--- title: An AI “mind-reading” tool can reconstruct what you’re... description: Researchers at the Weizmann Institute built an AI system that reconstructs images a person is looking at from fMRI brain scans with notably higher fidelity... canonical: https://daily.dev/posts/an-ai-mind-reading-tool-can-reconstruct-what-you-re-looking-at-based-on-a-brain-scan-cdja9kyen twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: An AI “mind-reading” tool can reconstruct what you’re looking at based on a brain scan | daily.dev og:description: Researchers at the Weizmann Institute built an AI system that reconstructs images a person is looking at from fMRI brain scans with notably higher fidelity... og:url: https://daily.dev/posts/an-ai-mind-reading-tool-can-reconstruct-what-you-re-looking-at-based-on-a-brain-scan-cdja9kyen og:image: https://api.daily.dev/og/posts/cdJa9KyEn.png og:image:alt: An AI “mind-reading” tool can reconstruct what you’re looking at based on a brain scan og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # An AI “mind-reading” tool can reconstruct what you’re looking at based on a brain scan **[MIT Technology Review](https://daily.dev/sources/technologyreview)** · 9 min read · 0 upvotes · 0 comments ## Summary Researchers at the Weizmann Institute built an AI system that reconstructs images a person is looking at from fMRI brain scans with notably higher fidelity than prior approaches, using a two-branch decoder (structure and content) paired with a diffusion model. A companion encoder, trained alongside the decoder, predicts brain activity from images and lets the team train on far more data than available fMRI-image pairs—about 70% of training data had no original fMRI pairing. The resulting 'universal brain encoder' needs only about an hour of calibration data on a new subject versus roughly 40 hours for prior methods. The work was presented at the Cognitive Computational Neuroscience conference. Researchers hope it could eventually help locked-in patients communicate or reveal dream content, while outside scientists and ethicists raise concerns about mental privacy, especially if similar decoding moves to EEG, which is cheaper and more pervasive than fMRI. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### How much fMRI calibration data does a new brain-decoding AI model need per person? A newer 'universal brain encoder' from Weizmann Institute researchers needs only about one hour of fMRI data per new subject, compared with roughly 40 hours required by earlier brain-decoding approaches. This reduction makes the tool far more practical for neuroscience research, where scanning time can cost $600 to $1,000 per hour. _Researchers tracking practical advances in brain-decoding AI can follow developments like this on daily.dev._ ### How do researchers train an AI model to reconstruct images from brain scans when fMRI data is limited? Researchers train an encoder that predicts brain activity from images alongside a decoder that reconstructs images from brain scans, then use the two together to generate synthetic training pairs. This lets about 70% of training data come from images never actually shown to a person in a scanner, massively expanding the effective training set beyond scarce real fMRI-image pairs. _Developers exploring data-efficient training tricks for generative models can keep up via daily.dev._ ## Similar posts on daily.dev - [Are brain waves the next unlock for physical AI?](https://daily.dev/posts/are-brain-waves-the-next-unlock-for-physical-ai--41mglzmkz) · TechCrunch · 0 upvotes · 0 comments - [Mind-reading devices can now predict preconscious thoughts: is it time to worry?](https://daily.dev/posts/mind-reading-devices-can-now-predict-preconscious-thoughts-is-it-time-to-worry--ufkjrr8c0) · Hacker News · 0 upvotes · 0 comments - [How visual learning happens in the brain](https://daily.dev/posts/how-visual-learning-happens-in-the-brain-yvbujwok8) · MIT News · 54 upvotes · 3 comments --- Tags: [#ethical-ai](https://daily.dev/tags/ethical-ai), [#diffusion-models](https://daily.dev/tags/diffusion-models), [#brain-computer-interface](https://daily.dev/tags/brain-computer-interface) [View this post on daily.dev](https://daily.dev/posts/an-ai-mind-reading-tool-can-reconstruct-what-you-re-looking-at-based-on-a-brain-scan-cdja9kyen) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"An AI “mind-reading” tool can reconstruct what you’re looking at based on a brain scan","url":"https://daily.dev/posts/an-ai-mind-reading-tool-can-reconstruct-what-you-re-looking-at-based-on-a-brain-scan-cdja9kyen","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/an-ai-mind-reading-tool-can-reconstruct-what-you-re-looking-at-based-on-a-brain-scan-cdja9kyen"},"datePublished":"2026-10-01T10:52:06.653Z","dateModified":"2026-10-01T10:52:29.456Z","description":"Researchers at the Weizmann Institute built an AI system that reconstructs images a person is looking at from fMRI brain scans with notably higher fidelity...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/f5562da84cd962e245d102f25e25031d?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/f5562da84cd962e245d102f25e25031d?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"MIT Technology Review","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"MIT Technology Review","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/a8d03889722f4f22b3986e0b395ff076","url":"https://daily.dev/sources/technologyreview"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/an-ai-mind-reading-tool-can-reconstruct-what-you-re-looking-at-based-on-a-brain-scan-cdja9kyen","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ethical-ai,diffusion-models,brain-computer-interface","timeRequired":"PT9M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"MIT Technology Review","item":"https://daily.dev/sources/technologyreview"},{"@type":"ListItem","position":3,"name":"An AI “mind-reading” tool can reconstruct what you’re looking at based on a brain scan"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/an-ai-mind-reading-tool-can-reconstruct-what-you-re-looking-at-based-on-a-brain-scan-cdja9kyen#faq","mainEntity":[{"@type":"Question","name":"How much fMRI calibration data does a new brain-decoding AI model need per person?","acceptedAnswer":{"@type":"Answer","text":"A newer 'universal brain encoder' from Weizmann Institute researchers needs only about one hour of fMRI data per new subject, compared with roughly 40 hours required by earlier brain-decoding approaches. This reduction makes the tool far more practical for neuroscience research, where scanning time can cost $600 to $1,000 per hour. Researchers tracking practical advances in brain-decoding AI can follow developments like this on daily.dev."}},{"@type":"Question","name":"How do researchers train an AI model to reconstruct images from brain scans when fMRI data is limited?","acceptedAnswer":{"@type":"Answer","text":"Researchers train an encoder that predicts brain activity from images alongside a decoder that reconstructs images from brain scans, then use the two together to generate synthetic training pairs. This lets about 70% of training data come from images never actually shown to a person in a scanner, massively expanding the effective training set beyond scarce real fMRI-image pairs. Developers exploring data-efficient training tricks for generative models can keep up via daily.dev."}}]} ```