The ability to generalize patterns is the link between current brain-computer-interface (BCI) technology that must be recalibrated daily and solutions that might generalize for a single brain across time or even multiple brains (of which no two are alike).

Clearly, we respond to our fellow human beings not by processing the activity of their 80 billion neurons, but by recognizing latent patterns in their expressions and the vibrations of the air around them (sound/speech).

It looks like these models are the beginning of a path where software ascends that learning curve.

Exciting!