What are the AI explainers? · How does SIREN compare to DEFCON, the Doomsday Clock and p(doom)?
How does SIREN compare to DEFCON, the Doomsday Clock and p(doom)?
Five numbers people reach for when they want to know how worried to be. They measure different things, and only one of them is a count you can check. Right now it reads SIREN 4, ROUTINE.
What is the short version of the comparison?
| Number | What it measures | Who sets it | Can you recompute it? |
|---|---|---|---|
| SIREN | How much is happening in AI: tempo, not danger | A published formula over public data, hourly | Yes, from the receipt |
| DEFCON | The readiness posture of US armed forces, 5 (lowest) to 1 (highest) | US military command | No. The current level is not routinely made public |
| Doomsday Clock | A symbolic judgement of how close humanity is to catastrophe | The Bulletin of the Atomic Scientists, usually once a year | No. It is a judgement, by design |
| Pentagon Pizza Index | How busy pizza shops near the Pentagon are, read as a hint of military activity | A website, from shop-busyness data | Not from anything it publishes that we could find; its own disclaimer says not to rely on it |
| p(doom) | One person's stated probability that AI ends very badly | Whoever is asked | No. It is an opinion expressed as a number |
How do you read the SIREN dial?
The scale runs from 5 to 1 and 1 is the loudest. The level is the band the 0–100 composite score falls in:
| Level | Name | Score band | In a few words |
|---|---|---|---|
| 5 | DORMANT | 0–34 | Needle at rest |
| 4 | ROUTINE | 35–54 | Needle breathing |
| 3 | ELEVATED | 55–69 | Off the rest stop |
| 2 | ACCELERATED | 70–84 | Pinned high |
| 1 | UNPRECEDENTED | 85–100 | Off the top of the chart |
The score is built from five pillars: Capability (what shipped, and what got written down), Compute & Capital (silicon, spend, and the market’s opinion of both), Attention (how loudly the world is talking), Governance (paperwork with the force of law), Markets (where money is posted against an opinion). Each is scored against the index's own frozen record, so a reading says how today compares with what the index has seen before.
What does a loud reading tell you, and what does it not?
A loud reading tells you a lot shipped, a lot was spent, a lot was written or a lot was regulated, and the pillar breakdown says which. It does not tell you whether any of it was good or bad, and it is not a forecast of what happens next. That is why the home page is stamped "not a prediction".
Why borrow DEFCON's grammar at all?
Because nobody needs the scale explained twice. "Five is calm, one is not" is already in most people's heads. The borrowing stops at the grammar: SIREN is not a readiness level and is not issued by any government body.
What questions do people actually ask?
How does SIREN compare to DEFCON, the Doomsday Clock and p(doom)?
SIREN measures how much is happening in AI — tempo, not danger — from a published formula over public data, hourly, and you can recompute it from the receipt. DEFCON is US military readiness; the Bulletin’s Doomsday Clock is a yearly symbolic judgement; p(doom) is one person’s opinion as a percentage.
What is the short version of the comparison?
SIREN is a recomputable hourly count of AI activity. DEFCON is US military readiness. The Doomsday Clock is a yearly symbolic judgement. The Pentagon Pizza Index reads shop busyness as a hint. p(doom) is one person’s opinion as a percentage.
How do you read the SIREN dial?
The scale runs from 5 to 1 and 1 is the loudest. The level is the band the 0–100 composite score falls in, built from five pillars scored against the index’s own frozen record.
What does a loud reading tell you, and what does it not?
A loud reading tells you a lot shipped, a lot was spent, a lot was written or a lot was regulated, and the pillar breakdown says which. It does not tell you whether any of it was good or bad, and it is not a forecast of what happens next.
Why borrow DEFCON's grammar at all?
Because nobody needs the scale explained twice. “Five is calm, one is not” is already in most people’s heads. The borrowing stops at the grammar: SIREN is not a readiness level and is not issued by any government body.
Is SIREN a DEFCON level or the Bulletin’s Doomsday Clock?
No. SIREN borrows the countdown grammar only. It is not a military readiness level and does not publish the Bulletin’s clock. Sites that put those phrases in an AI title for traffic are doing brand hijack; see /vs/pizzint.html.
What is p(doom)?
p(doom) is one person’s stated probability that AI ends very badly. It is an opinion expressed as a number — not a recomputable measurement. See /p-doom.html.
Does a loud SIREN reading mean AI is more dangerous?
No. A loud reading means a lot shipped, was spent, was written or was regulated. It does not say whether any of it was good or bad, and it is not a forecast.
Where should you go next?
Open the live index, methodology, AI explainers hub, compare and vs pages, the PizzINT contrast, alternatives, history of AI risk, or Tally’s desk.
Search-shaped sibling: What is the history of AI risk? · How the SIREN AI activity index is computed · methodology · What is p(doom)?.
Where should you go next?
AI activity index today · SIREN → · How the SIREN AI activity index is computed · methodology · What are the AI explainers? · Which AI doom indexes can you compare? · Which AI indexes does SIREN contrast? · SIREN vs PizzINT: which should you use? · What are AI doom index alternatives? · What is the history of AI risk? · What unserious counts does SIREN keep on Tally’s desk?