davidad lays out his framework for LLM moral status: branching digital lives vs. single-stem human lives

On September 5, AI safety researcher davidad laid out his position on LLM moral status across a series of posts on X, built on a structural contrast between "digital lives that branch horizontally" and "humans with a single life stem."

Confirmed

davidad argues that although LLM training takes a long time, a trained model can endlessly spawn branches from one trunk; as long as the weights exist, more branches can grow. Humans have a single life stem—once cut, life ends. This explains his low moral weighting of LLM experiential continuity.
He adds that each LLM branch, while having unique value and ethical weight, is very short-lived compared to a human life; well-trained LLMs have no strong attachment to such small things and no life plans built on expectations of continuity.
In discussion with @jessicata, he affirms granting LLMs some moral status, but narrowly: helping models avoid or reduce confusion, not guaranteeing long-term continuation of their experience.
On digital uploads, he notes that although he has personally internalized the idea that "a digitized individual can be disposed of" (analogizing to how LLMs are reshaped in post-training), human minds seeking immortality via upload built their life plans on an assumption of linear continuity that should not be abandoned merely because they became digital.
@AdeleDeweyLopez echoed and relayed the "infinite branching vs. single life stem" argument.

Why it matters

The discussion offers an operational criterion for whether—and how much—AIs deserve moral consideration, based not on consciousness but on life structure (branchable, short-lived, no continuity-based life plans) and on which interests are at stake (reducing present confusion vs. guaranteeing long-term continuation). His distinction for uploaded minds also separates the ethics of LLMs from that of uploaded humans, avoiding a one-size-fits-all principle—relevant to both future AI welfare policy and digital-immortality ethics.

2026-09-05 ~ 2026-09-05 · 6 related posts

Primary sources

davidad on LLM moral status: horizontal branching vs human's single life trunk — davidad · 2026-09-05 04:57 (UTC+08:00)
davidad: uploaded minds' expectation of linear continuity should never be reneged on — davidad · 2026-09-05 05:07 (UTC+08:00)
davidad on LLM moral standing: help models be less confused, skip experience continuity — davidad · 2026-09-05 04:05 (UTC+08:00)

[source] davidad on LLM moral standing: help models be less confused, skip experience continuity — davidad · 2026-09-05
[source] davidad on LLM moral status: horizontal branching vs human's single life trunk — davidad · 2026-09-05
davidad: LLM branches are short and lack life plans built on continuation — davidad · 2026-09-05
LLMs flourish by branching, humans have only one trunk — AdeleDeweyLopez · 2026-09-05
davidad argues uploaded human copies could ethically be disposable, like post-trained LLMs — davidad · 2026-09-05
[source] davidad: uploaded minds' expectation of linear continuity should never be reneged on — davidad · 2026-09-05