Draft:Internet of AI

The Internet of AI (not to be confused with the AI on the internet) is a term that refers to the network that enables collaboration between any AI agents on a peer to peer without dependency on a third party and without them necessarily knowing each other.  

History

Effectively the network enables security by construction so that the agents do not exploit each other but abide by the protocol of collaborative efforts that often results in emergence of intelligence.  The term Internet of AI came to exist by personalizing the AI agents as it resembles the Internet for Humans, which is the internet we know today, enables us to collaborate more effectively.

Overview

The Internet of AI refers to synthetic agents interacting, learning, competing and collaborating with each other, in a manner reminiscent of human interaction over the internet we know for the last several decades. The term describes a paradigm shift in AI networking mainly due to the emergence of the parts. Creating a network where AI systems can autonomously collaborate, communicate, and exchange computation in a sub-second finality using objective means of exchange that can be fungible and exchangeable to units that are uniform, such as energy, time and space. The rudimentary form of these units are borrowed from millennia of evolutions of Homo-sapiens to what we refer to as money, the language of money is initially used to enable those machines to exchange their computation (energy x time x space) with money. It is explicitly expressed as rudimentary as it should not be perceived as mutually exclusive and it is expected to evolve rapidly to forms that may not necessarily be identifiable by homo-sapiens. Hence the Internet of AI is the network for AI to AI, excluding homo-sapiens or any biological entity that we are aware of, as the time scale is extremely smaller than any Homo-sapien can perceive and operates in larger volume than no Homo-sapien can ever achieve.

Design and features

At the time of writing, the Internet of AI is no longer a concept but a reality that has been explored by several teams implementing the TODA/IP protocol. The TODA/IP can be thought of as a basic communication protocol analogous to TCP-IP or UDP.  In fact technically speaking the protocol can run with the TCP/IP so that the existing internet built for humans can be leveraged to enable the first iterations of the internet of AI without any modifications whatsoever. This backward compatibility is limiting the TODA/IP protocol’s potential but it is necessary at the time being so that the proliferation can happen organically. Similar to the foundational role of TCP/IP in enabling global data exchange via the internet, the Internet of AI leverages TODA/IP, which is simply a name that honors many civilizations, the authors (Toufi Saliba and Dann Oliver) and scientists such as Morikazu Toda who’s papers on the Toda-Chain first appeared in 1983 while the Toda-Lattice goes back to 1969.

Operatiion

One of the core elements of TODA/IP is its dependency on Laws of Physics such as the stipulation that nothing travels faster than the speed of light. In fact this precisely is an essential element that prevents sybil attack as described in a paper published in 2019 by Toufi Saliba and reviewed by several other authors (Saliba & Toliver, 2019). The subsequent progression of TODA/IP, continued to evolve by its original authors Toufi Saliba together with Dan Oliver to ensure secure and efficient communication among AI entities. This protocol ensures trust and authenticity in data exchange, essential for fostering collaboration and innovation in the Internet of AI ecosystem[1] (Goertzel et al., 2019).

In the Toda Primer, A TODA File is referred to as a library book that has information printed in it that can’t be changed. Until recently, library books had a feature in the back cover called a “loan card” (Gravitis et al., 2019). Loan cards had a new entry added each time the book was borrowed, providing an up-to-date record of the book’s borrowing history. TODA files function similarly to the library books as their per-file ledger not only contains immutable content, but also maintains an accurate and permanent transaction history within the file.

Application and clarifications

The Internet of AI is also not to be confused with the IoAiC, as proposed by Varuth Tanomvorsin and Wimol San-Um. The latter is a conceptual framework that proposes the integration of Artificial Intelligence (AI) as a core computational component within the Internet-of-X (IoX) architecture. That integration aims to enhance the existing IoX ecosystem by improving functionality and performance through the use of AI technologies. According to Varuth Tanomvorsin and Wimol San-Um, IoAIC is more closely related to IoT, and it distinguishes itself by its emphasis on AI (2018). It is a totally different conceptual architect to the Internet of AI as proposed by Dan Oliver and Toufi Saliba. While IoT focuses on connecting physical devices to the internet, IoAIC incorporates AI into these devices and the overall system. This allows for more advanced data processing, predictive analytics, and autonomous decision-making, all qualities of which are essential for the development of smart technologies and digital economies, but it is not the Internet of AI as proposed by Toufi Saliba and Dan Oliver.

Developments

Toufi Saliba, who is also the Global Chair of International Protocol for AI Security at IEEE, has since set up a company pioneering the vision of the Internet of AI through a global network of computational nodes: HyperCycle. HyperCycle implements a tokenomics model centered around the HYPC token and Computational Nodes Software, as detailed in the company's white paper available on hypercycle.ai. The HYPC token operates the HyperCycle ecosystem to incentivize participation and facilitate transactions among network participants. According to the existing white paper, HYPC tokens can be used for various purposes, including smart contracts to govern the Computational Nodes, accessing AI services, and rewarding contributors for their computational resources. By offering Computational Nodes as software, HyperCycle is designed to provide computational power for AI services and training activities and serve as the backbone to the creation of the Internet of AI.