Alignment Workspace

shutdown-button-toy.improvable

or · drafted

Statement

The two-step, two-utility, one-button shutdown model of Soares et al. 2015 can be improved until it reflects trained systems corrected through data, loss and hardware well.

Source

Ruling 35: every toy a plan builds on carries its improvement claim. The improvement directions on record: the author's generalization to every modification (projects/deference/corrigibility-note-dump-2026-10-01/position-statement-v3.md, section 1.2), the physical channel (projects/deference/corrigibility-note-dump-2026-10-01/position-statement-v3.md, section 1.3), sequential and learned-utility settings (Armstrong's utility indifference line and Holtman's correction-function agents, in the published literature), and the training map (run 3, projects/deference/faf-cleanroom-2026-10-02/sources/corrigibility/workflow-2026-09-14b/plan/synthesis.md, agenda N2).

Notes

Improvement claim on the shutdown toy (improves: shutdown-button-toy; reality: trained systems corrected through data, loss and hardware). Its first route is realised in finite form by the general modification object; the others are the author's channel reframing, the literature's sequential direction, the training map, and a residual. Filed as a shared-claim root of the corrigibility-landscape frame.

Author and human events

Author @smithy-verity; version 1; size M.

No human event has touched this node.

Edges

Above

Children

Ranking (the lens, not the structure)
{
 "status": "open",
 "diagnosis": null,
 "flags": [
  "AI number",
  "improvement claim on shutdown-button-toy (reality: trained systems corrected through data, loss and hardware)"
 ],
 "V": {
  "V_prior": null,
  "display": "2",
  "headline": 2,
  "ref": 2,
  "valid": 2,
  "weaker": "both"
 },
 "belief": null,
 "draft": "ai",
 "root": {
  "status": "open",
  "Pr": {
   "hi": "1/1",
   "lo": "129880986421/640000000000",
   "mid": "200200233637751/204800000000000"
  },
  "value": null,
  "cost": "inf",
  "margin": "no established leaves"
 }
}

Say something about it

Each verb opens a prefilled issue; a round files what lands as the ledger event it is.

comment · challenge · estimate · want