MLIR is being used as another solution to implementing and optimizing Tensorflow logic. This informative message is benign and is saying MLIR was not being used. This is expected as in TF 2.3, the MLIR based implementation is still being developed and proven, so end users are generally not expected to use the MLIR implementation and are instead expected to use the non-MLIR feature complete implementation.
Update: still experimental on version 2.9.1. On the docs it is written:
DO NOT USE, DEV AND TESTING ONLY AT THE MOMENT.