Comment by bionhoward | Hacker News Reader

Comment by bionhoward | Hacker News Reader

Eh, here’s mine:

``` ### Feelometry (A Geometry of Feeling) Discovered in 2018 by Bion

*TL;DR*: Time, Valence, and Locus of Control

1. *Past* 1.1. *Positive* 1.1.1. Internal - Pride 1.1.2. External - Gratitude 1.2. *Neutral* 1.2.1. Internal - Reflection 1.2.2. External - Acknowledgment 1.3. *Negative* 1.3.1. Internal - Guilt 1.3.2. External - Resentment

2. *Present* 2.1. *Positive* 2.1.1. Internal - Flow 2.1.2. External - Pleasure 2.2. *Neutral* 2.2.1. Internal - Calmness 2.2.2. External - Awareness 2.3. *Negative* 2.3.1. Internal - Distraction 2.3.2. External - Pain

3. *Future* 3.1. *Positive* 3.1.1. Internal - Motivation 3.1.2. External - Hope 3.2. *Neutral* 3.2.1. Internal - Anticipation 3.2.2. External - Curiosity 3.3. *Negative* 3.3.1. Internal - Anger 3.3.2. External - Fear ``` Yikes at this markdown formatting lol

can you tell us a little bit about what's going on here

Sure! It’s a hierarchical numbered outline of the emotion words according to time (past, present, future), then valence (positive, neutral, negative) and finally, locus of control (internal or external).

That’s cool because it lets us understand emotion in terms of logical relationships between points in a geometric space.

When we understand emotion in this way, it becomes clear how emotion isn’t separate from logic, but rather, emotion IS logic. Logic of the memory, experience, and prediction of pleasure, “zen,” and pain!

Consider Star Trek. Commander Data wouldn’t exist without emotion, he couldn’t, it wouldn’t make sense, that would imply he had no world model about time (duh) and good vs bad.

Future Good: Hope Future Bad: Fear

I dare you to find one emotion word that doesn’t fit this framework.

The consequences for AI could be huge. Specifically, we may not be able to avoid developing emotional AI, because emotion emerges directly from the structure of reinforcement learning over time.

We might be forced to deal with “scared AI,” whether or not we want to deal with it or want to create that, because fear is a generic term for the apprehension of future loss. Don’t expect AI to be unfeeling. Evolution may work by repeated accidents, but the cumulative result is NEVER an accident.

Sorry for the formatting, but the content is crucial, and I find it infinitely more clear than “emotion as arousal” (the referent of the “James-Lange Theory” eponym) — that’s overly simplistic.

Reply on news.ycombinator.com