Page 12 - Preliminary Analysis Towards a Standardized Readiness Framework - Interim Report

Preliminary Analysis Towards a Standardized Readiness Framework

2. Introduction

In this cross-domain study, we analyzed use cases related to the use of AI in different verticals
such as transportation, agriculture, disaster management, and health, with an aim to find
patterns in applications of AI in different scenarios. The goal was to derive a standardized
metric that could be applied to measure the readiness to use AI to solve relevant problems.
Our analysis of 16 use cases included the following outstanding characteristics of use cases to
be considered while evaluating AI readiness: The type of data used in each use case, domain-
specific research needed in the use case, infrastructure requirements, human factors supported
by standards, interoperability, data management, and data storage. These characteristics are
analyzed in “Table I – General use case analysis and AI impacts” in Appendix A.

The main AI readiness factors identified in this report are:

1) Availability of open data
The Kingdom of Saudi Arabia set up an Open Data Platform [3] providing datasets to the
public to enhance access to information, collaboration, and innovation. The major areas of
dataset availability in this open data platform are Health, Agriculture and Fishing, Education
and Training, Social Services, and Transport and Communications. The transportation system
in the major cities enables advanced use cases such as tracking vehicles with excessive speed
to guarantee pedestrian safety, providing the best driving routes to reduce the number of traffic
jams, and reducing the mortality rate caused by collision. These use cases utilize diverse data
such as imagery data collected by Closed-circuit television (CCTV), a detailed map of the city,
traffic signal information, and vehicle Global Positioning System (GPS) details. This is a prime
example of the collection and hosting of open data and enabling analytics for traffic safety.

Private entrepreneurs, startups, and industries are developing applications or designing
algorithms to achieve Sustainable Development Goals (SDGs) such as safe transportation.
However, there are still challenges in data collection, cleaning, and preprocessing which hinder
the opening of data for everyone. A well-designed open data strategy would make sure high-
quality open data is available for scholars, developers, and analysts to design solutions based
on real-world problems, thus enhancing the impact of AI on society.

2) Access to Research – domain-specific vs. AI-specific

The importance of domain-specific biological and medical data in predicting intoxication
level and modeling and validating AI models in predicting with accuracy shows that access to
domain specific and AI specific research, both remain equally important.

For example, while assessing the safe driving behaviors under the influence (see Clause 4.1.13),
not only monitoring of driver behavior was considered, but even biological data such as chest
movement and breath were collected. Chest movement was collected, analyzed, and predicted
heartbeat would serve as a reference of blood alcohol level.
3) Deployment capability along with Infrastructure

Networks interconnect various nodes in the AI/ML pipeline [ITU-T Y.3172] such as the source of
data, pre-processing, model, and distribution of inference. For instance, in agriculture use cases
(see clauses 4.1.7 and 4.1.8) soil sensors or water sensors should be deployed in the field with
high quality and numbers so that the volume and variety of data are sufficient to train models
with accuracy. Disease detection for wheat crops discussed in [38] provides an exemplary

4