Kozminski Techblog

Reading Time: 4 minutes

Those who actively follow the recent news and trends in the new technologies’ world must have noticed the great rise in popularity of AI and blockchain- two technologies that work on utterly different rules.

While AI uses computer programs to act smart like humans, blockchain technology uses a shared and secure system to keep a reliable record of transactions without a central authority.

Indeed, both of them pose the crucial background for other tech areas like cryptocurrencies, DeFi, tokenisation of assets (based on blockchain), ML, NLP, personal assistants (based on AI).

But what would happen if these two prominent technologies- blockchain and AI were interconnected?

What would it look like and how would such system work? Is it the same as the federated learning approach? What could be cons and pros of the decentralised AI? Will we ever breathe life into such an approach in-real-life issue?

In this blog, we will delve into all of these questions, and we will try to analyse essential aspects regarding the merge of two, the most ground-breaking technologies of our era.

What is decentralised AI?

Decentralized AI systems use blockchain technology to distribute, process, and store data across a network of nodes. Users can benefit from AI-generated insights on their local devices without sharing their data with a centralized authority. They can process data on their device using a prebuilt AI model and share the results without revealing their personal data.

So, is the decentralised AI and federated learning the same?

Shortly speaking no, it is not the same thing but, indeed, both approaches share many common characteristics.

The key distinctions between federated learning systems and decentralized artificial intelligence (AI) systems lie in the control and processing of data.

In federated learning, organizations have centralized control over the AI model used to process datasets, while the data itself is stored and processed in a decentralized manner, typically on user devices.

On the other hand, in a decentralised AI system, there is no central entity in charge of processing the data, and the decision-making process is distributed across multiple nodes without a central authority.

*For those who have not heard of it yet: federated learning is a machine learning approach where the model is trained across multiple decentralised devices or servers holding local data samples, without exchanging them. This allows for training a global model while keeping the data localised, addressing privacy concerns, and reducing the need for centralised data storage.

What are 5 core components of the decentralised AI?

According to Professor Longbing Cao, AI researcher:

“DeAI refers to the AI thinking, methodologies, technologies, systems, and services for developing, managing, and deploying decentralized intelligence in decentralized settings.”

Thus, the core components of decentralized AI include:

• AI platforms or decentralized apps (dApps)
• Blockchain-distributed ledgers
• Smart contracts
• Federated learning
• Homomorphic encryption technologies

These components form the foundation for decentralized AI systems, enabling users to provide data to AI training models without disclosing it to a third party, independent processing and decision-making, distribution of pre-built training models across a network of nodes, and increased transparency over an AI model’s processing activity.

Generative AI vs Decentralised AI

Now with the basic knowledge on the DeAI, it is crucial to distinguish between generative and decentralised ai systems’ features.

Generative AI:

• Focuses on creating new data, such as images, text, or music, based on patterns learned from existing data.
• Typically operates on a centralized system or server where the AI model is trained and deployed.
• Often requires large amounts of computational resources for training and inference.
• Can be used for creative tasks like art generation, text generation, and music composition.

Decentralized AI:

• Utilises blockchain technology to distribute, process, and store data across a network of nodes.
• Allows users to make use of pre-trained AI models on their local devices without sharing their data with a centralised authority.
• Empowers users to process data on their own devices and send results to third parties without sharing their underlying personal data.
• Offers increased privacy and data control for users by avoiding the need to centralise data in one location.

Benefits of the decentralised AI

Decentralized AI architecture provides several advantages for both AI developers and users. These include:

• Possibility for users to take advantage of AI-driven decision-making without compromising their data privacy.
• Increased transparency and accountability in the decision-making process of AI systems.
• Greater chances for independent researchers to contribute to the advancement of AI.
• New possibilities for secure encryption offered by blockchain technology.
• Integration with web3 and the metaverse enabled by the decentralisation.

 Drawbacks of the decentralised AI

Despite many prominent aspect which the emergence of the DeAI systems might bring, there are also certain challenges that we should bear in mind:

• Lack of centralised control- Decentralisation may lead to challenges in coordinating and controlling the development and deployment of AI models, as there is no central authority overseeing the entire process.

• Complexity of coordination- Coordinating communication and collaboration among geographically dispersed autonomous or semi-autonomous agents in a decentralised model may introduce complexities and overhead.

• Potential security vulnerabilities- Decentralised systems may be more susceptible to certain types of security vulnerabilities, especially as they rely on communication and coordination among multiple independent entities.

• Lack of standardised approach- In the field of DAI, there may be inconsistency in definitions and ambiguity surrounding the implementation and structure of these systems due to different organisations, projects, developers, and researchers following different approaches.

• Data privacy concerns- Decentralised models may face challenges related to data privacy that must be resolved, and this is where cryptography (homomorphic encryption, SMPC, etc.) plays an important role.

Is the decentralised AI the future?

Undeniably, the emergence of such the technology would be the great milestone in the digital era. Nevertheless, prior to applying such the novel approach, we should, firstly, face and, secondly, overcome the challenges which DeAI currently poses.

What is your opinion on this topic? Will one day DeAI revolutionise our daily life? Share with your own views in the comment section ;))

References

• https://builtin.com/artificial-intelligence/examples-ai-in-industry
• https://www.sciencedirect.com/science/article/pii/S1319157822000891
• https://ieeexplore.ieee.org/document/9839452
• https://www.techopedia.com/definition/decentralized-ai-dai
• https://www.researchgate.net/publication/341327134_Decentralization_of_Artificial_Intelligence_Analyzing_Developments_in_Decentralized_Learning_and_Distributed_AI_Networks
• https://bdtechtalks.com/2018/01/10/decentralized-ai-blockchain/
• https://www.v7labs.com/blog/federated-learning-guide

Prompts used in Monica AI:

• Summarise the following text where you clearly explain what DeAI is: (…).
• List the key differences between generative ai and decentralised AI.
• Based on the following article, explain what 5 key components of DeAI are.
• Based on the following article, list the main challenges which DeAI pose.