Hypothesis about human language and understanding

Assuming that we are not living in a giant simulation (See Simulation hypothesis to fill you daily dosage of self-doubt):

We can model our ways of understanding the world as the following process

Real-life => (Physical Process) => Sensor Organ => (Neural circuitry + learning) => Language => (Neural circuitry + learning) => Logic

Machine learning design implication:

When designing machine learning agent, we could have similar transformation process:

Real-life => (Physical Process) => Sensors => (Electronic circuitry) => Raw Data => (Machine-learning layer) => Label/Category/Prediction => (?) => Higher-level logic