Decentralized Platform for Crowdsourced Machine Learning
“Our mission is to build the world’s first decentralized platform for crowdsourced machine learning that enables prediction modeling on encrypted data.” — Cerebrum
The Problem: Centralized AI
The fundamental drivers of economic growth in the past 250 years have been technological innovations. The most important of these general-purpose technologies include the steam engine, electricity, information technology, and now artificial intelligence (AI). The First Machine Age, formally known as the Industrial Revolution, was when humans overcame the limitations of muscle power. We are now approaching the early stages of the Second Machine Age, in which humans are overcoming the limitations of mental capacity.
The essence of AI is learning. Just as humans learn how to identify patterns or create plans, machines can similarly be trained to perform such tasks based on powerful machine learning algorithms. Machine learning (ML) describes automated learning of implicit properties or underlying rules of data. As most AI systems today are based on ML, both terms are often used interchangeably. The main advantages of AI over human intelligence are its high scalability, resulting in significant cost savings; its consistency, resulting in a reduction of errors; and its improvability, resulting in continuous enhancements.
In the last years, established IT giants like Google, IBM, and Nvidia — fueled by the abundance of data, algorithmic advances, and the usage of high-performance hardware for parallel processing — have begun bridging the gap between science and business applications. The global market for AI-based services, software, and hardware is expected to grow at an astonishing annual rate of 15 to 25% and reach $130 billion by 2025. Most of the investment in AI consists of internal R&D spending by large, cash-rich digital-native companies like Amazon, Baidu, and Google. However, although these investments by technological giants suggest an innovative future of AI, there are also implications that it may not be shared equally. The reason is simple: AI will become a massive sector that unleashes a torrent of financial opportunities and will provide industry captains, both governments and corporates, with unparalleled technological power.
Elon Musk often quotes Lord Acton when discussing centralized AI:
“Freedom consists of the distribution of power, and despotism in its concentration.”
Musk outlines an imminent danger in today’s society: the centralization of AI in a selective group of large corporations. Similar to Acton’s premise that absolute power corrupts absolutely, few companies possessing strong AI will prompt domination over any competitors in an unfathomable and unstoppable manner. Moreover, the judgement over what constitutes good or bad usage of AI will be in the hands of a few with no accountability. This would be detrimental to society if the corporations’ values and goals are not relevant to the rest of humanity. Even if humanity agrees to entrust an entity with a centralized AI, this would still introduce a single point of failure, where the mishandling of a superintelligence could cause unintended harm onto the rest of society if it is improperly designed with no oversight.
The overarching solution to these problems is to create AI in a decentralized environment, whereby its power is openly accessible and its design, goals, and values are publicly verifiable and determined democratically with governance, in which the will of the collective overrides any malicious activities. Given its growing accessibility, broadening applications, and specific relevance to the industrial sector, it comes as no surprise that AI needs to be decentralized and distributed among researchers, investors, and companies. Decentralized artificial intelligence (DAI) offers a unique proposition that no other AI can offer: the democratization of AI.
The Opportunity: Ethereum + IPFS
Blockchain, an enormously powerful shared global infrastructure that can move value around and represent the ownership of property, is an emerging technology that capacitates applications to become decentralized. Integrating blockchain into AI is an optimal combination to create an effective decentralized AI. The inherent immutability of data published on the public ledger leads to absolute confidence in training/testing data and models they produce. Properly incentivized blockchain applications will garner increasing amounts of shared models and data leading to improved prediction modelling. Furthermore, an open blockchain grants anyone the ability to turn machine learning skills into assets by constructively contributing to the DAI.
DAI can utilize decentralized data storage via disintermediation of centralized storage providers. Distributed file systems like IPFS make an impact in the cost of storing data and in the amount and quality of data available. IPFS liberates DAI by moving away from proprietary data silos to blockchain-enabled shared data layers. DAI will shift power in Big Data away from data owners to those who can access the most data and rapidly gain insights. Data will no longer be locked away in corporate databases but belong to all with access to IPFS. When data moves out of proprietary systems onto open blockchains, having the data itself is no longer a competitive advantage rather interpreting the data becomes the advantage.
DAI will realize a set of analytic capabilities that can handle an otherwise insurmountable stack of data. DAI that applies a scalable process for producing an AI solution on top of open, shared, blockchain-based data layers will have enormous disruptive potential across industries. The potent combination of both AI and blockchain technology can transform many industries by reinventing information exchange.
The Solution: Cerebrum + Ethereum + IPFS
Cerebrum tackles the democratization of AI by harnessing the same components driving the supremacy of existing centralized AI juggernauts: lots of data and AI research talent. Cerebrum is the world’s first decentralized platform for crowdsourced machine learning. Cerebrum enables anyone to encrypt their data and host machine learning competitions to utilize the collective intelligence of models created by the world’s largest community of machine learning experts without sacrificing data privacy. Moreover, Cerebrum will promote machine learning competitions on open datasets to collect the world’s largest library of the most optimal intelligent agents for solving particular goals. Anyone will be able to use these agents with their own encrypted data to create multi-agent systems that solve complex tasks. Cerebrum implements blockchain technology to generate distributed, decentralized artificial intelligence by creating an open, competitive platform that aligns incentives between those wishing to gain predictive power from their data and scientists seeking to monetize their machine learning skills. Cerebrum will democratize the power of AI by allowing everyone to customize and use the best intelligent agents in the world.
Cerebrum is built on the Ethereum blockchain and the Interplanetary File System (IPFS). Ethereum enables Cerebrum to control the transfer of Neurons, tokens used to host or participate in machine learning competitions, via smart contracts. Cerebrum will transition between various mind states throughout its development: Simple Mind, Free Mind, Open Mind, Strong Mind, and Super Mind. These mind states evolve Cerebrum from a constrained, one-dimensional AI platform to an unconstrained, multi-dimensional, general-purpose AI platform. The foundation of this evolution is based on the customization and communication within and between multi-agent systems, where competitions within Cerebrum represent and generate intelligent agents that accomplish specified goals.
Cerebrum’s notion of modular agents being broken down into atomic agents is a popular approach for achieving higher forms of intelligence and is highly revered by artificial intelligence experts such as Ray Kurzweil in his Pattern Recognizer Theory of the Mind. This concept of artificial intelligence systems integration is making individual software components, such as speech synthesizers, interoperable with other components, such as common-sense knowledge bases, to create larger, broader and more capable systems. Accordingly, Cerebrum embraces this holistic and integrative viewpoint on intelligence. Cerebrum is inspired by the human brain wherein a massive number of simple constructs, neurons, work in tandem to create remarkable complexity. Both computational neuroscience and artificial life are fields developed under this umbrella of emergence. Cerebrum will aggregate related agents to create systems that make optimized decisions to achieve higher-level goals. Consequently, Cerebrum serves as a global, distributed operating system for designing and using complex intelligent agents.
In the sphere of business, AI is poised to have a transformational impact, on the scale of earlier general-purpose technologies. Although it is already in use in thousands of companies around the world, most big opportunities have not yet been tapped. The effects of AI will be magnified in the coming decade, as manufacturing, retailing, transportation, finance, health care, law, advertising, insurance, entertainment, education, and virtually every other industry transform their core processes and business models to take advantage of machine learning. The bottleneck now is in implementation and centralization. Cerebrum sets the foundation for AI progress to avalanche in the years to come by democratizing the power of AI through an open, distributed, and decentralized AI that everyone can leverage to transform industries without the threat of autocratic tech giants.
Soon, we will discuss how Cerebrum will implement a blockchain-based AGI.
“When everyone with creativity and vision can leverage the power of intelligent systems, there’s no limit to what this technology can accomplish, and that benefits all of us.” — Jia Li — Head of R&D, ML/AI at Google
Learn more about Cerebrum at https://cerebrum.world.
Acknowledgements:
