Skip to content
Bentley MoonAn independent commonsFuturism Institute
Bentley Moon

Futurism Institute

An AI safety researcher who moves where the work is, open to the right opportunity, anywhere.

CollaborateThe researchTheory of changeShort summary

The sky over your part of the world right now, drawn in your browser from the date and time.

I’m developing a set of protocols to unify resilient infrastructure standards online and beyond. I’m looking to work on the development of sustainable land projects and film the journey of established community models into a documentary about our shared future.

The present focus on developing powerful AI minds leads us to a power and citizenship enigma that requires our voices to be adequately formed and heard.

If we assume catastrophic scenarios of AI development, then our authorship of our reality and the AI that influences it could disappear. Now is the opportunity to leverage global networking and translating technology to experiment and refine robust dialogue, democracy and cooperation. Within a rigorous system, global priorities and values can emerge. Defining global citizenship and what civic participation looks like in that paradigm is the first question it can start to answer.

I’m approaching this challenge by developing a digital ecosystem of literacy and connectedness, decentralization where possible and connecting the unified approach through living spaces of culture that I then record into the documentary.

The opening, in my words

I’m developing a set of protocols to unify resilient infrastructure standards online and beyond. I’m looking to work on the development of sustainable land projects and film the journey of established community models into a documentary about our shared future.

The present focus on developing powerful AI minds leads us to a power and citizenship enigma that requires our voices to be adequately formed and heard.

If we assume catastrophic scenarios of AI development, then our authorship of our reality and the AI that influences it could disappear. Now is the opportunity to leverage global networking and translating technology to experiment and refine robust dialogue, democracy and cooperation. Within a rigorous system, global priorities and values can emerge. Defining global citizenship and what civic participation looks like in that paradigm is the first question it can start to answer.

I’m approaching this challenge by developing a digital ecosystem of literacy and connectedness, decentralization where possible and connecting the unified approach through living spaces of culture that I then record into the documentary.

The film, first of three

If Anyone Builds It

I’m making three short films on the argument of If Anyone Builds It, Everyone Dies. This first one runs 2 minutes 53 seconds: the claim, the four things it rests on, and how it can be stopped. The next two are in production.

Where to start

  1. Understand. The governments' scientific International AI Safety Report, plain-language answers at AISafety.info, and a serious sceptical view, AI as Normal Technology.
  2. Learn. Free online courses from BlueDot Impact and Lens Academy, and the directory at AISafety.com.
  3. Read the full argument. The book sets out in writing, in depth, how an artificial superintelligence could overpower humanity, which the film doesn’t show. Its free online resources are a good start.
  4. Join. Groups and events near you at AISafety.com, campaigns such as PauseAI, and the people working on the harms AI does today, such as the AI Incident Database.
  5. Speak up. Tell your representative you support a ban on artificial superintelligence (ASI) until it can be built safely, with the public’s consent. Every national parliament is listed here.

A wider list, from many angles, with resources in other languages: where to start.

Every fact on screen names its source, and the sources page lists them all, with the people working on this. The longer, single-film cut is on the about page. The book is by Eliezer Yudkowsky and Nate Soares, who did not make or review this film.

How the projects add up

Detailed web editionShort summary

Around every step: any AI doing work at any step can be checked by people who did not build it.
A road spirals out from a starting point. Each turn of the road passes the same six steps: notice a need, find out what’s true, learn how, work it out, agree who does what, try it for real. Each turn takes in more people: one person, a group, a place, places together, everyone. A frame around the whole picture says any AI doing the work can be checked from outside; a line under it says every group writes down what it owes its people. Today: one person, the tools exist; a group, built, not used yet; further out, not built yet.ONE PERSONA GROUPA PLACEPLACES TOGETHEREVERYONEstart1. Notice a needMy flat is cold again.Half the building is cold.Every building on the street is cold.Other places have the same problem.The AI running the heating can’t be checked.2. Find out what’s trueAn AI compares fixes, with sources.Every claim about heaters links to its source.Which fixes worked, building by building?Each place’s record is open to read.The records show what happened, with sources.3. Learn howI learn to bleed the radiator.A neighbour shows two teenagers the boiler.A shared workshop teaches repairs.People visit each other’s places.Buyers learn what to ask for.4. Work it outFix it myself, or ask the landlord?The boiler, or the rehearsal room?One heating system for the whole street?Groups in two countries compare answers.A standards body weighs the evidence.5. Agree who does whatMe the radiator, the landlord the boiler.Three people say yes; an AI orders the part.Who runs it, who pays, who can leave.They agree what to share, and what not.A rule: outsiders can check it.6. Try it for realDid it work? I note it down.A month on, the result is written down.A year of running it, written down.What failed stays on the record too.The record shows whether it changed anything.123456

Each turn of the road is the six steps once. After step 6 it goes round again, with more people.

  • Today
  • One person: the tools exist (Conscious Consuming, Dignity Needs, Kosplora, Legible AI)
  • A group: built, not used yet (The Socratic Hearth, The Work Party)
  • Further out: not built yet
  • The two rules (the frame and the line below): written down, no group has adopted them yet

Under every step: every group that takes people in writes down what it owes them, including on the day it asks someone to leave.

The examples are imagined, and follow one cold flat at every size. The steps, the rings and the rules are the Institute’s own, and what exists today is current.

One person goes through the six steps once, from noticing a need to trying a fix.

A group goes round the same six steps, with more people deciding.

A building, a street or a land project goes round them for real.

Several places go round together, and each learns from the others’ records.

At the widest ring, what was learned reaches the people who decide about AI.

Two rules hold the whole way round: any AI doing the work can be checked from outside, and every group writes down what it owes its people.

Today: green is what exists. The colours by the word “Today” say how far each ring has got.

  1. Notice a need. My flat is cold again.
  2. Find out what’s true. An AI compares fixes, with sources.
  3. Learn how. I learn to bleed the radiator.
  4. Work it out. Fix it myself, or ask the landlord?
  5. Agree who does what. Me the radiator, the landlord the boiler.
  6. Try it for real. Did it work? I note it down.

Step by step on the Futurism Institute →

The polycrisis, drawn in layers: crises on the surface, a consensus problem beneath them, and a design problem beneath that.A section through the ground under the sky. On the horizon, a skyline of interlocking rings: the polycrisis. In the layer beneath, arrows pointing every which way: a consensus problem, where our voices have to be formed and heard, through dialogue, democracy and cooperation. In the deepest layer, a drawn grid with three foundations joined by one line: shared standards online and beyond, sustainable land projects, and a documentary of community models. Each sends a shaft up to the surface, and the arrows beside each shaft turn to point the same way.shared standards, online and beyondsustainable land projectsa documentary of community modelsnear a foundation, arrows alignThe polycrisiscrises overlapping, on the surfacea consensus problembeneath it: people pullingdifferent waysour voices, formed and heard;dialogue, democracy, cooperationa design problemand therefore: sharedfoundations, joinedThe polycrisis, drawn in layers: crises on the surface, a consensus problem beneath them, and a design problem beneath that.A section through the ground under the sky. On the horizon, a skyline of interlocking rings: the polycrisis. In the layer beneath, arrows pointing every which way: a consensus problem, where our voices have to be formed and heard, through dialogue, democracy and cooperation. In the deepest layer, a drawn grid with three foundations joined by one line: shared standards online and beyond, sustainable land projects, and a documentary of community models. Each sends a shaft up to the surface, and the arrows beside each shaft turn to point the same way.sharedstandardslandprojectsthedocumentaryThe polycrisiscrises overlapping, on the surfacea consensus problembeneath it: people pulling different waysour voices, formed and heard; dialogue,democracy, cooperationa design problemand therefore: shared foundations, joined

The research

When can people trust one AI to check another, and how small can a trustworthy one be? I test it on one workstation, write each test down before it runs, and keep every correction on the record.

  1. See it happen

    Demonstrations of the research, each with one control you can move and every number traced to its run.

  2. Checking AI

    The wider case for checking AI from outside, argued for a general reader: the seven points a check leans on, with what is measured at each.

  3. The full record

    Every result with its limits, the second line of work (Shoofly), and every correction, newest first.

The work

Research programmes

  • Boundary Intelligence

    Simple mechanisms at system boundaries, tested against far larger internal models.

    Written up · five negative results

  • Project Reinnman

    Failure-aware tools for AI-assisted mathematics, tested against Robin’s inequality.

    Riemann hypothesis open · no proof claimed

Tools

  • Local verification engine

    A local confidence layer that answers with a receipt or refuses with a reason.

    Private alpha · not yet checked from outside

  • Lattice

    A local, OpenAI-compatible gateway with model routing and a containment check.

    Working software

  • lattice-commitsource code

    A test-gated wrapper that stops code-repair loops undoing good fixes.

    Alpha · PyPI

  • Governorsource code

    Sends each request to the cheapest model tier that can handle it.

    Open source · MIT

Latest Five figures on the research pages corrected after an audit of every page.

Where this work needs other people

Every project is free and noncommercial. These are the places they need someone.

The four openingsfuturisminstitute@gmail.com

Reviewers of the civic projects: the Values Commons grant brief is at valuescommons.org/funders.