[LOCKSUIT]

You want to build a brain ? AGI?

Like korrelan said, absolutely don't use real gel neurons...you can't control/access all the data bits like you can in a computer simulation/algorithm of the nodes/connections. And we already have net learning algorithms that learn/prune useful/less connections/weights on the connections/etc. The AI field is astronomically advanced/amazing/wide/deep but I'm a little farther. Korrelan is a neuro nut btw, he decides to simulate the whole brain and there realistic many-connections.

Secondly, I think I read you said something like "and then it'll get going and acquire this and learn this etc...", it's easy to say that but harder than that...there's a lot to human intelligence/AGI, it is sorta simple, but to grasp those simple parts takes a little understanding...

First off, yes rewarding actions & it'll do dis and dis & dis !! is sorta true, BUT, and this is where I failed myself.....it needs to learn facts!! Like "water is wet", "gravity pulls objects"......without a visual/auditory/text sensory knowledge, it can't learn any SOLUTION to ex. CANCER / STORAGE DEVICES etc, it needs to **discover new actions/facts/tools, to CHANGE the Earth! That's what we did to get this far! We pass down what we discovered/got, then it builds on those facts(knowledge)/actions(skills)/tools. The whole process is acquire and discover.

[ranch vermin]

No his idea is good,  the neuron soup transmitting to itself is an excellent idea, i dare you to put all the wires in,  instead,  that is what would be impossible.

You can think of a neural network in this simple way,   that your not programming the robot exactly,  your programming it to maximize a score, due to itself and its environment.

But its a program the same,  its not wholey self developing, of course! or it would be bulldust.

You can give it all types of programs tho,   you can make it analyze itself,  and reorganize itself,  as well as generating motor - which is the final product of the system.

Viewing them as a search engine, is another way to look at them,  but they never arise at the exact result,  but at the closest one it could find.

[HS]

You want to build a brain ? AGI?

Like korrelan said, absolutely don't use real gel neurons...you can't control/access all the data bits like you can in a computer simulation/algorithm of the nodes/connections. And we already have net learning algorithms that learn/prune useful/less connections/weights on the connections/etc.

Yeah, I'm going for an AGI. Seems less useful than a specific intelligence that's really good at Pokemon Go, but a broad intelligence still seems cooler. Nooo! The gel is the best part! Then you can have mili-bots or micro-bots (what's this obsession with nano?) slowly winding their way around, pinging neurons and replacing them if they stop responding. Besides , I don't like the idea of calculating a thing with a computer which can be effortlessly done by physics (it's bonkers , massively parallel and no lag or anything!)  With a PC even with multiple threads/cores it's still an hourglass and you have to push to get that information through. Besides besides, simulations are too orderly and clean for cool accidents to happen. Takes away potentially useful ingredients from the primordial soup.

[LOCKSUIT]

Hmmmmm......

Ok

How do you save/teach in-to your gel a school fact like "AI is powerful" or "everyone needs education" ? What sensor and what memory storage device will you use ??