Україна 
info@agroglorytime.io
AGTIntellegence © network – The concept of distributed hybrid˟ intelligence
Copyright P.Drobyshev, D.Novhorodkina
AGTIntellegence © network – the concept of distributed hybrid˟ intelligence
(˟hybrid intelligence is understood here as the combined workflow of a human and his personal gadget)
This concept is designed to build and develop a system capable of generating monetary value from distributed computing by solving a broad class of problems. The development perspective is to create some kind of “megabrain” with low-limit capabilities.
Back in 2011, not being familiar with blockchain technology, the authors started to think (the idea was in the air then) about a system of virtual value creation through distributed computing, generation of a new type of virtual money and the resulting payment system. The second and third questions had already been ingeniously solved by S. Nakamoto in 2009, although the world made this decision closer to 2013. The issue of value generation is still in its infancy. Yes, there are games and other applications whose tokens people are willing to buy because they are useful. However, distributed computing, in the form of blockchain, has nothing to do with the value of tokens that are generated and maintained by applications (exempli gratia games, booking services, AI, etc.). And globally, for example, bitcoin and many other coins are essentially just distributed dynamic tables. Since the launch of smartcontracts by V. Buterin, these tables have started to be managed by decentralized programs. New methods of anonymization, confidentiality, and security have emerged. But the tables remained tables. Here we try to give a conceptualization of the idea of value generation by the network (in a broader sense than blockchain allows) in the real world, namely the value of the ideas, knowledge and solutions created.
Our concept is based on factors such as:
– sufficiently high bandwidth of Internet protocols to perform distributed computing
– high computing power and memory capacity of modern gadgets
– Each gadget is operated by a person who wishes to receive network services or participate in network activities
AGTInw is a gadget application program that connects gadgets and people working on them via the internet, hereinafter referred to as the application.
AGTIntelligence network (hereinafter – the Network) – a subnetwork on the Internet that connects people and gadgets with the AGTInw application installed in order to receive or provide work on the Network.
AGTIneuron (Agent, Neuron) – association in the work in the Network of a person (hereinafter referred to as the Owner) working in the Network on the AGTInw application with his Gadget.
AGTIgadget (AGTIgadget, Gadget) – a gadget connected to the Network and operated by the Owner
SubnetAGTI – An association of Agents that perform a specific task.
AGTIsynapse – A connection between two Agents that has its own weight, and possibly its own operating program.
The structure of the application is aimed at helping the Agent to participate in the Network by creating AGTI neural structures, performing computations on the gadget using human reasoning, evaluations, goal-setting, working with the Internet, Agents and just other people, and thus solve problems and put them to the Network by making or receiving payment with AGTI tokens.
Combining the computer power of a gadget, human intelligence with inherent intuition, will and goal-setting, the power of distributed computing and a potentially huge number of Agents will make it possible to create subnetworks of agents of completely unusual classes to set and solve the most complex problems, including those of unknown types.
Thus, by installing the AGTInw application on his gadget, the agent gets an opportunity on the Web:
– Pose questions, requests and tasks to the Network, and pay for their fulfillment.
– Discuss and develop ideas and concepts.
– Store and process information including creating and participating in the development of databases and knowledge, and being paid for doing so.
– Train and configure Network structures, share information between agents.
– Evaluate answers to questions (by assigning weights or by other means) of other agents, created subnets in the Network (hereinafter referred to as Subnets) or connections within and between them in different Subnets.
– Participate in various Subnetworks, develop ideas and concepts, and participate in answering questions and inquiries.
– Participate in the management of the Subnetworks.
– Pay for Agent or Subnetwork services with AGTI token.
– Get paid for capacity, work, knowledge, solutions, etc. AGTI token.
AGTInw application architecture requirements
(hereinafter referred to as the Annex)
1. The application is adaptable to any personal gadget and computer, but can be installed on any more sophisticated computer system as well
2. The owner of this gadget (hereinafter referred to as the Owner) himself determines how much computing power and memory he devotes to the Application, when and how he can himself work on it as a constituent (second component – the Owner’s gadget) part of Neuron.
3. Participation in Subnetworks and other substructures of the Network is selected by the Application by default or by the Owner.
4. In the Annex and the Network, it should be basic very strictly stipulated that settlements are made only in AGTI, and in the future, possibly by a fork of AGTI held only by AgroGloryTime Corporation, or by its own blockchain.
Hence the proposed architecture:
1. Network Participation Unit – the computing power offered for participation in the Network, the memory of the Gadget, the possibility (in terms of time and topics of interest) for the Owner to participate in the Network – configured by the Owner or default settings are used.
2. Communications Block – means of connection with other Agents, assigning to each communication its own weight (tensor function calculated by this Gadget in a short time), participation in subnetwork operation, receiving and transmitting information, participation in assigning weights, creating and managing subnets, setting and solving tasks.
3. block creation, training, participation in sub-network creation, including through AGTI smart contracts.
4. Information and Program Storage Unit – information for use by or in the Network, programs for processing incoming and proprietary information.
5. Information Processing Unit – the Network’s own programs, imported or independently developed programs.
6. Gadget and Owner Interface Unit – means for Owner to work, ask questions and tasks, exchange information, get answers, pay and other.
7. Security block – means of protection in various ways against hacker attacks, phishing, network paralysis, looping, etc. In the future, perhaps from unmanaged activity of the Network or Subnet.
8. Message Block – messenger between Owners related to the management of Subnetworks, weights, algorithms, etc.
9. Payment Block – AGTI wallet for offering and paying for services of the Network, Subnetworks and Agents, accepting offers and rendering services to Agents, Subnetworks, Network. Maintaining current (unbilled) debit and credit liabilities. Designing smart payment contracts.
10. Design block – the network will evolve and with it the AGTInw application. In this way, the Agent can quickly suggest updates to the Application, and once tested on the Web, they can be quickly installed.
11. A standard library of programs that can be customized for each Agent.
12. the AGTInw application modification block.
First tasks for the AGTIntelligence network
- Internet search tasks
- Information about a person or company
- Airline tickets
- Real estate
- Monitoring, analysis, forecasting
- Professional translation and debugging of neural network translators
- Analytics
- Identification and classification
- Encryption and decryption
- Post-quantum cryptography and blockchain
- Distributed encrypted storage of information with pay-per-use access.
- Patenting and access to inventions, including using blockchain and NFT.
- BigData storage and processing.
- Scientific information – retrieval and processing
- Public opinion
- Forecasting complex processes
- The Art of Hybrid Intelligence
- Nonlinear, fuzzy and symbolic programming problems
- Bringing agents together to create products on the Web, constructing algorithms, Creating subnets on the Web
- Scientific challenges
- New types of forecasting
- New types of art
- New types of blockchains
- Taking Network interface functions to new levels
- Taking Network tasks to new levels
- New types of Subnetworks (fractal, cluster, etc.) with tools for their creation and training, and a new level of problem solving.
Initially, the Network itself is centralized and managed by AgroGloryTime. Subnetworks, on the other hand, are created by Agents as centralized, hierarchical and decentralized. However, as the network grows, capacities grow, structure is built, and Agents gain weights and means, the Network decentralizes to a large extent. In this case, the extent of the Agent’s leadership function will be determined by its administrative weight in the Network and the amount of AGTI in its wallet.
Different kinds of blockchains may be present in the architecture of the Web to store dynamic information that requires a distributed registry.
Understanding the novelty of hybrid intelligence and visible algorithms.
The Level 0 Agent sets the task and the planned payment for the solution, for which the Level 1 Agents are charged.
They break the problem into subtasks, partially solve it, partially pass it on to Level 2 agents, and so on until a complete solution is reached, which is passed on and formalized in descending levels, followed by payment and evaluation of each participant and each Synapse.
The concept of an Agent is much broader than the usual neuron in neural networks. Therefore, the algorithms and types of networks can be radically different and can solve a wider class of problems and at a higher level.
Agent 0 needs to search the internet for some requirements.
He proposes a request and payment for completing it. The payment algorithm offers its own, or from the Application library. The 1-N agents that have taken up a query (also the 1-N agents may be selected by default by the Network, given the ties to Agent 0 and the weights in this type of query) propose their solutions, some of which are paid for and evaluated by Agent 0.
Several agents voluntarily create a Subnetwork to solve a particular class of problems. For this purpose, sub-networks of development, evaluation and training of working AGTIneurons are formed. The payment process for network creation and operation is determined by the algorithm in the specified Subnetwork.
The agent establishes a network for monitoring and forecasting oil prices. Engages skilled agents, and uses computational power to create a subnetwork that learns to predict.
Further, network members can, for example, receive payment from people who ordered forecasts, from playing on the stock exchange, etc.
ALGORITHM DEVELOPMENT IN THE NETWORK
The currently developed neural network algorithms can easily be adapted to the AGTInw network. The hybrid and distributed computing, self-learning and self-constructing capabilities of the Web provide a much broader class of networks and problem-solving capabilities. Undoubtedly, new opportunities will gradually open up online.
The authors estimate that a full launch of the Network is possible with 10,000 Agents and a development cost of about $5 million dollars. It is the AGTI token model that makes it possible to pay developers with it, as AGTI is based on a real business and generates revenue. This way tokens can’t turn into junk tokens. On the other hand, as the AGTIntelligence network grows, there will be an increase in the value of the token, which will synergistically entail the growth value of the AGTI token business – and increase overall sustainability.
The authors plan to release the first version of the app by around the 2024th year, with the ability to solve simple problems, construct networks, pay, and suggestions for development. We are already assembling a team to launch this program. Further, when the network is running, the application decentralizes fairly quickly to perform different classes of tasks.
