A box jellyfish has 24 eyes and no brain. And it still learns.
This is all of it: a 1 cm bell, 4 eye clusters, ~1,000 neurons in each.
There's no central brain anywhere. The work happens in 4 little clubs hanging off the bell, called rhopalia. Each one carries 6 eyes and a heavy crystal that keeps it looking up.
The layout:
24 eyes, 4 types
4 rhopalia, 6 eyes each
~1,000 neurons per rhopalium
1 nerve ring linking them
0 brains
Scientists put it in a tank with grey, low-contrast walls. It kept bumping into them.
After 3–5 bumps it figured out grey = wall and started turning away early.
Then they cut out a single rhopalium and tested it in a dish. It learned too.
Its upper eyes look up through the water surface at the mangrove trees to find its way home.
Learning without a brain. Box jellyfish have been doing it for over 500 million years.
Videos were never meant to stay flat.
This is a new way to experience memories in 4D —
not just watching a moment, but stepping into it.
Frames become space.
Memories become structure.
Time becomes something you can explore.
A new way to see video. A new way to remember.
https://t.co/h5ojx9dkSz
The Alien Neuron
Alien neurons are an idea I'm exploring within hybrid mind uploading. It came through our early work on biohybrid and gradual brain replacement (gradually replacing ageing brain tissue with biofabricated tissue from the lab that is surgically implanted into the patient's brain). In bio-to-bio replacement, young, healthy tissue cannot be assumed to take over from an existing piece simply because the two initially look alike. It has to survive implantation, establish connections to the host brain and join a brain that already has "a history". The appropriate starting architecture may therefore differ drastically from the configuration we want later in the final, fully integrated state.
So you design backwards, including all the transition parameters themselves. Depending on the situation, that might mean some cells are concentrated near the outside, space for guided inward growth, allowances for tissue deformation, and an architecture adapted to the surrounding vasculature and so on => a pretty big combinatorial explosion problem to solve. "Optimised for temporal integration" is how I've been putting it, meaning that whatever you put in at first is probably going to look "very alien" compared with its final, functionally integrated state.
Now, extrapolate this reasoning from biological tissue through engineered and biohybrid systems towards increasingly non-biological replacements (i.e. fully mechanistic tissue), and an "alien neuron" becomes a broader proposition: a neural replacement whose materials and geometry are chosen for the dynamics it must support during the integration process that it must undergo. Other (potentially de novo) materials may require unfamiliar architectures. In the attached video, I'm playing with conductive filaments, ionic compartments and folded dielectric surfaces as possible ways to do some of this. Change their dimensions or how they connect, and you could change the electrical coupling, ionic transport, charge storage and response times around different neuronal types. Composition, active mechanisms and energy supply would matter alongside geometry, and several interacting components might share the work of one biological neuron (this is all just a very naive and simplified view and exploration; what the final version will look like, including what kinds of materials and geometries it will use, is exactly the one-trillion-dollar question we're trying to answer at @synconeticsorg).
Btw, the connection to hybrid mind uploading is the proposed handover. The idea is to get the engineered structure working with the living circuit first and, as they interact and adapt, the circuit would then gradually rely more on the replacement and less on the original biology.