Topic: Weak / Narrow Blockchain AI (Read 5669 times)

Yeah! Applied for slack and following the forum

Here we go https://bitcointalksearch.org/topic/ann-codex-multi-wallet-custom-pos-spv-assets-decentralized-finance-1812410

,,

Great Idea, the very concept i am working on is that the initializing data of every AI will be different, much like we all generate different addresses. every user has a local  AI that crunches data and makes recommendations, and yes AI all evolve but slight deference is given to AI that are exceedingly accurate in their predictions and those lagging behind re-intialize with new data and past checkpoints. This allows the system to evolve as a whole while not being centralized, and not reliant on "tiers" as i had envisioned previously.

One of the biggest problems people who want AI and blockchain to work together don't realize is that they are different technologies. A blockchain is sequential, Ai is not... and BOOM...the idea dies a sad death. We then get to your second paragraph... after some deep thought.

If you follow my posts you will see the evolution of my idea into exactly what you describe and more. The AI will make comprehensive analysis and make local recommendations that benefit the user while submitting network recommendations that help in the overall network policy.

there is indeed a major difference between a normal blockchain and an "intelligent" one. A normal blockchain works on  "hard-coded " parameters, whereas an intelligent chain actually modifies it's parameters to fit the environment based on input variables. An intelligent chain is an open system whereas the typical chains are closed.

I was thinking it would be interesting to make a virtual world where AIs can compete for bitcoins or other coins, maybe even trade them perhaps. The AIs that are most successful can evolve onto the next generation. What has there been done in regards to virtual world AIs? Probably has been discussed a lot in the past, but with the blockchain based digital currency there are new opportunities in regard to persistence and rewards.

What about something more of a narrow AI, as in the title of this thread, and more applicable to the digital currency sphere? How can narrow AI be used to, for instance, evaluate the integrity of a coin network, or perhaps make a better metric than the market cap to measure the amount of money that the network contains? The market cap can be really volatile, and doesn't apply if there is not enough liquidity to cover a large order that wants to enter or exit a currency. Can narrow AI be used to detect fake volume trading on exchanges? And is this term really any different from just any code that is already being used in the software to run the blockchains? (Since any logic and mathematical operations are the building blocks of some type of intelligence that can be referred to as "narrow AI")

Yes, that is true , but the more we immerse ourselves in studying evolution patterns the more concerned i am that perhaps these approaches may not lead to the results we want. We are Humans.. trying to apply human/biological concepts to the evolution of a computer. i know that people like to say our brains are biological computers, but i think it is an over-simplification. No two people are alike even if they are put through the same env.

Some researchers i chatted with likened a Ai's progression to that of a child and the vehemently tried to justify that, but my belief is that a 9 month old baby already far outpaces this. The human mind is processing different kinds of different types of different levels of data input all at once. There is not just one temp gauge , literally every part of our body is a temp sensor, wind vane, pressure gauge, Our ears are listening to hundreds of sounds and filter out noise just to leave us with what we want. We can talk about our eyes, but that is obvious, what bout the less obvious things... you can tell there is rain coming by complex combinations of factors , like smell, air quality, reduction in animal noise etc. Then we have the autonomous functions which probably outnumber conscious functions on magnitudes of thousands

While our minds generally "ignore" a lot of input, the right kind of stimulus can bump anything to the top and it comes into focus, intuition is also part of the human decision making process. Let's not forget risk taking and wrecklessness.

I am strongly leaning towards the survival of the fittest school of thought. I think that each AI that out performs others and assimilates them has far better potential to make it closer to a GAI than these current iterations.

I think trial and error would be the way to go, because it shows that the system will be learning from its previous mistakes, unlike survival of the fittest were the fittest might not be the best. Thats my take on it i guess!!

I don't think these two are mutually exclusive... in biological forms of life, an organism learns in real time through trial and error (learning to walk for example), while the evolutionary mechanism is used to rebuild the whole system, changing any aspect of the architecture to adapt to macro level trends.

Very deep conversation with fellow AI enthusiasts and we left it at a question i thought i'd share here.

Should AI develop through trial and error...or survival of the fittest ?

Meanwhile on dev island...




So, small changes:-

1) The one wallet idea....not so hot. Let's use multiple wallets, although asking for a backup will give a single backup of all wallets.

2) The Multi-chain client currently supports 3 chains, with code lying around for a fourth (DASH)

3) Next stage (RPC Refactor), will come after i add something that seems may be needed if we want to use the same client on mobile SPV.

4) So, once SPV and RPC are done, it will be the GUI, i just completed the Wallet refactor.


The Oracle decentralized exchange can potentially work better and ensure more security and privacy because each oracle user is potentially running a full node. Ie for every supported currency.

Immediately after completion of the decentralized exchange, i will add interfaces for centralized exchanges, maybe two or three...

Once all this is stable, I go into AI development full swing since most of the material will be done and there will potentially be adaquate data with which to train the AI and for it to analyze process and operate on.

I am going to need help with launching first the plain oracle wallet, then the multi-chain wallet, anyone who has the time, please PM me.

One thing i have just realized is that this possibly could be a way to create incentive for hosting Bitcoin and Litecoin full nodes.

Unfortunately the flurry of work around the multi-wallet has left work on the AI dead in the water for now. Since the chains are syncing, i have to mod the wallet and GUI to show these additional chains and allow users to transact over them.

The biggest hurdles are

1) RPC refactoring
2) Wallet Refactor
3) GUI refactor

Other big ticket items on the books are

1) Decentralized Exchange (AT)
2) Integration of custom trading software. (for centralized exchanges)
3) Fresh start on the AI front.

So, henceforth i am now preparing to launch this project. I will open a thread in the ANN section and post the link here. The rollout plan will be detailed in the thread.

Hi

I have been considering this as well and have come up with the following :-

Desktop GUI version will run litecoin and bitcoin in pruned mode unless user specifies otherwise.

Android version will run all coins in SPV mode (not sure how i'm going to pull off SPV staking)

Daemon/Server versions will run full nodes.

it will be fine for windows and mac, but i think android will get you some troubles, you will have to take down many options to get it to work as it have to synce several chains at the same time.
I have been said that in order to make an android wallet, you must remove all CPU intencive tasks, i guess that block verification and database indexing are some of them.
Can it actually work with pruned chains ?

I'll be following this project, sounds interesting

So the Oracle multi-wallet can now fully sync 3 chains or more. I think this will be enough for launch. I have removed the Dash integration code, it will serve two purposes in future.

1) To see if there is demand for more coins to be supported in the multi-wallet.
2) Guide to teach developers how to integrate new coins into the client

So , once i perfect this and build for windows and mac, (Maybe android)

i will start laying out how the launch will be handled.

The coin itself will be used for services in and outside the network, but the real material is in the AI, in data mining, in predictions and in modelling. The vast majority of the features are supportive of this.

Some light reading :- https://www.wired.com/2017/02/bots-crowds-bitcoin-driving-next-hedge-fund-wave/

How will be used the coins? For what?

I think it's very bad idea. Don't forget that this system failed in bitcoin. This system is good for ordinary customer but not good for miner.


Yes, but reward helps to encourage miners work. If you know, now mining (in bitcoin) can't exists without reward and commissions (electricity and mining devices very expensive for now). How do you make miners to mine without reward? Noone will want to do it and just spend their money and time.

For those asking about an ICO, i do not like to think of it as such, the term has become so abused on this forum that it no longer means or serves the purpose it is meant to.

Oracle will have a crowdfund , maybe two rounds. The funds collected will be used in development, paying devs, designers and consultants as well as maintaining a minimum node threshold to keep the network active and the AI from becoming biased

Unfortunately my current level of c++ does not allow me to make dynamic coin adding, so not in the first release. But once i master templating and we can entice better devs with funds, we will have that ability, the only limits will be your bandwidth, ram and processor power.

An alternative could be that if a user wants a coin added, they can start a campaign, raise funds and once the funds reach a certain mark, the work to add the coin will be done. this has the added advantage of only allowing coins that are actively supported. Maybe we can come up with criteria for coins that want to be added and see how that works out.

The multiwallet also has an advantage for coins that need more nodes, every client becomes part of the network.

I will respond to Pms in a short while, i just need to get this working.