Bitcoin Forum

I wasn't even sure where to post this.

Ive been doing some research. The human brain contains approximately 86billion neurons. There are around 1billion computers connected to the internet. If a program was devised that worked in a p2p fashion, like bitcoin, but to simulate 1000 neurons on each computer and allow them to communicate with all the other simulated neurons on the p2p network we could make a full scale simulated model of the human brain and possibly make the internet self aware. I know it sounds crazy. Depending on the amount of neurons emulated per client you would only need a small percentage of the worlds computers to run the client for it to be more powerful than a human brain.

To give an incentive for the running of the network it could be designed around a cryptocurrency. All you need is each client to simulate a few thousand neurons. A floating blockchain that is flushed after an arbitrary set of time, say 24 hours, after 24 hours the oldest data is lost off of the end of the blockchain, that allows the network to have a working memory the neurons can write directly to this working memory blockchain to be seen by all nodes on the network, this blockchain is coupled with the cryptocurrency block chain that is a long term storage of transaction data and possibly the emulated brains long term memory storage.

This could lead to the first self aware cryptocurrency.

What do you guys think?

Ready! Set! Gooooooooo! Sounds interesting.

http://en.wikipedia.org/wiki/Blue_Brain_Project

How did someone with 420 in their name come up with this?

 :P

I've read up on projects like that, the limiting factor is raw computing power. Also distributing the neurons into clusters that talk to each other would make it more like a neural network, p2p networks look a hell of a lot like neural networks. I read about one scientist that is building custom boards that each emulate a few thousand neurons and he plans to network them to make a simulated brain, the problem being that these boards are not free so its hard to get enough emulated neurons to get anywhere near a real brains level. I think the key is to make each node a little different, this creates chaos which leads to consciousness eventually emerging hopefully. The hope is once there is enough neural processing power that the emulated neurons will automatically map themselves to smaller neural networks that have specific functions in the whole. It would be like natural selection and an evolving neural network, changes that work will stick around, changes that don't work will be recycled into new changes.

The working memory block chain could be read by all nodes and neurons which means that any part of the network can talk to any other part, each emulated neuron would need an address. This is different than our brain, in our brain any single neuron cant talk to any other, just the ones its currently connected to, ofcourse the human brain can and does change these networks on the fly so our simulated brain should do the same.

I'm thinking you could have a mechanism that automatically randomizes neurons that are not in use so they get new connections to other neurons on the network. Like "if neuron(428ckeirnec9543emnc) no communication for 10 blocks connect to random neuron". Each emulated neuron would be able to communicate with multiple other neurons at the same time so once one connects to an active neuron it becomes part of its network.

The working memory block chain would be just that, it would take all communication between neurons and store it in the block, kind of like sending mail, except the mail man goes around once every minute or 30 seconds. There could be memory locations like in ram, so each neuron can talk to all other neurons by looking at a memory location in the block chain.

This wouldnt be a full speed brain by any means but time is relative so the speed of a brain is too.

bump

Because it turns out that weed doesn't make you stupid.

I actually had a similar thought. I think this might actually happen down the road. Like in 30-40 years. If you or anyone else starts working on actually discussing this, lemme know!

That's an interesting project phase. Why, I think we're working on actually discussing it now!  And here you are!

I've participated in a couple of conversations on this topic here. It never gets too far because, well, skynet.  

Also network propagation between the "neurons" and some other issues make this idea extremely far-fetched.  Not impossible (who the hell knows) just, very improbable.

This is really great thing to know about. I am also interested in this if you start something like this plz let me know.

I think the main problem would be to simulate the neurons. We still don't know 100% about how the brain functions.

We know enough about how individual neurons function. Each way a neuron interacts with other neurons could be quantified(chemical and electrical) so that our emulated neurons act the same way and can communicate the same way over a network. If done properly you could possibly even interface real brain tissue with the network.

I don't think we have to emulate the structure of the brain, just emulating the neurons and let the structure come naturally later. Baby's don't have much of anything going on in their brains right after birth, neural nets are constructed and destroyed all the time until eventually consciousness arises. This is my hope with an artificial intelligence. Giving the network the ability to change itself on the fly.

The only other thing is how do u give this emulated brain input. A consciousness cant grow without input. Like being born with none of the 5 senses. I think input would possibly have to come later and the AI would have to learn to use it. Possibly a modified client that gives the AI access to the internet or a webcam or a Roomba remote body. lol

The problem with letting it interact with the real world is how slowly it processes. Say you give it control of a Roomba and the block time is 30seconds. That's 30 seconds per cycle of the emulated neurons being able to talk to one another, to put that into perspective our brain nearly instantly communicates between neurons, I would say instant but the Einstein says that's impossible. So were talking about an emulated brain that is thousands if not millions of times slower than our brain but functions in the same fashion.

To solve this we could change the block time but this would cause more orphans, this isnt a worry as the working memory block chain doesnt necessarly pay out, the transaction block chain does. By doing it this way we could drop the working memory block chain down to 5 seconds per block.

I'm thinking you could combine how its mined, the miner does what a miner does normally but with 2 block chains simultaneously. The wallet would have the option to run the emulated neurons on the cpu so a miner could run both. This would mean anyone with a wallet is part of the emulated brain and the miners just process all of the communication between all of the wallets whether it be transaction data or neuron communication.

It already is self-aware!! We are just pawns in its game!!

Hmmm,

Cryptocoinexplorer.com is running a total of 16 block chains on its two servers.

The explorers themselves create a secondary database of the block chains, and CCE serves as a seed node both officially and unofficially for many, if not all the block chains it supports.

I am now in the process of rewriting the explorers to make them lighter and faster so I can put more block chains on each server.

Could CCE become the "spinal cord" of our new Skynet master?  ;)

All I do all day is research shit like this on the internet and chat in the trollbox on btc-e lol

Don't get me started on time travel I'll blow ur mind. lol

What happens when an unstoppable logic meets an immovable monkey?

You get monkey feces everywhere.

Please just dont destroy the menkind!

Well... The AI ive described would be very very slow. But when you think in terms of macroscopic time, like time relative to planets forming and stars being born and dying it would be very fast. lol

The same experiment could be done with a faster AI brain, taking the cryptocurrency out of the equation or letting the AI neurons talk to each other directly and unencrypted. But then someone could just take out the AI portion of the code and have just a plain wallet. The only reason I think it needs a cryptocurrency backing is its a great way to reward processing power.



What I'm talking about is something along the lines of this http://www.neurdon.com/2010/10/27/biologically-realistic-neural-models-on-gpu/ but BIGGER waaaaaayyyyyy BIGGER.

You have described the machine on the show "Person of Interest"

Sort of. To complete the picture then, aside from the simpler matter of coding, all we need is an altruistic billionaire and volunteers for the role of an unafraid ex-special-service-agent-type ninja dude.

At the end of the film Terminator 3, John Conner explains what Skynet was and what it had done.  I replaced Skynet with Bitcoin in the quote below:

I can do that too.

By the time [The Dildo] became self-aware it had spread into millions of computer servers across the planet. Ordinary computers in office buildings, dorm rooms; everywhere. It was software; in cyberspace. There was no system core; it could not be shutdown.