Different ways to get a connectome idea to function. (from scratch)

  • 5 Replies
  • 2542 Views
*

goaty

  • Trusty Member
  • ********
  • Replicant
  • *
  • 552
Im thinking about making a neural network, not for ai, but just for running a normal logic procedure, to make maybe a pong game.   but I don't want to use electricity,   and just a 10hz oscillation just to see if it would work.

But im yet to come up with my method,  ill post it here once im done working it out,   id like some advice if anyones got anything for me.

As I see it->

Logic is all about force transfer, or the other way to name it, conduction.
Conduction doesn't have to be electricity, even tho maybe the term mostly refers to it.



If you put the complete program in, all the series involved,   then you should be able to get your program to function in a single hit,  so you only need to reach the frame rate in oscillation,  computers have to oscillate really fast to just make the frame at the end,  and all a custom connectome system would need to do, is just make it to the visible fps.

Off the top of my head I see this cool mechanical pong game,   but that's only the beginning.   Ive got this crazy idea, that if you were running a neural network for ai,  it could be externally emulated with a fixed connection on the outside, that's running the ACTUAL ai network on the inside,  so its going through a level of abstraction,  instead of just representing it directly.

So any gift ideas?   And ill have my paradigm here as soon as I come up with it...
« Last Edit: April 04, 2019, 03:55:21 am by goaty »

*

AndyGoode

  • Guest
Re: Different ways to get a connectome idea to function.
« Reply #1 on: April 03, 2019, 10:38:24 pm »
It's not clear to me what you're trying to do.
Do you mean traditional logic, with AND, OR, NOR, XOR, NAND, etc.? That's extremely clear-cut and can be done with a neural network if you want, but there's no advantage to using a neural network for logic, as far as I can tell, especially since you'll be throwing away the neural network's most useful features, like fuzzy matches and learning ability.
A 10 Hz oscillation of what? Sound? Electricity?
I could probably give you some good ideas, but your goal and methods aren't making sense to me.

*

HS

  • Trusty Member
  • **********
  • Millennium Man
  • *
  • 1175
Re: Different ways to get a connectome idea to function.
« Reply #2 on: April 04, 2019, 02:14:28 am »
My idea is vague, sorry, but I think some sort of resonance frequency could keep everything moving in sync extremely efficiently. Like water on a speaker, its an economic way to order a complex system.
« Last Edit: April 04, 2019, 03:29:29 am by Hopefully Something »

*

goaty

  • Trusty Member
  • ********
  • Replicant
  • *
  • 552
Re: Different ways to get a connectome idea to function.
« Reply #3 on: April 04, 2019, 03:36:25 am »
Hopefully Something, you were making your own neural network hardware before right?

Thanks for the comments, sorry for being a bit vague im a bit permanently dizzy in the head.
What im trying to do, is to just to build a hardware network, easiest way wins, so I can have my own computer, I made from scratch.


Im not thinking of using electricity or sound,  just good old mechanical connections... but its not certain. I actually just want what would be easiest.

I figure I should be able to put down 100 thousand brain cells (or network nodes) mechanically,  it would make in a 25cm metre block, half a centremetre per cell.

Making the "computer" or "network" mechanical, could make things easier to put it the robo-brain together from scratch.   But the actual bits and bobs im going to fab together, not sure of yet,    im not even sure of the topology im going to use,

it may end up looking more like traditional logic than a connectome, but my theory is more towards a network type idea.

Im thinking about using a mechanical oscillation like this (watch?v=lRHlrOW7UkY&t=33s) <-utube, for a "lego" neural network.
That oscillator I linked to from utube, runs on a compressed air battery,  so id be pumping its battery and it would outputting the oscillation. if you wanted to speed up the hz you could use this -> (watch?v=5_1aZW-AXTA) but that could heat up a lot if you pushed it too hard.  Strangely I don't think going for a hz multiplier is so important,  having a good count of cells could be just as good.





« Last Edit: April 04, 2019, 04:43:47 am by goaty »

*

HS

  • Trusty Member
  • **********
  • Millennium Man
  • *
  • 1175
Re: Different ways to get a connectome idea to function.
« Reply #4 on: April 04, 2019, 05:05:49 am »
Hopefully Something, you were making your own neural network hardware before right?
Still trying!

Quote
sorry for being a bit vague im a bit permanently dizzy in the head.

No i meant I was being vague.  :)  And I feel your pain, part of the reason got interested in  AI was cause I drank some apparently unsafe water while abroad and consequently my nervous system got scrambled, autoimmune or parasitic I don't know. But It took me a year to recover depth perception and coordination etc. What can you do, right? So I took it as a call to action for my AI superhero origin story.   O0

*

goaty

  • Trusty Member
  • ********
  • Replicant
  • *
  • 552
Re: Different ways to get a connectome idea to function. (from scratch)
« Reply #5 on: April 05, 2019, 05:14:04 am »
I still cant think of it...  but if I had it in KNN mode->
If u had a small key, 100 thousand cells could near match a thousand 100 bit keys,  and the amount of accesses u get a second would be the hz u hit it at.
That's not too bad, especially for a home made job,  and that's what its looking like for realistic use,  its a hard budget you have to hit for doing a.i. with.

Performance in the toilet is pretty much what everyone has to deal with.   Those deep mind go killers probably spend an hour just to set one synapse weight, the going rate for the cutten edge, SLLLLLLLLLLOOOOOOOOOOOOOOWWWWWWWWWWWWWWWWWW.....

 


OpenAI Speech-to-Speech Reasoning Demo
by ivan.moony (AI News )
March 28, 2024, 01:31:53 pm
Say good-bye to GPUs...
by MikeB (AI News )
March 23, 2024, 09:23:52 am
Google Bard report
by ivan.moony (AI News )
February 14, 2024, 04:42:23 pm
Elon Musk's xAI Grok Chatbot
by MikeB (AI News )
December 11, 2023, 06:26:33 am
Nvidia Hype
by 8pla.net (AI News )
December 06, 2023, 10:04:52 pm
How will the OpenAI CEO being Fired affect ChatGPT?
by 8pla.net (AI News )
December 06, 2023, 09:54:25 pm
Independent AI sovereignties
by WriterOfMinds (AI News )
November 08, 2023, 04:51:21 am
LLaMA2 Meta's chatbot released
by 8pla.net (AI News )
October 18, 2023, 11:41:21 pm

Users Online

377 Guests, 0 Users

Most Online Today: 396. Most Online Ever: 2369 (November 21, 2020, 04:08:13 pm)

Articles