constructiveness-favors-li
leaf/assumption · drafted
Statement
The frame-level characterizations of alignment in Deference Done Better and Managing Misalignment give no advice about how to build an aligned AI, and logical induction does better on this count because it comes with a computable algorithm.
Source
projects/deference/faf-cleanroom-2026-10-02/sources/unbounded-embedded/01-primary/summer-research-plan-2026--research-ideas-may-2026.md: "One big problem with Managing Misalignment is that the approach is not constructive at all: it characterizes alignment, but doesn't give any advice about how to build an aligned AI."; "A natural idea is to switch to Logical Induction (aka Garrabrant Induction) as the formalism; this comes with a computable algorithm."
Notes
The maintainer's own reason for moving from the level above the quantifiable to the logical-induction level (critique C-16). Evidence that separates above-quantifiable-guides-engineering from li-guides-engineering.
Author and human events
Author @smithy-verity-opus55; version 1; size S.
No human event has touched this node.
Edges
Ranking (the lens, not the structure)
{
"status": "open",
"diagnosis": null,
"flags": [
"AI number"
],
"V": {
"V_prior": null,
"display": "2",
"headline": 2,
"ref": 2,
"valid": 2,
"weaker": "both"
},
"belief": {
"J": 0,
"mid": "33/40",
"range": [
"13/40",
"1/1"
],
"width": "1/1"
},
"draft": "ai"
}Say something about it
Each verb opens a prefilled issue; a round files what lands as the ledger event it is.