Application of Machine Learning in Blockchain applications
It is amazing to see two different technologies have a meeting point and let us see how they meet each other. Machine Learning and Blockchain. Machine learning is about 4 things Data, Compute power, Domain Knowledge and Models (Algorithms + Data). Blockchain is also about four things - Software engineering, Domain Knowledge, Decentralized data and Cryptographic Science. By this time we have identified that there are two things common between them and for a moment let us leave domain knowledge out of the context as it is required for business or technology to succeed. The left out is Data and it is a common attribute in blockchain and Machine Learning. Blockchain is all about how to decentralize and eliminate middleman in the chain. Blockchain allows the data to be owned by the users and trust is major element in the outcome of blockchain and actionable intelligence is an outcome of Machine Learning with the right prediction.
The biggest advantage of blockchain data compared to a centralized data which many of the platforms currently use is the immutability of the data. As the data is immutable in Blockchain the usage of decentralized trusted data in machine learning makes the prediction much more accurate than the changed data. This gives better predictions and therefore better business outcomes. For the sake of this discussion let us take an application where both machine learning and blockchain technologies. I would tend to take instant messaging - an application which is widely used by all of us to do an analysis of how machine learning and blockchain can be used.
- Immutable Data: Instant messaging has become a touch point for many E commerce websites to handle customer queries. These could be specific to the site or it could be the one that is widely used in the market like Messengers. If I want to analyze the messages to do the sentiment analysis for a customer the data is stored in the central platform. However it is not guaranteed to be immutable. A possibly mutable data is not good for sentiment analysis and may well give wrong result on prediction. So we moved away from a data that is being used for sentiment analysis to a stage of using immutable data (Better Quality Data) for doing a sentiment analysis. An immutable data gives better prediction than mutable data. The first advantage is the use of Immutable Data for customer touch points guaranteed by blockchain technology. Any review done on a product becomes more immutable and therefore the recommendation systems appear better.
2. Data Security: The main advantage of blockchain is the decentralization of trust and therefore the data is distributed across network nodes that participates in the block chain. The biggest challenge for Machine Learning Coders is the data. The biggest challenge for the people who own the data is to ensure that the data they give out is not being leaked out and at the same time the advantage can be taken from usage of Machine Learning. Any data that is being shared it is shared with the consensus of the miners in the blockchain. The security of the data that is getting mined is ensured through the blockchain technology. This apart most of the blockchain uses asymmetric cryptography providing better security of data.
3. Data Oligopoly: Two revolutions are happening(Cloud and AI) and the revolution has lead to the oligopoly in the field of AI. AI requires Data and the Data is owned by only few players who were able to see the data as the new oil. Large cloud service providers have become the owners of large amount of data. Unlike the conventional oil which is at least owned by few countries the data is owned by four to five major cloud service providers or platform service providers and some of the names like Google, Facebook, AWS etc. Blockchain brings in a new paradigm of data being localized and is owned by multiple miners. Chances of Data Oligopoly being broken is high using Blockchain technology and this leads to machine learning applications. Blockchain possibly provides an opportunity to move away from a Data Oligopoly based Bias leading to more useful business outcomes.
In the conclusion machine learning is all about predictions based on data and blockchain is all about the security of data. Both of them can complement each other to produce better predictions leading to better business outcomes. I hope you will read this blog and let me know your feedback whether this makes sense.
