How AI could make future wireless networks fix themselves

Today's 5G networks rely on separate AI models for different tasks—traffic prediction, fault detection, optimization—each trained independently. This vision paper argues 6G should instead use a single foundation model as the backbone, with task-specific knowledge distilled into lightweight models for edge devices, orchestrated by multi-agent AI systems that autonomously manage the network. The shift moves from building "networks for AI" to "AI for networks," enabling self-healing infrastructure that diagnoses and recovers from failures with minimal human intervention.