6GMedia - work topic 2- Characteristics of AI-enabled applications
Proposition 1: SA4 should study the support of additional media modality and codecs or their enhancements for 6G, building upon 5G and 5GA studies.
Proposition 2: SA4 should define terminology applicable for AI/ML data (feature, token, embeddings, latent, intent, etc.), study the relevant AI representation formats to ensure a common understanding across WGs and study which interchangeable format/codecs are applicable.
Proposition 3: SA4 should identify and study a set of spatial compute functions that may benefit from off-device processing for mobile AI-enabled applications and services
Proposition 4: When studying traffic characteristics for AI-enabled applications and services, SA4 should aim at developing generic QoS and QoE mechanisms suitable across the diversity of traffic patterns.
Proposition 5: SA4 should study the necessary enhancement to QoS framework enabling finer granularity and context awareness.
Proposal 6: SA4 should specify the procedures for real-time QoE-based adaptation of multimodal media and define QoE metrics for real-time and delay-bound AI inference.
Proposal 7: SA4 should characterize the impact of QUIC-based protocols on the delivery of AI data and on the traffic characteristics of AI applications and services over QUIC-based protocols especially for real-time or delay-bound applications.
Proposal 8: SA4 should study the potential integration of the extensions SA2 has defined for QUIC-based transport solutions into the media delivery architecture. For the RTC architecture, this would be done by leveraging the FS_Q4RTC-MED study and applied to these AI-enabled applications.
Proposal 8: SA4 should study the impact of multi-devices on the QoS and QoE framework.
Proposition 9: SA4 should consider heterogenous multi-devices associated with the same user for the definition of QoE metrics and for the study of potential enhancement to QoS for real-time and delay-bound AI inference.