While the “see one, do one, teach one” paradigm of learning by doing is prone to such an error mode, personally I don’t find it as much of a problem when taking a more “supervised learning” approach—inferring your function from labelled training data. If you just shadow someone doing your exact job for a few days, chances are you’ll avoid most mistakes just by copying the observed behavior. For a restaurant job—or most others—that should cover a majority of situations.
What about the iceberg iceberg, when noticing your first iceberg you realize there was a metric ton of icebergs under the iceberg.
I find that similar to the concept of fractal wrongness. What distinguishes an iceberg from a fractal is that you’re in situations where someone is resisting exposing the whole iceberg for one reason or another. In the dishonesty scenario, you realize one lie reveals many others but only because that person has left you a tidbit of information that cracks their facade and allows you to infer just how deeply they’ve lied to you—or in the case of attraction, an event or action that only would occur if they had a much greater level of attraction existing below the surface.
or in the case of attraction, an event or action that only would occur if they had a much greater level of attraction existing below the surface.
This seems misleading, à la Sherlock Holmes’ “Eliminating the impossible”. A charitable reading would parse as:
“or in the case of attraction, an event or action where the most probable world (as calculated with Bayes) in which it happens also requires a much greater level of attraction existing below the surface.”
Just wanted to make sure I’m not inventing new interpretations and that there’s no hidden inferential distance.
While the “see one, do one, teach one” paradigm of learning by doing is prone to such an error mode, personally I don’t find it as much of a problem when taking a more “supervised learning” approach—inferring your function from labelled training data. If you just shadow someone doing your exact job for a few days, chances are you’ll avoid most mistakes just by copying the observed behavior. For a restaurant job—or most others—that should cover a majority of situations.
What about the iceberg iceberg, when noticing your first iceberg you realize there was a metric ton of icebergs under the iceberg.
Or a recursive iceberg, where you realize there’s a whole nautical mile worth of rabbit hole left to go down?
I find that similar to the concept of fractal wrongness. What distinguishes an iceberg from a fractal is that you’re in situations where someone is resisting exposing the whole iceberg for one reason or another. In the dishonesty scenario, you realize one lie reveals many others but only because that person has left you a tidbit of information that cracks their facade and allows you to infer just how deeply they’ve lied to you—or in the case of attraction, an event or action that only would occur if they had a much greater level of attraction existing below the surface.
This seems misleading, à la Sherlock Holmes’ “Eliminating the impossible”. A charitable reading would parse as:
“or in the case of attraction, an event or action where the most probable world (as calculated with Bayes) in which it happens also requires a much greater level of attraction existing below the surface.”
Just wanted to make sure I’m not inventing new interpretations and that there’s no hidden inferential distance.