Conversation

the latest in brain decoding. someone looks at the pic on the top, and the fMRI readings are used to generate the pic on the bottom. need to feed some dreams into this 🙏 mind-vis.github.io

Two rows of six images showing a giraffe, insect, baseball player, firetruck, building, and bathroom interior. First row text reads "Visual Stimulus" second row "Our Reconstructed Images".

read image description

ALT

7:12 AM · Nov 17, 2022

@ChuckBaggett Chuck Baggett

@ChuckBaggett

Nov 17, 2022

Replying to

@kcimc

If true, or even partially true with big qualifiers, this is a huge story.

Replying to

woah what did i just see

Replying to

and

This is what my dreams feed would look like

The following media includes potentially sensitive content. Change settings

View

Replying to

Wait, what

Replying to

is this real?

Replying to

I'm still trying to understand what they did here exactly (good morning!), but sounds like without much annotated training data they can now translate fMRI recordings into images (yes

@kcimc

?) I guess it makes sense considering Transformers abilities to translate. Wow.

Replying to

and

But your eyes aren't transmitting image data when you dream.

Replying to

and

I wouldn’t be surprised to see the first conviction resulting from the use of an AI-based Lie Detector test to happen within a decade, probably earlier in some countries.

Replying to

Curious if this improvement comes significantly from better interpretation of scanning, or simply just better image production with the diffusion revolution (e.g. vs biorxiv.org/content/biorxi )

Replying to

Insane