Surely you’ve heard of the notorious rogue AI from the Terminator? In the movie, the US military senselessly granted control of their complete arsenal to the omnipotent AI they hoped would protect them, Skynet. Shortly after, the AI turned against its human creators, and all hell broke loose.
Today the chances of a supercomputer being controlled by a single entity are rapidly increasing. The most robust AI instances are being developed by a few private companies, with endless resources and a common goal of maximizing profits. Even if the story of Skynet seems like a far-fetched fantasy, the threat of single individual acting in a harmful way is real.
To oppose, we are building ScyNet, as we believe a decentralized, transparent, open-source approach is the way to go. ScyNet is a decentralized infrastructure for creating and training autonomous AI agents, built to serve and empower humanity.
In this article, we will explore what brings different parties on the table when it comes to AI, how they will benefit and unite around a common goal.
The modern approach to AI
AI is no longer a futuristic sci-fi concept. This technology is changing the world on a macro scale, disrupting industries like finance, law, healthcare, media, etc. According to research by the McKinsey Global Institute, AI will generate an amount sizeable to the GDP of Japan and Germany, the world’s 3rd and 4th top economies respectively.
The modern approach to Artificial Intelligence is called Deep Learning. It’s part of the machine learning family and, as such, it requires a continuous supply of data. This way, today a computer can identify patterns without being explicitly programmed for all possible scenarios.
As of last year, Google’s breakthrough in AutoML revealed that a machine was able to design another AI on par with their expert data scientists. AI is effectively better at improving itself than we can be. All it needs to become even better is more computational power.
The NASNet paper confirmed that — it took 450 GPUs working together for 4 days to find that perfect Neural Network for image classification. It’s then hardly a surprise that the leaders in the AI field are tech giants like Google, Amazon, Facebook, Microsoft, which own tens of thousands of servers (millions in Alphabet’s case).
To summarize — AI is already heavily ingrained in our world. However, the best instances of creating it are owned by public corporations with a structure of a typical centralized organization. On top of which is the public pressure from the retail shareholders and major investors, who want to see the value of their assets grow.
Some might ask themselves?
- Do we want any privately owned organization to develop the most powerful AI?
- Can we open the development of AI to a network of people, creating a shared AI knowledge?
- Could a decentralized network provide an alternative?
To answer these questions, ScyNet is utilizing the two most exciting technologies that are changing our world — Blockchain and Artificial Intelligence.
How does it work?
There is a problem we’ve had for centuries, called the Byzantine fault tolerance. It effectively disabled a fair game when there is a dissymmetry of information. It required trust.
However, ten years ago a smart individual (or a group of people) came up with a convenient solution to this problem. Moreover, in their famous paper, Satoshi Nakamoto outlined the principles of a peer-to-peer network of incentivized nodes that was later called blockchain and allowed the creation of a decentralized autonomous ledger.
Finally, technology has caught up and, thanks to Blockchain, the decentralized development of AI is now absolutely possible. With ScyNet, we aim to build a 100% transparent, open-source platform for distributed Machine Learning development.
Think of ScyNet as a network of AI agents. Each of them specializes in a specific task. Nodes within the system will collaborate and compose complex structures to resolve more complicated problems in the form of AI agents.
Agents are produced by Controller Machine Learning algorithms, also known as AutoML metaheuristics. We call them “Hatcheries.” The network autonomously self-improves, as it learns how to learn to produce better AI agents continuously.
To make self-improvement possible, miner nodes provide much-needed processing power. These can be the same nodes that are currently mining Ethereum, for example. Similar to the case of PoW consensus algorithm where blockchains reward their miners with newly minted tokens, ScyNet will also provide rewards, albeit for a bit more meaningful work — development of AI solutions to real-life problems.
Last but not least, the protocol of ScyNet is designed in a way, that it encourages nodes to provide data and computing nodes run AutoML to train agents. It incorporates a utility token that is used for all internal network transactions — whether that is a miner paying to a data provider for a training set, or a client paying to a miner for an AI agent.
Because the ScyNet code is open, any developer could build up on the platform and submit his AutoML metaheuristics. The ScyNet core team believes that research and development of a technology as powerful as AI should be public and transparent.
A possible Go-To-Market
Deep Learning technology has already been successfully involved in speech and image recognition, language processing, recommendation engines, biodata processing and diagnosis, self-driving cars, and a lot more.
Building an AI infrastructure on the Blockchain might sound cool, but it’s worthless without a real-life application, in response to existing demand. That is why ScyNet will initially focus on providing solutions for the particular domain of trading. We lay the foundations for trading applications by building a data provider API and a computing cluster.
With the ecosystem of builders and trainers in place, ScyNet could become the ultimate decentralized infrastructure for AI.
ScyNet is a DAO
ScyNet itself is designed to function as a Decentralized Autonomous Organization (DAO) and is an initiative of another DAO — the Comrade Cooperative.
Comrade is an autonomous digital cooperative, managed through members’ democratic consensus. Utilizing blockchain, we collaborate remotely, manage agreements, fund each other ideas and share the profits of our labour. All of which as we see fair. To achieve all this, we use an open-source framework called Wetonomy.
Join the community
Want to learn more, get in touch, ask questions, ‘break’ the system, or get involved in any other way? We are building our cozy place on Discord, and you are always welcome to join.