en
Back to the list

Blockchain, machine learning and AI

Blockchain

en.cryptonomist.ch 15 May 2020 18:50, UTC
  
Reading time: ~2 m

The blockchain can also be used in the machine learning sector and in combination with artificial intelligence (AI).

Let’s see how they can be integrated or how they have already been used to take these technologies to the next level. 

The blockchain in machine learning 

Blockchain technology allows to have a considerable degree of reliability from the point of view of data security: when data is recorded on the blockchain, first going through a “reliable” source such as an oracle (possibly decentralized) it can be retrieved in a decentralized way and without being tampered with.

This concerns any kind of data, hence also those managed by machine learning or artificial intelligence systems. 

For example, the project undertaken by the company SingularityNET allows to decentralize algorithms and have a data market on the blockchain that anyone can utilize, so not a closed and centralized market. This is how the Sophia robot was developed.

Another project in this field, based on the EOS blockchain, is Effect.ai which creates a bridge between the blockchain and the various dApps (decentralized applications). 

Machine learning 

There is an important element that is closely related to artificial intelligence, namely machine learning, which provides computers with the ability to learn something without explaining anything to them, but by accessing the various data they find on the network or on the blockchain.  

Machine learning relies on data which, when uploaded on the blockchain, and thus to some extent “certain” and “verified”, can be analyzed to generate a model or prediction of a certain event. 

Going further, with the integration of these elements it could be possible to reach an evolution also of the consensus protocol, passing from the more traditional PoW, PoS, DPoS to PoI (Proof of Intelligence), for example used by AION.

This is an evolution of PoW, but in which validators train an AI to solve increasingly complex mathematical problems so as not to increase the difficulty, since it will be the AI itself to set the “difficulty” to prevent attacks by subjects with enormous power. 

At the moment there is still a long way to go before seeing something complete and working in this area, but it is not necessarily impossible to achieve it in the future, given the progress that is being made in the quantum computer sector.


   Source
Back to the list