Integration of BOINC, YaCy and Open Assistant

[Kulintsov-PI]

Hello.

There is an old decentralized search engine YaCy, which is capable of crawling various parts of the Internet on a variety of nodes.

If each node crawls a piece of the Internet according to a distributed task, and then trains a small neural model, then these models will then be able to effectively train each other.

Probably each node will create a "non-intelligent" neural network. However, after mutual learning they will be able to give reasonable answers and remember everything they crawl.
Each node does the following:

1. Crawls a part of the Internet and store a set of data, let's say 1 GB.

2. Receives a small neural model, trains it on stored data, passes it to the next node.

3. Performs benchmarks of various models with simple tests.

4. Conducts a dialogue of 2 or more models with each other.

5. Periodically audits the stored data for relevance and novelty of the data.

Large nodes implement a genetic algorithm for selecting the most successful small models and combine them into large ones.

The project manager updates the tasks, comes up with new versions and distributes them to the nodes.

I used the decentralized search engine YaCy a few years ago. It was impractical, since many of the desired results ended up on the 5th page of the search.

A decentralized search engine is the final missing element in creating a fully decentralized Internet.

Modern search engines are spoiled by SEO. Sometimes it's hard to find what you need, because there is a lot of irrelevant spam. For example, it's hard to find a blogger's video if you don't remember his name.

AI could solve this problem, but now AI is in the hands of the same corporations that are to blame for the existence of aggressive advertising and SEO.

BOINC project is very old, one of the progenitors of modern decentralized projects. The GridCoin cryptocurrency network is organized on top of it. Participants receive rewards according to the Proof-of-Research algorithm. Now this coin is forgotten, but it is unlikely that it will die, since it cannot be banned. In fact, this is the only practically useful cryptocurrency. And all the coins on the PoW algorithm are slowly dying.

If the participants of the network with a decentralized AI and the search engine receive rewards for supporting the network, then it will be possible to create an Internet free from advertising and inadequate censorship.

Thus, a free AI will appear, which will be constantly improved to resist spam, SEO and censorship.

I'm not an AI expert, but as I understand from Yannic's videos, there are combinations of RNNs and transformers that can be well parallelized.

At first, the whole system will be impractical, but it has no limits on self-improvement.

Small decentralized parts can be easily updated, specialized for narrow tasks. It will be easy to attract new open source developers. It is even possible to pay them using cryptocurrencies such as GridCoin. Maintaining a decentralized network is pretty cheap.

The main reason for this cooperation is creating of a fully decentralized internet.
The BOINC network is slow, but you can hang important simple tasks on it, such as crawling, testing and a genetic algorithm. Participants will be able to receive rewards in GridCoin.