Proof of Intelligence
|
Proof of Intelligence
we propose that the problem to solve by miners is the training of an artificial intelligence model, and more particularly a deep learning architecture (probably a convolutional neural network or a variation of it, although the specifics can remain open and are not described in this paper). Training a deep learning model out of data involves learning its parameters (weights) using some kind of gradient descent algorithm. This is an iterative process which is generally regarded as a computationally expensive task, therefore fulfilling such requirement.
In order to append a new block to the blockchain, first the hash value from the last block will be taken. Next, the miner will choose a maximum of N transactions from among the set of pending transactions, which are those which have not been yet included into the blockchain and thus remain unconfirmed. The value of N represents the maximum number of transactions which can be stored in a block. Miners will have freedom to choose whichever transactions they want to include in the block, either randomly or according to some policy of their preference, although it is reasonable to expect that the decision will be based on the fee included in the transaction, as it happens in other blockchain networks such as Bitcoin or Ethereum. |
 | Artist/Author: Fork
Date: Monday April 8, 2024 10:27:26 pm
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7515252/ |
| |
|
|
|
|
