Page 39 - AI Ready – Analysis Towards a Standardized Readiness Framework

AI Ready – Analysis Towards a Standardized Readiness Framework

Table 1: Characteristics of the AI Readiness factors (continued)

AI Readiness Characteristics Notes/Description
factor
Data collec- Number of Sandboxes Number of trusted sandboxes available [ITU-T
tion and Y.3181] for validating models and data related
model vali- to the use cases under study.
dation via
Sandbox Number of published Feedback loops (AI/ML feedback loop, collect
pilot exper- controllers [ITU-T Y.3061] data, infer, action recommendations and policy
imental application, pre-built traffic plan for specific
setups occasions.
Number of edge deploy- AI-based resource allocation for low latency,
ment options higher throughput, and edge intelligence.
Number of connectivity Networks (4G, 5G, satellite networks, ad hoc
options networks).

Number of interface Protocols (e.g. MQTT [56], Transport protocols
options with error resilience such as TCP/IP).

IoT system (LoRa-based), Number of sensors deployed related to the
wireless sensors use case.
Percentage of geographies Coverage scale (Roaming across countries and
Deployment covered regions for seamless connectivity [50]).
capability Number of customizations Domain-specific physical infrastructure (In-ve-
along with needed for domain-specific hicle Safety accessories (belt, airbags).
Infrastruc- applications
ture
Deployment options and Computation capability (cloud, ground station,
resources available such as available on the edge.
devices, CPU, GPU cores, Optimizations for model deployment in
and memory budget. resource-limited settings [68]
Efficiency of energy sources Energy source: solar panels (for energy auton-
omy).
Number of public services Regarding the public services offered by
visualization dashboards governments, the number of citizen dash-
and mobile apps boards that integrated inferences and models,
including mobile applications.

32