How would you explain the trade-off between latency and accuracy in edge computing at Motorola Solutions?