Common basic knowledge for artificial intelligence algorithms
Completed on 05-Jul-2015 (19 days)
The preliminary character of the proposed system
should never be misunderstood as having an ambiguously-defined essence; in fact, this project (and my personal impressions on this front) might almost be defined as of the take-it-or-leave-it type.
The main goal of all the ideas in the current project is to propose (the preliminary structure of) a system which is expected to always verify the following principles:
- Adaptable & Scalable. The reliability of a system in charge of standardising a so complex and variable reality is highly conditioned by its ability to adequately address any unexpected issue.
- Collaborative. The intended systematisation can only be accomplished in case of having a relevant enough amount of resources. In principle, the most logical way to collect all these resources is by accepting (voluntary) contributions of organisations which have developed relevant AI-focused projects.
- Overall Compatible. Curiously, a big proportion of universally-utilised systems do not aim to create highly compatible outputs, but rather show monopolistic behaviours on the lines of "you have to do everything my way". This principle is important because of the aforementioned absolute dependence upon voluntary collaborations, which are more likely to occur in case of having an attractive enough structure.
- Interdisciplinary. Computing and data standardisation will undoubtedly be the most relevant fields of expertise, but not the only ones. Different areas of knowledge will also have to be considered in order to give a comprehensive answer to so peculiar situations (i.e., machines of any kind running almost autonomously). For example: non-technical specialists to analyse the ethical or political implications of certain issues.
- Monitored. Any standardised (data) format expected to have general applicability has to be controlled by an authority ensuring its adequacy. In this specific situation, such a generic requirement is even more important on account of two issues: on one hand, Artificial Intelligence is already a very sensitive matter (which is even likely to become exponentially more sensitive within the medium term); on the other hand, the proposed open-to-collaboration format reinforces the necessity of a monitoring instance.
All the monitoring-related issues are analysed in the Supervision section. Nevertheless, I can already anticipate that these tasks are expected to be managed by an international organisation, with a national-government level support in each country.