Decoding a monkey's brain — David Reich

[DwarkeshPatel]

This short was taken from my conversation with David Reich, a geneticist focused on ancient DNA and human origins. I had no idea how wild human history was before chatting with him!

If you want to know more about the research he’s discussing in this clip, Dr. Tsao published her findings in the 2020 Nature paper “A map of object space in primate inferotemporal cortex.” In 2017 her lab was able to accurately reconstruct faces by monitoring the activity from 200 neurons!


Now, many labs and companies are using deep learning to try reconstruct faces from DNA analysis, but the field has yet to have its alphafold moment. What are your thoughts on this line of research?

[Donnerwamp]

Note: The reassembled image was easily recognizable, but not a carbin copy of what the monkey saw. Kinda like building a Lego set from memory and having a few parts left at the end.

[terrifictomm]

There's a reason we feel terrified and frightened by the prospect of such a technology. Its best summed up by this quote from C.S. Lewis:

"What we call Man’s power over Nature turns out to be a power exercised by some men over other men with Nature as its instrument.”

[aleksanderwishman]

Fyi, Japanese researchers did this 15 years ago with humans, and were vaguely able to tell what the people were dreaming! 😮 Paper called: "Visual Image Reconstruction from Human Brain Activity using a Combination of Multiscale Local Image Decoders"

More from the same author(s) if you search for it too.

[astararrialt]

they did something similar with a cat and found out that not only can they see the invisible stripes all people have but they also literally see us as cats. we have cat faces to them

[GrzegorzDurda]

Monkey could tell that a blurry picture of a banana is still recognizable as a banana. Monkey smart. Keep up the good work!

[mamavswild]

This can be misused so dangerously. To society, to free will, and a number of other ways.

[bobbyv]

The neural image visualization is itself brilliant. It'd be cool to communicate with animals

[DarthHoosier]

That’s completely terrifying. That could transform into Orwellian systems so easily. You wouldn’t even need telescreens.

[Philip-hvkc]

He's seeming to have described plain old signature analysis. It's not predictive of anything, it merely identifies from an already established long recorded archive of previous results like those substance identifiers at airports that can identify say 20, 000 different substances and compounds from the spectrum analysis of X ray laser readout. If it ain't a substance previously scanned and in the archive, then it can't be identified
It's only that one specific monkey here that they have an archive for, the archive would be not so useful if used for a different signature analysis.

[carolynbrightfield]

I wonder how many times that study has been replicated and verified.

[MichellePuckett-td]

YOU ARE GOING TO FIGURE IT OUT!! KEEP UP THE GREAT WORK

[tommydennen]

Fascinating extrapolation he’s making, and he’s right, it’s not that hard to imagine that being possible.

Also, whoever’s doing your graphics is excellent. Made me chuckle 😅

[AleixSalvador]

Neural network models do not generalize well when outside the training region. Asking it to predict something to which it has no similar training data will not yield correct results.

[uncrativ]

This is a thing with people too, throughout every language we all register the same response for each thing.

[sirnonapplicable]

It's time to brush up on Linear Algebra

[the_ther_ne_]

And... how could this be used against the plebs, to the advantage of... THEM? I have no idea.

[raptorsan]

somehow we'll try to use this for porn applications

[floior]

Fascinating. The possibilities for this tech and science are broad and very cool.

[sethmathieus]

Incredible research, keep up the great work