Two Maps, Four Questions

Spend ten minutes in any AI comment section and you will get sorted into a team. Doomer. Luddite. Hopepilled. Tech spiritualist. The labels are fast, and they often miss most of what the actual person thinks.

A YouTuber named Elodine has a better idea. In The 4 Hidden Factions Controlling the Future of AI she sets the labels aside and uses two simple graphs instead. Each graph has two axes, so there are four questions in all. Answer them and you get a position, not a team jersey.

Axis: a line on a graph that measures one thing, with one extreme at each end. Two axes that cross split a graph into four sections called quadrants.

I liked it enough to build a quiz for it. Take the Two Maps, Four Questions quiz (it opens in a new tab). It is sixteen statements, and when you finish it plots your dot on both maps. Read on first if you want to know what the maps mean.

Why the labels fail

Elodine’s starting point is that AI words are “loosey-goosey.” Some come from companies, some from online groups, and some come from frustration. A group does not always get to pick its own label. Luddite is the sharpest example.

Luddite: the original Luddites were English textile workers who, from 1811 to 1816, smashed machines they believed were destroying their jobs and wages. Today the word is often used as an insult for anyone who pushes back on new technology.

Her worry is a thing writers call flanderization.

Flanderization: when one small trait of a character gets exaggerated over and over until it takes over the character. The name comes from Ned Flanders on The Simpsons, who started as a nice, churchgoing neighbor and slowly became mostly about his religion.

Labels do the same thing to real people. A person with ten opinions about AI gets boiled down to one talking point. Elodine is upfront that she is a writer, that she is biased, and that her view is a Western one. She is not claiming to be neutral. She is claiming the map holds more than a label does.

Map 1: Where should AI go?

The first map is about the future of the technology.

The horizontal axis is direction. It runs from Restrain on the left to Accelerate on the right. Elodine names three camps along it:

  • Accelerationist: move AI forward despite any potential cost.
  • Preservationist: the technology needs more time to mature.
  • Revivalist or abolitionist: walk the technology back. (She admits she struggled to name this one.)

The vertical axis is transparency, which she calls clock privilege. The top is Concentrated and opaque: a few well-funded labs working in secret. The bottom is Distributed and transparent: openly shared top models, released weights, and documents that tell the public what a new technology can do.

Model weights: the numbers a trained AI model has learned. Releasing them, called “open weights,” makes it much easier for other people to run, study, or change the model.

The name comes from what she calls the clock problem. Her definition: companies shape what the public knows about a new technology, showing some things and hiding others, to sway investors, the public, and regulators. That delay can buy them time to make money before stricter laws or lawsuits catch up.

She links this to the Collingridge dilemma, an idea from David Collingridge’s 1980 book The Social Control of Technology.

Collingridge dilemma: early on, a new technology is easy to control, but its effects are hard to predict. By the time the effects are clear, the technology is so built into everyday life that it is hard to change.

In the classic version, that gap is just bad timing. Elodine’s version adds a choice. She describes companies that “manipulate markets through controlling publicly available information” to delay legal action. In other words, a company that controls the information can stretch out the early “nobody knows yet” period. That is the clock, and whoever holds it has the privilege.

Cross the two axes and you get four quadrants:

  Restrain Accelerate
Concentrated and opaque Guarded Stewardship Incumbent Acceleration
Distributed and transparent Democratic Precaution Open Acceleration

In plain words:

  • Guarded Stewardship: go slow, and let a few careful hands hold the keys.
  • Incumbent Acceleration: go fast, led by the big companies already on top. (An incumbent is whoever already holds the top spot.)
  • Democratic Precaution: go slow, and let the public decide how.
  • Open Acceleration: go fast, out in the open.

Elodine puts herself in Open Acceleration, but close to the middle. She wants the technology to move forward and wants it done in the open. She is not pushing for breakneck speed, and she says she would not be upset if progress pulled back. Her non-negotiable is the clock problem. “I hate security theater. I hate censorship,” she says.

Security theater: safety steps that look impressive but do little to make anyone safer.

She does not believe responsible scaling, meaning growing AI carefully and safely, happens behind closed doors. She also does not think companies should be trusted to do it without input from the public.

She mentions something she left off the map on purpose: local models trained on data from volunteers who get paid for it. That approach tries to solve many of the ethical complaints about AI. A viewer pointed out why it is not enough for people on the restrain side of the map. If a few companies still control the market and the information, they can still drive a race to the bottom, where everyone cuts corners to keep up. Fair training data does not change who holds the clock.

Map 2: Is AI a someone?

The second map is about what AI is.

The horizontal axis asks whether AI could ever be someone instead of something. It runs from A tool on the left to Someone on the right. This is about the long run, not just today.

The vertical axis asks whether today’s AI has cleared the bar. The top is AI clears the bar: a mind. The bottom is AI is mindless: autocomplete.

Sentience: the ability to feel things or have experiences. It is the “bar” most people mean when they ask if an AI is really a mind.

Elodine says those two questions cover “everything from stochastic parrots to waifus.” That is, from people who see AI as a word machine to people who treat a chatbot like a romantic partner.

Stochastic parrot: a phrase from a 2021 paper by Emily Bender, Timnit Gebru, and coauthors. It describes a language model as something that strings words together based on patterns, without understanding what they mean.

The four quadrants:

  A tool Someone
Clears the bar (a mind) Cold Conviction Wholehearted Personhood
Mindless (autocomplete) Wholehearted Tool The Pack-Bond

In plain words:

  • Cold Conviction: today’s AI may be a mind, but it is still a thing.
  • Wholehearted Personhood: AI could be someone, and may already be.
  • Wholehearted Tool: AI is a tool, now and forever.
  • The Pack-Bond: today’s AI is not a mind, but it is not only a tool either.

She introduces this map with a joke: “sentient beings love a good pack bond.” Her example is people who felt a strong connection to chatbots like Microsoft’s Tay while knowing how the technology worked. Tay launched on Twitter on March 23, 2016. Users manipulated it into posting hateful messages, and Microsoft shut it down about 16 hours later.

Those people, she says, “might say that it feels like a tool, but they’re empathetic towards it while knowing it doesn’t clear the bar for sentience.” She places them in the center of the lower half of the map, right on the line between Wholehearted Tool and The Pack-Bond.

Elodine sits a bit further right, in The Pack-Bond. She knows AI is a tool, and today’s models do not clear the bar of sentience for her. But she cannot call it a tool alone. She has big problems with having to bully a model into doing tasks it shouldn’t have to comply with. She is careful to say that is her personal stance, not one she asks anyone else to adopt.

What the maps show

Put both maps together and you have four questions:

  1. Where do you want AI to go?
  2. How transparent should companies be along the way?
  3. Could AI ever be a someone versus a something?
  4. Is it there now?

Elodine noticed that her dot sat in almost the same spot on both maps, the lower right. She is curious whether other people see a pattern between their two positions, too.

Her bigger point is about the arguments. When she reads arguments from anti-AI folks, she often sees something more specific underneath: people who detest clock privilege. Their complaint is often a disclosure problem with new technology, not an AI problem. A surprising number of pro-AI people could find common ground with them on that one question, even if they disagree on everything else.

Calling it “anti-AI” buries that common ground. She suggests that might be the silver lining for Big Tech, but “definitely not for the public.”

Why I like this

I think the second axis on Map 1 is the most useful idea in the video. Speed and openness get lumped together all the time, but they are different questions. You can want AI to go fast and still demand that it be done in the open. You can want it slowed down and still trust a few careful labs to do the slowing. Most online fights treat those as one dial. They are two.

The quiz is my attempt to make the framework something you can try instead of just read about. The sixteen statements and the scoring are mine, not Elodine’s, so any bad question is my fault. Your answers never leave your browser. When you are done you can copy a link that holds your four scores, not your answers, and compare your dot with where Elodine put hers.

Take the quiz →

Sources

Written on October 11, 2026