Page 127 - AI Standards for Global Impact: From Governance to Action
P. 127

AI Standards for Global Impact: From Governance to Action



                   21�3  Standardization gaps and needs


                   Speakers identified priority areas where standards are needed:
                   •    Definitions and taxonomy                                                                    Part 3: Future AI
                   •    Component-level clarity on building blocks and interoperability
                   •    Simulation-to-reality gap: Bridging synthetic training data and real-world environments.
                   •    Benchmarks: Application-oriented benchmarks are needed to reflect real-world
                        requirements.
                   •    Socio-technical standards: IIn addition to technical specifications, standards development
                        should consider socio-technical dimensions such as human rights,

                   Standardization activities of ITU-T on embodied AI is shown in Box 5 below, including standards
                   under development and published.


                   21�4  Collaboration across standards bodies

                   Participants agreed that international collaboration is essential and made for:

                   •    Multistakeholder inclusion, with startups, emerging companies, academia, and civil society
                        engaging alongside large institutions.
                   •    Parallel pre-standardization discussions supporting formal standards-development
                        processes.
                   •    The consideration of human rights, privacy, dignity, children’s rights, and accessibility
                        alongside technical discussions.
                   •    accessibility, and environmental impact.




                       Box 5: ITU standardization activities on embodied AI

                       ITU-T Study Groups are actively engaged in standardization work on and related to
                       embodied AI. Information about related published standards and ongoing work items
                       are provided below.

                       ITU-T Study Group 2

                       ITU-T Study Group 2 has published two ITU-T Recommendations related to embodied
                       AI.
                       •    Recommendation ITU-T M.3167.1 (03�2025)
                            With the continuous development of Internet of things (IoT) technology,
                            the application of intelligent maintenance robots (IMRs) in the field of
                            telecommunication smart maintenance (TSM) is increasing. Recommendation
                            ITU-T M.3167.1 provides the requirements for the interface between the IMR-
                            based smart patrol system (IbSPS) and the telecommunication smart maintenance
                            system (TSMS) at a protocol-neutral level. It describes the position of the relevant
                            interface and specifies the high-level requirements for interface interaction, as
                            well as specification level use cases for each requirement.












                                                            115
   122   123   124   125   126   127   128   129   130   131   132