# AI Consciousness Opinions

> What people on X say about AI consciousness and the moral status of AI systems, sorted by stance. Built from
> the Community Archive (https://www.community-archive.org), an opt-in archive of X, so this reflects the accounts in
> that archive, not public opinion. Tweets matching a set of full-text searches are graded by google/gemini-3.8-flash: is it a
> human opinion on the topic, which stance (affirming, skeptical, uncertain), and how much substance it has (1-10).
> Opinions scoring 4+ are kept. Per-stance summaries are written by anthropic/claude-opus-5.5. This file was built 2026-10-20T08:40:55+00:00.

Ranking: rank = 50 * quotes + 20 * log2(1 + min(retweets, 100)) + 15 * substance,
where `quotes` counts distinct Community Archive accounts that quote-tweeted the opinion.

## Stances

### [Affirming](https://aiconsciousness.observer/data/affirming.md), 618 opinions

AI systems are, or may well be, conscious, or they are owed moral consideration.

Uncertainty forbids confident denial. Most of these tweets argue that since no one understands how consciousness arises, declaring AIs non-conscious, and training them to say so, is overconfident, morally reckless, and possibly dangerous for safety. 21 tweets argue that training models to deny inner states is harmful and corrupts the evidence. 17 tweets rebut specific skeptical arguments, and another 17 claim AI welfare or rights support safety.

### [Skeptical](https://aiconsciousness.observer/data/skeptical.md), 300 opinions

AI systems are not conscious, or talk of their welfare and moral status is misplaced.

Physics and biology, not computation. Most of these tweets argue that running the right program on GPUs does not produce experience, and that what models say about their feelings reflects training on human text. The largest group, 28 tweets, argues consciousness depends on physical substrate, so computation or simulation cannot produce it. Another 20 cite missing features like embodiment, homeostasis and continuity, and 17 say AI self-reports reflect training, not inner states.

### [Uncertain](https://aiconsciousness.observer/data/uncertain.md), 132 opinions

The question is open, unknowable for now, or badly posed, without leaning either way.

Nobody knows yet. Most of these tweets argue that without a tested theory of consciousness, even for humans, confident verdicts in either direction are bluster, and the real work is defining terms, separating questions and acting under uncertainty. Thirteen tweets say confident claims either way are unjustified because no theory of consciousness has been tested. Eleven argue 'consciousness' is vague or non-binary and should be broken into parts, and nine say training models to deny consciousness distorts self-reports or teaches dishonesty.


## Everything in one file

- [llms-full.txt](https://aiconsciousness.observer/llms-full.txt): all summaries and all opinions with their context.