Post

Contemporary frontier models are not conscious.

Indeed, the word itself is sufficiently void of anything but casual and subjective meaning such that it is counterproductive to apply it in any rational discourse at the intersection of the mind and computing.

Contemporary frontier models are complex stateful reactive stochastic software-intensive systems with certain degrees of self-referential reflection, capable of coherent albeit unreliable outputs. Their behaviors are for the most part taught, not learned.

But we are not the same as these software-intensive systems, in several important ways.

We are mortal, a constraint that shapes and colors our behavior. We have agency. We generate abstractions. We build models of the word, of others, and of our selves, and we have mechanisms to test the structure and consistency of those models against the reality we perceive.

Coupled with our agency, we can learn, a process that involves the intentional identification of noise, gaps, and incoherence such that when driven with self-directed behavior to devise then carry out plans to mitigate those flaws, they yield homeostatic actions that lead us to self-improve.