

“If pizza was never shown during training, brain decoding cannot reconstruct pizza.” This is a misconception. Since Miyawaki et al. (2008), reconstructing unseen stimuli has been a core requirement of reconstruction, distinct from classification or retrieval.

Moving intentions from brains to machines Opinion by Christian Beste, Heleen A. Slagter (@haslagter.bsky.social), Christian Herff (@cherff.bsky.social), Yukiyasu Kamitani (@ykamit.bsky.social), Sabrina Coninx (@sconinxphil.bsky.social), Richard van Wezel, & Christian Frings tinyurl.com/mr2ch69z

Moving intentions from brains to machines www.cell.com/trends/cogni... by @haslagter.bsky.social @ykamit.bsky.social et al.; #BCI #NeuroTech #neuroscience

Moving intentions from brains to machines

Brain–computer interface (BCI) research has achieved remarkable technical progress but remains limited in scope, typically relying on motor and visual cortex signals in limited patient populations. We...

cell.com

Advancing credibility and transparency in brain-to-image reconstruction research: Reanalysis of Koide-Majima, Nishimoto, and Majima arxiv.org/abs/2511.07960 by @ykamit.bsky.social et al. via @seelikat.bsky.social; #neuroscience #AI

Advancing credibility and transparency in brain-to-image reconstruction research: Reanalysis of Koide-Majima, Nishimoto, and Majima (Neural Networks, 2024)

A recent high-profile study by Koide-Majima et al. (2024) claimed a major advance in reconstructing visual imagery from brain activity using a novel variant of a generative AI-based method. However, o...

arxiv.org

arxiv.org/abs/2511.07960

Advancing credibility and transparency in brain-to-image reconstruction research: Reanalysis of Koide-Majima, Nishimoto, and Majima (Neural Networks, 2024)

A recent high-profile study by Koide-Majima et al. (2024) claimed a major advance in reconstructing visual imagery from brain activity using a novel variant of a generative AI-based method. However, o...

arxiv.org

Between October 13-17th we ran the second edition of Ibn Sina Neurotech School hosted and sponsored by NYUAD Center for Brain and Health and @ibroorg.bsky.social We welcomed 18 students from across the world to learn hands-on about fMRI data collection and processing

SfN 2025 satellite Symposium DATES: Thursday 13 November ~ Friday 14 November 2025 (cf. SfN2025; November 15~19) VENUE: CORTEZ HILL Room (3rd floor), MANCHESTER GRAND HYATT SAN DIEGO 1Market Place, San Diego CA 92101, USA. www.jst.go.jp/kisoken/cres...

New preprint from our lab led by Onoo-san. We propose readout representation, defining neural codes by what can be recovered, not what caused them. Inputs remain recoverable from distant features, revealing expansive, redundant codes that align neural activation with meaning arxiv.org/abs/2510.12228

New preprint from our lab, led by Otsuka-san. In brain–AI alignment, linear transformations are common—but when features ≫ samples, outputs collapse onto the training set, breaking zero-shot predictions. We mathematically show how data and model sparsity can help avoid this. arxiv.org/abs/2509.15832

Overcoming Output Dimension Collapse: How Sparsity Enables Zero-shot Brain-to-Image Reconstruction at Small Data Scales

Advances in brain-to-image reconstruction are enabling us to externalize the subjective visual experiences encoded in the brain as images. Achieving such reconstruction with limited training data requ...

arxiv.org

Our article is out in Annual Review of Vision Science: “Visual Image Reconstruction from Brain Activity via Latent Representation” We trace the path from early brain decoding to modern NeuroAI, highlight progress & pitfalls, and discuss future directions www.annualreviews.org/content/jour...

Reconstructing sounds from #fMRI data is limited by its temporal resolution. @ykamit.bsky.social &co develop a DNN-based method that aids reconstruction of perceptually accurate sound from fMRI data, offering insights into internal #auditory representations @plosbiology.org 🧪 plos.io/4fhNw1Z

Can we hear what's inside your head? 🧠→🎶 Our new paper, led by Jong-Yun Park, presents an AI-based method for reconstructing arbitrary natural sounds directly from a person's brain activity measured with fMRI. journals.plos.org/plosbiology/...

Out now! @ykamit.bsky.social and colleagues present a neural code conversion method to align brain data across individuals without shared stimuli. The approach enables accurate inter-individual brain decoding and visual image reconstruction across sites. #compneurosky www.nature.com/articles/s43...

Inter-individual and inter-site neural code conversion without shared stimuli - Nature Computational Science

A neural code conversion method is introduced using deep neural network representations to align brain data across individuals without shared stimuli. The approach enables accurate inter-individual br...

nature.com

Spurious reconstruction from brain activity https://www.sciencedirect.com/science/article/pii/S0893608025003946 by @ykamit et al.; more information in the Bluesky thread https://bsky.app/profile/kencan7749.bsky.social/post/3lri44htxek25 #BCI #NeuroTech #neuroscience "Our findings suggest that […]

Our paper is now accepted at Neural Networks! This work builds on our previous threads in X, updated with deeper analyses. We revisit brain-to-image reconstruction using NSD + diffusion models—and ask: do they really reconstruct what we perceive? Paper: doi.org/10.1016/j.ne... 🧵1/12

Our overview of visual image reconstruction from brain activity is in press at Annual Review of Vision Science. We explore its foundations, how the field evolved, where it went astray—and how to set it right arxiv.org/abs/2505.08429

Visual Image Reconstruction from Brain Activity via Latent Representation

Visual image reconstruction, the decoding of perceptual content from brain activity into images, has advanced significantly with the integration of deep neural networks (DNNs) and generative models. T...

arxiv.org

Yukiyasu Kamitani, Misato Tanaka, Ken Shirakawa Visual Image Reconstruction from Brain Activity via Latent Representation https://arxiv.org/abs/2505.08429

Updated preprint, now in press at Neural Networks arxiv.org/abs/2405.10078

Spurious reconstruction from brain activity

Advances in brain decoding, particularly visual image reconstruction, have sparked discussions about the societal implications and ethical considerations of neurotechnology. As these methods aim to re...

arxiv.org

I promised to write about my thoughts on the status of the field of neuroAI, some of the big challenges we are facing, and the approaches we are taking to address them. This is super selective on the topic of finding a good model but in my view it affects the field as a whole. Here we go. 🧵

Brain-to-AI and Brain-to-Brain Functional Alignments speakerdeck.com/ykamit/brain...

Brain-to-AI and Brain-to-Brain Functional Alignments

Presentation at a meeting of JSPS Transformative Research Area (A) "Unified theory of prediction and action" at Kyoto University (2024.11.7)

speakerdeck.com

New preprint, led by Ken Shirakawa. Using inappropriate generative AI methods and naturalistic data, seemingly realistic but spurious visual image reconstruction can be generated from brain data (even from random data). We describe it and formulate how it occurs arxiv.org/abs/2405.10078

Spurious reconstruction from brain activity

Advances in brain decoding, particularly visual image reconstruction, have sparked discussions about the societal implications and ethical considerations of neurotechnology. As these methods aim...

arxiv.org

While investigating the issues of questionable brain decoding and reconstruction (osf.io/nmfc5/#!), Misato Tanaka from my lab found that the seminal paper by Nishimoto, Naselaris, Benjamini, Yu, & Gallant (2011) appears to use highly similar stimuli across both training and test sets. ...

New preprint from our lab! Led by Wang-san, this work introduces a content-loss-based functional alignment of brain data, which does not require shared stimuli between subjects/datasets; greatly expanding the potential of data reuse arxiv.org/abs/2403.11517

Exhibition of Pierre Huyghe "Liminal" From 17 March 2024 to 24 November 2024 At Punta della Dogana, Venice, Italy www.pinaultcollection.com/palazzograss... We provided brain-decoded images and moves for some of the works

Exhibition of Mogens Jacobsen"Restruktion" at Ringsted Gallery, Denmark, where one of the works was created in collaboration with my lab kunsten.nu/artguide/cal...