In order to implement the automation of advanced processing hardware
such as robots, an infinite number of training data needs to be
applied and taught. What is important is what is learned and
generated by the AI.
Couger has developed an application that guarantees the credibility
of the Blockchain-based AI’s learned history.
The application’s possible conduct is as follows :
・Store invariable learning models and trace each record
・Confirm which data to base learning on and whether the learning
model has been generated accordingly
・Learning model rollback according to administrator
This application can be roughly divided into three functions :
learning phase, inferring phase and rollback. In the learning phase,
the AI’s learned history is stored within Blockchain, and the
learning data is stored in a distributed file system (IPFS). In the
inferring phase, the designated learned models are summoned from the
IPFS and displayed as predicated results. In rollback, in the
circumstance that an error has occurred within a learning model,
only those with administrative authority are able to rollback the
said model.
