---
title: 'Mind Reading: The Science of Decoding Thoughts and Dreams'
source: 'https://youtube.com/watch?v=AgbeGFYluEA'
video_id: 'AgbeGFYluEA'
date: 2026-07-19
duration_sec: 1342
channel: 'Vsauce'
---

# Mind Reading: The Science of Decoding Thoughts and Dreams

> Source: [Mind Reading: The Science of Decoding Thoughts and Dreams](https://youtube.com/watch?v=AgbeGFYluEA)

## Summary

This video explores the scientific field of thought identification, where researchers use neuroimaging and machine learning to reconstruct visual memories and even dreams from brain activity. The host participates in experiments at the University of Oregon and Kyoto University, demonstrating how fMRI and EEG can decode mental images, raising profound questions about the nature of thoughts and the ethical implications of mind-reading technology.

### Key Points

- **Introduction to Thought Identification** [00:05] — Mind reading is framed as a scientific endeavor using neuroimaging and machine learning to understand what thoughts are made of, not just spying.
- **fMRI and Memory Patterns** [01:31] — Brice Kuhl's lab uses fMRI to detect patterns of brain activity when memories are formed and reinstated, allowing them to deduce what is being remembered.
- **Experiment Setup: Memorizing Celebrities** [02:32] — The host memorizes 12 celebrity faces, then tries to vividly recall them while in an fMRI scanner, with the goal of reconstructing the images from brain activity.
- **How fMRI Reconstructs Faces** [04:08] — The brain is divided into voxels; fMRI detects blood flow changes indicating activity. Different voxels respond to facial features like mustaches, enabling reconstruction.
- **Second Phase: Viewing Faces for Training** [05:44] — The host views over 400 faces in the fMRI to train the algorithm on how his brain reacts to different facial features, which will later decode his memories.
- **Results: Reconstructed Memories** [07:20] — The team shows reconstructions from the host's memory of John Cho and Megan Fox. The John Cho image captured face shape and expression, but Megan Fox was less accurate.
- **Self-Image Reconstruction** [09:13] — A reconstruction of the host remembering his own face showed some features but lacked the beard, highlighting limitations in memory fidelity and technology.
- **Improved Results from Viewing Phase** [10:37] — Reconstructions from the viewing phase (when the host actually saw faces) were much closer to the originals, showing the algorithm works better with direct perception.
- **Dream Decoding with EEG** [12:14] — Dr. Yukiyasu Kamitani at Kyoto University uses EEG to detect when a person is dreaming and then predicts dream content across 20 categories, such as 'character' (written language).
- **First Photographs of Dreams** [15:48] — Kamitani's lab has created early reconstructions of dream imagery, representing the first 'photographs' of dreams, though accuracy is still limited.
- **Ethical Implications** [17:30] — Julia from Fathom Computing discusses ethical lines: technology can empower, but policies and laws must guide its use to prevent misuse while enabling progress.
- **Applications for Communication** [18:51] — Mind reading could help locked-in patients communicate, allowing them to share thoughts that are otherwise inaccessible.
- **Humans and Technology Co-evolution** [20:17] — Julia argues that humans are already inseparable from technology; ethical concerns should be addressed proactively, similar to how traffic laws evolved with cars.

### Conclusion

Mind-reading technology is advancing rapidly, enabling reconstruction of memories and dreams from brain activity. While ethical concerns are significant, proactive collaboration between technologists and policymakers can ensure these tools are used for good, such as helping locked-in patients communicate.

## Transcript

I love reading. Look, mind reading might sound like pseudoscientific-- But its scientific counterpart, thought identification,
It's based in neuroimaging and machine learning, aren't just about spying on what someone is thinking.
They're about figuring out what thoughts are even made of. what does that mental picture actually look like? How high fidelity is a memory,
Well, in this episode, My journey begins right here at the University of Oregon. He's a neuroscientist who uses neuroimaging
without them telling him.
Well, I'm in the cognitive neuroscience program here, My lab primarily uses neuroimaging methods, or fMRI.
And how do you use fMRI to investigate memories? When you form a memory, there's a certain pattern. and then test whether that pattern is reinstated
Does that mean we can look at the patterns of brain activity and deduce what it is that is being remembered, or recalled, So it basically takes your input pattern
while you're remembering something. You can see how this sounds like mind reading. So, Brice, what are you going to do to me today?
is uncharted territory for us. of the experiment on you. But it represents where the field is
Today, you're going to participate in an experiment 12 pictures of celebrities. And you're going to try to remember those pictures.
Try to bring that picture to mind as vividly as possible. as you try to imagine these pictures. Essentially draw a picture of what you're remembering.
An actual picture that we can print out [Michael] The first step is for me to memorize Brice will later try to detect me thinking about.
I sat down to do this graduate student, Max. on my ability to recall these faces as vividly as possible while inside the fMRI.
I think I have a pretty good memory of all of those. With the celebrity faces hopefully memorized,
going through the metal detector where Brice will record and monitor my brain activity, and then later feed it into his algorithm to rebuild the faces.
to reconstruct faces from long-term memory, on how clearly I can remember the celebrity photos I love its eyes. Look at that.
Wouldn't the kid be like, "It's going to eat me"? by dividing it up into thousands of small cubes
Each of these voxels contains Using fMRI, we are able to detect which means that that part of the brain is active.
my brain will react to the features for each face. that is engaged throughout. That may be the area of my brain that reacts to mustaches.
if Brice notices that area is engaged, about a mustache. and he's seeing words appear on the screen one at a time,
remember the face in as much detail as possible. We get one of these brain volumes every two seconds. So these are refreshing in real time as we collect the images.
it's time for part two, where Brice and his team so they can later decode by brain scans.
[Michael] Yup. and record how my brain reacts They will then use this information
I thought about during the first phase of the scan. So we're going to basically keep him in there we could get in the fMRI.
But I was able to look at over 400 faces, Hey, Michael, you did it. That was great. [Michael] All right.
Some images of your brain. Max is going to analyze your data. where we try to actually reconstruct the face images
All right. Well, see you tomorrow. You better pull an all-nighter. I want this data to be perfect.
Overnight, his team crunched the data, and I can't wait to see what they think they saw me thinking. We're going to take a look in just a moment here.
-So can I just take a seat? -Yeah, have a seat. first of all... these are the pictures I actually memorized.
you've reconstructed from my imagination. [Brice] Okay, so this is one of the reconstructions [Max] So that's John Cho.
-Can we see the side by side? -Yeah. in the kind of facial expressions in general.
The shape of the face I also thought was-- -Yes. Yes. -So those are the things this image of John Cho,
I just kept thinking, he was the square guy. Excellent, all right. [Michael] Mm-hmm.
[Brice] You can see the picture you actually saw, I'll you this. Megan Fox, I was not able For some reason, this image of her was really hard for me
The sternness in the face was something that I did pick up on. And you picked up on the sternness. [Michael] Keep in mind that Brice and his team
But when I remember a face, with photographic accuracy? By reading my mind, they may be seeing
-Me! Me! -[Brice laughs] of me thinking about this image of myself. Where'd the beard go?
[Michael] For instance, this is a picture of me remembering my own face. how good am I at picturing myself?
so the strangeness in the result and mental picture of myself as flaws in the technology. [Michael] That's Jennifer Lawrence?
It looks like it's Jennifer Lawrence's much older uncle. Nothing here was too mind-blowingly close. But this is something that you're just starting out trying
What Brice and his team read in my mind might have been more accurate if they'd shown me thousands because then the algorithm would have learned
But regardless, the quality of my memories is cut out of the equation entirely. when I was looking at faces in the fMRI.
And those results were much closer Okay, so, what am I looking at right here? in the top row, these are images that you saw
Below that, in this bottom row, these are the reconstructions that we draw from the patterns of brain activity we collected. [Michael] These are from my brain.
Yeah, overall they were pretty close. These are-- you can see there's some variability in these. that the reconstructions that we generated,
there is some correspondence between the actual face. -when you're viewing them. -Right, right. Well, Brice, Max, thank you so much
Thank you. It's been a lot of fun. It's always useful for us to think about these things.
Dr. Brice Kuhl's memory research is showing that it's possible To figure out what they're thinking. I mean, if you want to know
it's still easier to just ask me to tell you. Dr. Yukiyasu Kamitani is a researcher, professor and pioneer exploring the frontier
I've come here to Kyoto University to read not what someone is thinking, but what someone is dreaming.
-Hi, I'm Yuki. -Yuki, nice to meet you. Dr. Kamitani has been at the forefront The subject is, you know, ready to go in.
his early experiments explored reconstructing images shown to subjects in an fMRI based on their brain activity. and the reconstructions were strikingly accurate.
Recently, Kamitani has focused on using deep neural networks while they view much more complex photographs. What you're seeing is the result of a deep neural network
looking at the photograph. for example, in criminal investigations and interpersonal communication.
But I think you still see some, you know, eyes and, you know... [Kamitani] Yeah, to some extent, yeah.
He's attempting something extremely ambitious: Would you call yourself a sleep researcher, Maybe a brain decoder.
That's a pretty cool job description. Can you show me anything from what you're doing with dreams? [Kamitani] Mm-hmm, yeah.
begins with a similar process to Dr. Kuhl's: while they are in an fMRI when it is thinking of certain things.
at identifying what images the subject is thinking about, with an EEG cap on their head, When the EEG waves indicate that the person is dreaming,
the subject is most likely dreaming about. Right now, the algorithm looks for 20 categories. and characters in a language.
ask them what they were dreaming about, and the person's recollection match. Here is actual data from one of Kamitani's experiments.
The name of each category get bigger or smaller that they are present in the subject's current dream. for the category "character," meaning written language.
and this is what they reported. -[laughs] -Right? I mean, you--
Yeah, in a way. But... Well, the accuracy's not that great but, you know, Right.
into predicting the content of dreams, actually reconstructing images from our dreams. that your lab has created...
...of dreams.
[Kamitani] Yeah. appreciate that what we're looking at on this screen are, in a way, some of the first photographs of a dream.
We are looking at the earliest phase One day, we may able to have images, And Dr. Kamitani is the only person in the world
He's a lone explorer journeying into our subconscious. No. -Thank you for showing it to me. -[laughs]
The insights that researchers like Dr. Kuhl and Dr. Kamitani because of mind reading But let's slow down for a second,
that can know us better than we know ourselves. Well, to address that question, neuroscience and artificial intelligence:
She's the director of strategy at Fathom Computing, an alum of Ray Kurzweil's Singularity University, a think tank specializing in future technologies
and their impacts. -Yeah, of course. -You are the perfect person But I think they're extremely important,
I think we're living in such an interesting time right now, So when it comes to being able
where are the ethical lines here? Like with any powerful technology, All these new technologies
are things that can make whoever uses them more powerful. So we want to not blame the technology, but we want to-- So how do we make sure that this technology
So I think it's very important to involve people who act on policy and law I am hopeful about the collaborative aspect of it.
I mean, what are the applications here? if he had a way of richer interfacing with the world
what he could have shared with us. They are there. They know that they are there. to see what it is that they are trying to say,
So, what do you say to people that have that kind of fear of technology, of us surrendering our true natural selves to technology?
There is something enticing about getting to the next level of what some people might call a human evolution In a way, we are already not living natural lives, right?
I don't know, 30 or 40. We would not wear this clothing. We wouldn't have antibiotics.
[Julia] We are already kind of to the human that was living 10,000 years ago [Michael] Yeah, we really are.
right now we basically have to either just ask people or observe their behavior. But reading thoughts directly would be a lot better.
and it's how Dr. Kamitani is studying sleep and dreams. it's easy to see how ethical questions
Well, here's the thing: there is no such thing as a totally wild human. We are co-evolving with technology.
Humans and technology today are inseparable. but we cannot change the fact that they will happen.
You know, we could have sat around forever and who should have the authority to enforce it. Instead, we went ahead and invented cars,
and responsibly figured out the details as we went along. do the most good when they facilitate the technology, not when they needlessly hinder progress.
And, as soon as you can, show them to me. And, as always, thanks for watching.