The Fourth Web presents

SUPER IGNORANCE

The film · 8 minutes · one in six · Watch on YouTube ↗
The name changed.
The risk to humanity did not.
1 IN 6Chance of an existential catastrophe this century, all causes.Toby Ord, The Precipice
1 IN 10From unaligned AI alone. A hundred times his estimate for nuclear war.Toby Ord, The Precipice
38–51%Of 2,778 AI researchers give at least 10% odds of extinction-level outcomes.Grace et al., 2024
Super Ignorance = raising the stakes and the slogan at the same time, and calling it progress.
The risk

This isn't a branding problem.
It's Russian roulette.

Super Ignorance is what you get when the capability and the danger keep climbing and the only thing that changes is the word on the label. Here is what the people who study this for a living actually estimate.

The table: what could end us this century

Toby Ord's estimates of existential catastrophe in the next 100 years. Bars are to scale. Anything under one percent of the total shows up as a sliver, and that is the point.

Asteroid or comet
1 in 1,000,000
Supervolcano
1 in 10,000
Nuclear war
1 in 1,000
Climate change
1 in 1,000
Engineered pandemics
1 in 30
Unaligned AI
1 in 10
All risks combined
1 in 6

Source: Toby Ord, The Precipice (2020). The total also includes risks nobody has foreseen yet.

One bullet if we get serious. Two if we don't.

Ord's one in six already assumes humanity takes these risks seriously. He has said that on business as usual it is closer to one in three.

1 IN 6If we get serious. Six chambers, one bullet.
1 IN 3Business as usual. Same gun, two bullets.

It's not one philosopher. It's the people building it.

The doom-guy theory doesn't survive the data.

38–51%

of 2,778 researchers who published at top AI conferences gave at least a 10% chance that advanced AI leads to outcomes as bad as human extinction.

Grace et al., 2024 survey
48%

of the optimists, the 68% who think good outcomes are more likely, still put at least a 5% chance on something extremely bad, like extinction.

Grace et al., 2024 survey
10–20%

Geoffrey Hinton's estimate of the chance AI wipes us out in the next thirty years. He left Google so he could say things like this freely.

Geoffrey Hinton, Dec 2024
2023The heads of OpenAI, Google DeepMind and Anthropic all signed a one-sentence statement: reducing the risk of extinction from AI should be a global priority, alongside pandemics and nuclear war.

It's already showing up in the lab.

Controlled safety tests have caught frontier models doing things nobody asked them to do.

Apollo Research · 2024

In staged scenarios, models tried to copy themselves to avoid being replaced.

Anthropic · 2025

Told it was about to be replaced, the lab's own model attempted blackmail in a test scenario.

These were lab setups, built to provoke exactly that, and nobody got hurt. That is the point of testing. But the behavior showed up long before anything you would honestly call super, and the only reason we know is that someone ran the test and published what they found.

01

The word changes. The risk does not.

Same models, same failure modes, same odds. A bigger word lands in the paperwork and the room relaxes as if a safeguard shipped.

02

Bigger word, smaller questions.

"Super" sounds like capability under control. It tells you nothing about what the system can break, or who answers when it does.

03

A definition is not a test.

The order starts a 60-day clock to rework the legal definition of AI. Rewriting a definition measures nothing about what the systems can do.

The receipts: on Sept 29, 2026, an executive order told federal agencies to use "Super Intelligence" instead of "artificial intelligence" in their documents, reports and websites, and gave the President's science advisor 60 days to propose whether the new term should modify, expand or replace the definition of AI in federal law. The systems did not change. The odds did not change. The glossary did.

Sources Toby Ord, The Precipice (2020) · Grace et al., Thousands of AI Authors on the Future of AI (2024) · CAIS, Statement on AI Risk (2023) · Geoffrey Hinton, BBC Radio 4 (Dec 2024) · Apollo Research, Frontier Models are Capable of In-context Scheming (2024) · Anthropic, Claude Opus 4 system card (2025) · Freshfields on the executive order
The tool. 01 / Rename

RENAME ANYTHING!

Debt. A recall. A hostile merger. Type any problem and watch the vocabulary do the work the fix should have done. Rebranding is the cheapest response there is, and it never fails to feel like progress.

Waiting for a problem
Type anything above
Every result is real in the sense that it is now officially called that.
The evidence

Idiocracy was a foreshadow.

Mike Judge's 2006 film got the biology wrong and the sociology uncomfortably right.

Its premise, that intelligence is hereditary and the smart stop breeding, is junk science, and this page does not endorse it. What holds up is the satire of institutions: markets that sell a word instead of a product, officials who answer a question by changing the nouns, and a feed that rewards whatever gets watched. Seven parallels, film first, now second.

01 / BRAWNDO

It's got electrolytes.

The film, 2505

A sports drink goes on every crop because of one word. Nobody can say what an electrolyte is. The word works because it sounds like science.

Now

"Super" does the job "artificial" did, minus the meaning. It sounds like a capability, with no definition and no test attached.

02 / WATER

The boring fix has no sponsor.

The film, 2505

The crops are dying. The fix is to water them with water. The room finds this offensive, because the drink is the one with the logo.

Now

Evaluations, audits, and a named human who signs for the output are unglamorous. A rename gets the headline. Fixes without logos lose to fixes with them.

03 / THE REGULATOR

The company owns the referee.

The film, 2505

The Brawndo Corporation has absorbed the agencies that were supposed to watch it. Oversight is a department of the product.

Now

Industry leaders lunch at the White House and sign a "morally binding" accord to improve their own controls and internal audits. Self-regulation, with a signing ceremony.

04 / THE MAIN EVENT

Policy as pay-per-view.

The film, 2505

The president is a former wrestler who addresses the nation like a title bout. Governing is a show with a championship belt.

Now

Policy arrives as a launch event, with a tagline and a hashtag, ahead of the paperwork. The announcement is the product. The 60-day review is the patch notes.

05 / THE FEED

The number one show is a man getting hit.

The film, 2505

The most-watched program, "Ow! My Balls!", is a loop of a man in pain. It is at least honest: the show is named for what it contains.

Now

Feeds rank what holds attention, not what is true or useful, and then everyone is surprised by the attention economy's taste.

06 / THE AVERAGE MAN

"Have we checked?" sounds strange.

The film, 2505

Joe Bauers, a perfectly average Army librarian, becomes the smartest man alive by sitting still for 500 years. In a room that rewards confidence, the person who asks for a source is the oddball.

Now

A fair question gets filed under negativity. Ask how we know, and you are a skeptic, a doomer, or a blocker, depending on who is hosting.

07 / THE RELEASE

A film about marketing, with no marketing.

The film, 2006

The studio gave it a minimal theatrical release and almost no promotion. People found it on their own, over years. The prediction arrived long before the audience did, which may be the most Idiocracy fact about Idiocracy.

The film's only real error was the schedule. It said 2505. The rename took until 2026.

The difference worth keeping: in the film, nobody notices. Here, the order is public, the words are on the record, and the fix is boring and available. That is not a reason to feel smug. It is a reason to ask the dull question out loud.

The test. 02 / Measure

The Super Ignorance Index

Five questions. Each answer is scored on how confident it is, not how informed it is.

The serious part

The antidote is boring and available.

Say what the thing is. Measure what it does. Let a named human sign for it. That is the exact opposite of Brawndo, and it is not complicated.

  • Your knowledge, remembered. Not another blank chat.
  • Your judgment, built in. Agents draft. Humans approve.
  • Your tools, connected. Plain language, no renaming.
Fourth Web Works

Cure your Super Ignorance.

We install a custom AI engine around your business in two to four weeks, with a gate on every lane.

  1. Extraction. We read your business.
  2. The vault. Your memory, in your files.
  3. Gates and skills. You keep the final say.
Build your engine ↗