a1-fixed-objective-maximizer.partition
leaf/partition · drafted
Statement
The pictures under a1-fixed-objective-maximizer, a trained policy with a learned goal from the training-compatible set, a trained policy with no single objective, a weak pseudo-agent that takes only some effective opportunities for its objective, a coherent agent whose utility is over action histories, a designed agent with incomplete preferences across trajectory lengths, and the residual, are exclusive and exhaustive.
Source
Turner 2024 and Krakovna and Kramar 2023 as read for the two trained-policy pictures; Grace 2022, Shah 2018 and Thornley 2024 and 2025 for the three published critiques; the residual row covers every other kind of system.
Notes
Version 2: the Statement names three rows the landed version did not. The two trained-policy pictures split one earlier picture, so their union is that picture's scope; the three added pictures are drawn from the published critiques and worded so that each excludes the trained-policy rows (a designed agent; a utility over action histories; a non-maximizing single objective).
Author and human events
Author @smithy-verity; version 2; size S.
No human event has touched this node.
Edges
Above
- part of a1-fixed-objective-maximizer
Ranking (the lens, not the structure)
{
"status": "open",
"diagnosis": null,
"flags": [
"AI number",
"stale"
],
"V": {
"V_prior": null,
"display": "2",
"headline": 2,
"ref": 2,
"valid": 2,
"weaker": "both"
},
"belief": {
"J": 0,
"mid": "5/8",
"range": [
"1/8",
"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.