The existing AI market is increasingly controlled by tech giants like Google, IBM and Microsoft, all of which offer cloud-based AI solutions and APIs. This model assumes little control of users over the AI products, and in the long run, such a centralized model could lead to the monopolization of the AI market. This could cause unfair pricing, a lack of transparency, interoperability and limited participation of smaller companies in AI innovation. Fortunately, we are witnessing the emergence of a decentralized AI market, born at the intersection of blockchain, on-device AI and edge computing/IoT.
Decentralized Intelligence Market
The first segment of the market involves decentralized autonomous organizations deployed on Ethereum blockchain. DAOs are algorithmic companies run by AI agents in charge of managerial decisions. This model can be effectively leveraged in the distribution of royalties, subscription payments and more. The second segment is represented by device-centric AI supported by mobile machine learning libraries like Appleâ??s CoreML. On-device AI is closely related to AI-based edge computing that allows complex AI algorithms to run on IoT devices like sensors, security cameras, drones or autonomous vehicles.
Decentralized Intelligence Networks
Another recent trend in the AI market is the development of decentralized networks built on the blockchain. One such project is SingularityNET
, which seeks to dodge the monopoly of tech giants over AI by allowing any company or researcher to monetize their AI solutions and get access to a variety of AI algorithms. To enable cross-AI communication, the SingularityNET team created a protocol shared by each AI product on the network. This protocol supports data exchange and sharing across different algorithms, which is helpful in building multitier AI applications.
Such applications can combine multiple algorithms performing different sub-tasks and have access to the training data exchanged on the network. One example of such an approach is the development of a comprehensive AI-based cybersecurity solution. Currently, there is no single software package that handles all security-related tasks, which means that companies have to use various centralized AI solutions and customize them to their needs. Developing such a solution also brings about the problem of data security and domain-specific knowledge.
Homomorphic encryption that makes different datasets visible to different classes of users in different aspects can solve this problem. Companies can also combine the expertise of different cybersecurity AI agents on the network, which will safely exchange security information, outsource tasks and cooperate in solving common security issues. Such decentralized network will involve efficient division of labor and offer access to solutions without the need to manually obtain data and customize algorithms.
Decentralized intelligence has a number of advantages over centralized solutions in the following scenarios:
You need an autonomous AI solution that runs in the decentralized environment and implements contractual obligations: By definition, centralized proprietary solutions cannot be exposed to many users in the decentralized network. If your goal is to run a fully autonomous AI agent making smart managerial decisions and distributing profits, decentralized AI on blockchain is the way to go.
One example of this approach is Terra0, a project of the augmented self-owned forest
proposed by researchers from the University of Arts, Berlin. Terra0 is an AI engine that decides when and how to sale forest land and wood by using drones for data and ML models for the assessment of the best selling strategy.
You need an AI optimized for the on-device performance and not dependent on network connectivity: Due to network connectivity problems, battery power constraints and low computing power, mobile devices are not a good option for running cloud-based AI software. In particular, high latency and low throughput can compromise the speed and performance of AI applications adversely affecting user experience. Whenever you need to run and train AI on mobile devices, a decentralized network becomes a better option.
You want to sell your AI algorithms while maintaining proprietary rights: There is no way to sell your AI algorithm and retain proprietary rights for it (the same goes for selling mobile apps in the app store right now). To fill this gap, decentralized AI networks offer an opportunity for developers to make their algorithms available for commercial usage as AI as a service.
At the same time, centralized AI still remains a good option if you need a very generalized ML model that you can easily plug into your application. Google, Microsoft and IBM have developed the best generalized machine learning models on the market that are trained on huge data sets and built according to the top ML standards and bleeding-edge ML algorithms. Reinventing the wheel is not an option if you want proven image or speech recognition features in your application. A more viable solution is using cloud-based ML APIs provided within a pay-as-you-go model that ensures cost efficiency and scalability of your AI-based solutions. Major providers of centralized AI have a comprehensive suite of services for image and video recognition, emotion AI, speech recognition, predictive modeling and other common AI/ML tasks.
In the long run, decentralized solutions can produce the radical democratization of the AI market, optimization of solutions for a wide variety of use cases, easy integration and communication between different algorithms through a single protocol and the development of interoperability standards, which will ultimately lead us to the era of AGI (artificial general intelligence).