[HS]

Hey All,

I bought 50 of these solar cells on sight. Theoretically, they are exactly what I need for my mechanical AI brain design.

Specs: rated open circuit voltage = 350mV (900nm, 1mW/cm2 light source). Short circuit current is 47μA.

Note, the working power, I'm guessing, would be around 8/10 of those values. I should be able to compensate by:
1. fine tuning the light source wavelength
2. MOAR POWER!!!!!
3. cooling

I need some components (memory, processor, emitter, receiver) that can work from this tiny amount of power.  So, If anyone knows of electronics like this, that would help a lot.

[HS]

OK, I found nothing. Inspired by ranch's crazy circuit I thought LETS GO LOW TECH!

So here, would this work:
-at the bottom we have light going into a solar cell, that charges the cap, that discharges on reaching the lower diode threshold voltage, that runs through transformer circle, that ups the volts and led pulses-





And many in series:

[ranch vermin]

Warms my heart.      This glow in the dark valve head robot would be a wonderful thing if u finished it.

Doesnt it need a reciever as well as a trasmitter.   so alot of the amps coming out of the solar cell, would have a wire which you then have to work out the frequency of the unshielded signal coming in,  all the cells would need a different ringer frequency, or something.

Its better the way it is but what about thinking about some home brew miniaturization for your evolving data patterns->
if u had a rizla paper - water proof it,  then you could stamp on conductor at 100 microns detail (thats 10 times thinner than a finger print line.),  so ud get,  by layering 10 papers for 100 microns thick, youd get 1.5 million plottable voxels - to put the memory of the cell in. 
if u made it 100 papers, would take ages to stamp them all, but ud get 10 times more.   ud just be able to see it with the naked eye. 

What do u mean about the red green blue - is it some kind of frequency spectrum knowledge u know?


[EDIT ->]   I just realized, its not power coming in the bottom its signal isnt it!   Is it actually getting powered by its signal?  thats a cool idea if thats what it is.

[HS]

if u had a rizla paper - water proof it,  then you could stamp on conductor at 100 microns detail (thats 10 times thinner than a finger print line.),  so ud get,  by layering 10 papers for 100 microns thick, youd get 1.5 million plottable voxels - to put the memory of the cell in. 
if u made it 100 papers, would take ages to stamp them all, but ud get 10 times more.   ud just be able to see it with the naked eye. 

Always good to have extra ideas you can go to if Something is not working, thanks.

What do u mean about the red green blue - is it some kind of frequency spectrum knowledge u know?

Originally meant to represent white light, probably gonna use red LEDs for signaling because that is the closest wave-range to max efficiency of these solar cells.

[EDIT ->]   I just realized, its not power coming in the bottom its signal isnt it!   Is it actually getting powered by its signal?  thats a cool idea if thats what it is.

Yes, the idea is (because max efficiency is 900 nm, which is infrared), I'm gonna use a diffuse 900nm light source to partially charge up the capacitors. This puts all the cells on a hair trigger, a small increase in local light levels, such as a red led flashing, will set of near cells more easily. It's gonna be easy to see the red LEDs flashing because that'll be the only light source for the human eye.

[ranch vermin]

Knowitall is back again. hehe, u may already know this, but im not sure.

Machine learning.    its not easy, and ur missing a major piece of the picture.

So, to do it, by "find superiority based on a rule through method of evolution."

Each cell, would have a current best connection,  but it has to have connections that stand for its new random attempts,  each random attempt must be scored,  once all scoring is done, we either swap to the best,  or we keep the best, if it wasnt defeated.
Thats just the easy bit, and its already complicated your system.

Then you need a rule,  and thats fairly simple too.

But the bad issue....   its a longer story.
You cannot score anything, unless u have a sensor, for something simple, (you can do that.) or a memory of a sensor with an internal model for something more complicated.  and thats the issue, that this part of the system is potentially bigger than what you have here.

[HS]

Yep, I've been avoiding that self correcting part. You've laid out the different things I need to do, and I agree, the mind should be able to evolve on its own. Otherwise it's just a random probably useless bunch of connections with nearly no chances of attaining sentience.
I'm not sure about the specifics yet. I'll probably need to connect a potentiometer to the prongs of the led so I can adjust the brightness. Then connect a computer wirelessly to the potentiometers. Then have the computer run a genetic algorithm, changing resistances until something good happens.