Brain decodingBrain decoding
All In The Mind
Imagine if it were possible to read people’s thoughts by detecting their brain activity with a brain decoder. We hear from the researchers who can already tell something about what you’re watching or listening to.
28:50•10 Jul 2016
Brain Decoding: From Movie Scenes to the Voice in Your Head
Episode Overview
- Gallant’s lab builds detailed encoding models that show how different brain areas represent visual and language information, and these can be flipped into decoders.
- By pairing fMRI data with thousands of hours of random YouTube videos, the team can roughly reconstruct unseen film clips from recorded brain activity.
- Natural, engaging stories from The Moth Radio Hour reveal that meaning is spread across large, overlapping networks in both hemispheres, rather than isolated spots.
- Data-driven, story-based experiments can reveal richer patterns than traditional fMRI designs that compare narrow categories like tool words versus animal words.
- Future applications could include brain–computer interfaces for people who cannot speak, but also raise serious ethical questions about privacy and decoding internal speech.
“"To my mind, the most interesting decoder that we can create in the immediate future is something that would decode internal speech, the little man or woman in your head that talks to you all the time."”
Curious about how others turn brain science into something that sounds a lot like mind reading? This episode of All In The Mind looks at how researchers are learning to "decode" what people see and hear just by watching their brain activity. Host Lynne Malcolm chats with Professor Jack Gallant and neuroscientist Alex Huth from the University of California, Berkeley, whose lab accidentally built powerful brain decoders while trying to do something else: understand how the brain represents information.
As Gallant puts it, their decoding demos are "really just a proof of how good our models of the brain are." You’ll hear how volunteers lie in an fMRI scanner watching hours of film clips, while computers link tiny patterns in blood flow to what’s happening on screen. With thousands of hours of random YouTube clips as a kind of visual dictionary, the team can then reconstruct rough versions of unseen movie scenes from brain data alone.
The conversation then shifts from vision to language. Using engaging stories from The Moth Radio Hour, Huth explains how meaning lights up the brain. Instead of one word in one spot, the team finds a "many-to-many" pattern: concepts like numbers or social relationships fire up widespread, overlapping networks across both hemispheres. The episode balances geeky detail with very human questions. Could these methods one day help someone who’s lost the power of speech communicate again?
Could internal speech – that constant little voice in your head – be decoded by a future portable headset? And what happens to privacy if "your innermost thoughts are your most private information" but become technically readable? If you’re fascinated by how science moves from basic research to unsettling possibilities, this one might leave you asking: how comfortable are you with someone else peeking into your mind?
Do you want to link to this podcast?
Get the buttons here!
More From This Show
The latest episodes from the same podcast.