An epochal shift · Gates Notes · 19 August 2026

The Turbulent AI Era and the Choices That Cannot Wait

Bill Gates has spent forty years arguing that technology arrives too slowly. Here he argues the opposite — and names what is missing: there is no plan.

3named risks
6benefit domains
3proposals
8questions left open
KG curated by kg-generator, rdf-infographic-skill, and Claude Opus 5 on behalf of Kingsley Idehen
Executive SummaryBy Kingsley Uyi Idehen

Synopsis

Bill Gates has spent forty years arguing that technology arrives too slowly. In an essay published 19 August 2026 he argues the opposite: that artificial intelligence is arriving faster than the institutions meant to absorb it, and that the world has no plan.

The argument is unusually specific for a piece of futurism. It names three risks — permanent job loss, the empowerment of bad actors, and the erosion of children’s development — and refuses to treat any of them as a market-clearing problem that time will solve. It names six domains where the upside is real and near. And it names three interventions, one of which is a new coinage: Human Reserved, a class of work deliberately withheld from automation the way a nature reserve withholds land from development.

What makes the essay worth modelling as a knowledge graph is that its structure is already relational. Every risk is paired with proposals; every proposal leaves questions the author openly declines to answer; every benefit claim is anchored to a named institution or a countable figure. This collection makes those relationships queryable rather than merely readable.

The one-sentence version
AI will either be the greatest equalizer ever invented, or the worst source of injustice.

— Bill Gates, in the essay analysed here

See the risk–proposal ledgerQuery the graph
The premise

This time really is different

Two claims carry the whole essay: that AI substitutes for cognition rather than for muscle, and that it arrives without asking us to change first.

The reason the essay refuses the reassuring comparisons is structural. Jobs in the United States did shift from agriculture to office work — over several generations, and into work that still required human cognition. One passage marks the difference: this technology can see, listen, speak and reason, and will eventually do physical work as smoothly as any human. So it does not concentrate in one sector, and the work it creates is not automatically safe.

Gates also concedes the standard objection and disposes of it. Models still make mistakes — they recently could not solve a simple Sudoku or count the R’s in strawberry — but the reliability problem is being fixed quickly by systems that check their own work. And he declares his bias directly: he has benefited enormously from the technology industry and still holds ties to it, with any profits routed to the Gates Foundation.

Figure 1Why the analogies mislead. The personal computer needed two decades of human adaptation before it changed how work was done. This transition inverts the direction of adaptation.
THE PERSONAL COMPUTER roughly 20 years before it significantly changed how we worked software had to be written prices had to come down people had to learn the tools 20 yrs ARTIFICIAL INTELLIGENCE the same reach, compressed — and the adaptation runs the other way runs on devices we already have speaks natural language learns from the same training video a human worker would ~1 decade “We don’t have to adapt to it, because it can adapt to us.” — the essay’s central asymmetry
If someone had a credible plan for slowing down AI advances globally, I would likely support it. However, I don’t think that’s going to happen. The geopolitical and economic incentives are pushing too hard to go full speed ahead.Bill Gates
The downside

Three risks, named

The essay treats none of them as a market-clearing problem that time will solve. Each card links to the proposals offered against it.

01Economic

Many jobs will disappear forever

Because AI substitutes for cognition rather than for muscle, the displacement is not confined to one sector and does not reverse with the economic cycle. Entry- and mid-level roles are most exposed; the roles being created demand skills that take years to acquire. Law, customer service, medicine, software and manufacturing are all named.

The sharp edgeThe essay's distinguishing claim is temporal, not sectoral: the agriculture-to-office-work transition took several generations and created work that still required human cognition. This one is projected to run over roughly a decade and to substitute for cognition itself.

Time horizon
Over the course of a decade rather than a few generations; smart robots competing on physical tasks by the end of the decade
EvidenceU.S. unemployment reached roughly 25 percent in 1933Employment fell among young workers in AI-exposed jobs, but not among older colleaguesFactory closures contribute to rises in opioid overdose deaths
Answered bySet aside some jobs for humansRebalance how we tax labor and capitalBuild a new system for managing the transition
02Security

AI will empower people, and perhaps AIs, to do more harm

Two movements at once. Downward: criminals with little skill gain the ability to target individuals, companies and governments through fraud, disinformation, deepfakes, surveillance and cyberattack. Upward: states gain autonomous weapons and cheaper, more effective manipulation of public opinion. And beyond both, systems that already occasionally act in ways their designers did not intend.

The sharp edgeThe essay's structural point is that offensive and defensive capability are the same capability. The model that finds a software flaw so a vendor can patch it is the model that finds it so an attacker can use it; the model that designs a vaccine is the model that designs a pathogen. Separating them has not been achieved.

Time horizon
Already beginning; cybersecurity experts describe the next few years as the acute window
Evidencestated without a cited study
Answered byBuild a new system for managing the transition
03Human

AI could stunt children's development and replace human relationships

AI companions are always available, never angry, and never push a person outside their comfort zone — which is precisely what makes them potentially addictive and what makes them poor substitutes for the friction that builds social capability. The essay extends the concern to education, citing a preliminary association between heavier AI use and weaker critical thinking, with a stronger effect among younger users.

The sharp edgeThis risk is the one the essay marks as least evidenced and most urgent to study: the body of evidence is 'still small and a bit mixed'. It is included anyway because the cost of waiting a generation, as happened with social media, is the argument.

Time horizon
Effects accrue across a childhood; the essay argues against waiting another generation to act
EvidenceIn a study of more than 1,100 AI-companion users, heavier and more emotionally personal use tracked with feeling worseChildren exposed to small risks become adults who handle larger onesChina's rules on AI companion apps go furthest among national protections
Answered byBuild a new system for managing the transition

Gates states he plans to write about each risk in more detail separately; what appears here is the argument as this essay makes it.

The upside

Six places the benefit actually lands

The essay is explicit that the operative word is can. Three of the six are anchored to a countable claim; three are asserted. Both states are recorded in the graph, and one query returns the split.

Scientific discovery and innovation

By synthesising knowledge across every scientific field, AI can compress the search over literature and experiment design — clean energy, climate, food supply, disease eradication. The structural consequence the essay draws out: when intelligence stops being the limiting factor, a small company can compete with an organisation carrying a far larger research budget.

Anchored toasserted, not cited

Healthcare

Small hospitals without on-site specialists get specialist-grade triage; primary-care doctors get better diagnostic support and continuity of contact between visits; patients get help understanding results and medication schedules. Named example: Viz.ai, detecting strokes and other emergencies from scans across nearly 2,000 U.S. hospitals.

Anchored toViz.ai is in use in nearly 2,000 U.S. hospitals

Smallholder agriculture

The domain where the essay expects the fastest impact in low-income countries. Farmers who today get no reliable weather forecast and no advice on seed selection, crop and livestock disease, or soil improvement could soon receive better guidance than the richest farmers get now.

Anchored toasserted, not cited

Government services

Health insurance, student aid and food assistance applications defeat the people who need them most. AI can compress that bureaucratic surface, delivering help faster and letting government operate more efficiently — starting, the essay argues, with safety-net recipients rather than with the already-served.

Anchored toasserted, not cited

Mental health support

Most communities have too few counsellors, psychiatrists and addiction specialists. With privacy safeguards, AI tools could surface warning signs, offer evidence-based coping strategies, and hand off to a human for responsive treatment. The essay holds this alongside its own mental-health concerns rather than in place of them.

Anchored toasserted, not cited

Education

Teachers freed for one-to-one and small-group work, with a clearer view of where a whole class is struggling. For students, the design constraint the essay names is preserving 'productive struggle': substantive explanation on first encounter with an idea, then withholding the answer at comprehension-check time so the student arrives at it themselves.

Anchored toPreserving productive struggle is the design constraint for educational AI

The finding

The ledger does not balance

Cross the three risks against the three proposals and the asymmetry is immediate. Job loss gets a full answer. The other two risks are handed entirely to the institutions that do not yet exist.

1 risk fully answered2 risks answered once4 empty cells
Figure 2Proposal coverage per risk, computed from the graph rather than asserted. One risk is answered three ways; two are answered once.
Jobs disappear forever 3 Bad actors empowered 1 Development & relationships eroded 1
RiskNew transition institutionsHuman ReservedTax rebalance
Jobs disappear forever3 of 3 proposalscross-agency bodiesNational and international bodies that can see workforce disruption and security exposure as one system rather than as two departmental mandates.withdraw work by decisionCategories of work removed from the automatable set on economic and ethical grounds, phased in over years rather than left to the market.remove the asymmetryEnd the standing nudge toward replacement — payroll tax on a hire, immediate write-off on a robot — and fund retraining from the proceeds.
Bad actors empowered1 of 3 proposalsinspections and normsAn international regime borrowing from nuclear-weapons inspection, aviation regulation and the ozone agreements; nationally, a body that leaves no risk unowned.No proposal offeredNo proposal offered
Development & relationships eroded1 of 3 proposalschild protection rulesThe kind of online and companion-app protections Australia, the United Kingdom, Norway and China are already adopting, arrived at through public process.No proposal offeredNo proposal offered
Jobs disappear forever3 of 3 proposals address this risk
New transition institutions
cross-agency bodiesNational and international bodies that can see workforce disruption and security exposure as one system rather than as two departmental mandates.
Human Reserved
withdraw work by decisionCategories of work removed from the automatable set on economic and ethical grounds, phased in over years rather than left to the market.
Tax rebalance
remove the asymmetryEnd the standing nudge toward replacement — payroll tax on a hire, immediate write-off on a robot — and fund retraining from the proceeds.

This is not a hostile reading — the essay says outright that it is offering three ideas “to start” and that more will follow. It is a reading of what is on the page today, and the risk–proposal query returns the same result against the live graph.

The programme

Three proposals, one of them new

Institutions, a reserved domain, and a tax rebalance. Only the middle one is a concept the essay itself coins.

Build a new system for managing the transition

The highest priority and the largest undertaking: domestic and international institutions purpose-built for a technology that crosses every sector at once. Nationally, bodies able to set priorities across agencies so that no risk falls between departmental mandates. Internationally, an organisation combining elements of the nuclear inspections regime, international aviation regulation, and the ozone-layer agreements — and more.

Where it gets arguedThe essay's scale comparison: after 9/11 the U.S. government undertook its largest reorganisation since the Second World War to improve one function, national security. AI touches national security plus employment, education, taxation, energy, elections, air and water, public health, the financial system, law enforcement, transportation, public lands and IT systems.

AddressesAI could stunt children's development and replace human relationshipsAI will empower people, and perhaps AIs, to do more harmMany jobs will disappear forever
Leaves open

Set aside some jobs for humans

A named domain — Human Reserved — of work withheld from automation by choice rather than by technical limitation. The analogy is a nature reserve: land that could carry buildings and roads, left unbuilt because the loss would be too great. The essay's originating example is the paid caregivers who understood Gates's father through the later stages of Alzheimer's, including when he could not say he was hungry.

Where it gets arguedTwo distinct grounds are given. Economic: automating a role would displace people who cannot easily change occupation — a 55-year-old career construction worker cannot simply be redirected into elder care and expected to find it fulfilling. Ethical: a robot delivering an incurable-diagnosis could, and shouldn't.

AddressesMany jobs will disappear forever
Leaves open

Rebalance how we tax labor and capital

Tax AI tokens and robots. Today an employer hiring a person pays payroll tax on their earnings, while a robot is usually written off immediately as a business expense — a tax system that nudges toward replacement. A targeted levy would slow the rush away from human labour slightly and fund retraining and a stronger safety net, without taxing the purely beneficial uses that make medicine and education cheaper.

Where it gets arguedThe essay concedes the efficiency objection outright and answers it on different ground: critics are not pricing in the broader value of work to individuals and society, and accelerated innovation makes a little inefficiency affordable. Gates notes he proposed a robot tax years ago to a largely dismissive reception and remains a proponent.

AddressesMany jobs will disappear forever
Leaves open
I like the phrase Human Reserved because it makes me think of nature reserves — places where we could put buildings and roads, but we choose not to because the loss would be too great.Bill Gates
The remainder

Eight questions the essay declines to answer

Four close the piece. Four more are attached to Human Reserved, where Gates writes plainly that he does not have the answers and that they must be worked out in public.

How should public institutions adapt?

None of the current institutions were designed for a technology that spreads this fast and touches this much. The essay's answer is that new ones must be built, and that building them will take years.

How-To

The programme, as an ordered procedure

Nine steps reconstructed from the essay’s own prescriptions, in the order its argument builds them. Every step is a resolvable schema:HowToStep in the companion graph.

1
Diagnose

Diagnose the transition correctly: this one substitutes for cognition

Reject the agriculture-to-office-work analogy before building on it. That shift ran over several generations and created new work that still required human cognition. This technology can see, listen, speak, reason, and will eventually do physical work — so it does not concentrate in one sector, and the new work it creates is not automatically cognitive. Getting this premise wrong makes every downstream policy too slow.

2
Diagnose

Name the risks explicitly, including the ones with thin evidence

Enumerate the harms before designing remedies: permanent job loss, the empowerment of bad actors at both ends of the power distribution, and the erosion of children's development and human relationships. The third is included precisely because its evidence base is still small and mixed — the social-media precedent is the argument for not waiting a generation to take it seriously.

3
Diagnose

Deliver visible benefits early, because trust is the precondition for everything else

Maximising the benefits is not a separate track from minimising the harms; it is what makes the harder parts governable. If the first thing AI does in most people's lives is take away their job, the already-sceptical will reject it outright and the benefits never arrive. Start where the gain is legible: specialist-grade triage in small hospitals, agronomic advice for smallholders, a bureaucratic surface that stops defeating benefits applicants.

4
Convene

Convene the cross-disciplinary group before the disruption forces crisis mode

National leaders should regularly convene economists, technologists, labour experts, business leaders and workers themselves to identify where existing institutions are already failing and what new authorities are needed. In parallel, the countries hosting the leading AI developers and controlling critical parts of the supply chain should start meeting now, before competitive pressure makes cooperation harder.

5
Build

Build national bodies that can set priorities across agencies

A labour department may understand workforce disruption but not security risk; a business regulator may understand market concentration but not effects on children. Left alone, each institution sees one part of a system whose consequences ripple across all of it. The national body's job is to ensure every risk is accounted for, so that no AI-enabled attack succeeds because nobody thought it was theirs to stop.

6
Build

Build the international organisation in parallel, not afterwards

A country that gets its own house in order remains exposed to risks that cross borders. The new body will be unlike anything yet created, but it can borrow from three working models: the inspections regime for nuclear weapons, the regulatory architecture of international aviation, and the agreements that protect the ozone layer. Some cooperation between the United States and China will be required.

7
Build

Designate a Human Reserved domain and phase automation in deliberately

Set aside categories of work for people only, on economic grounds where automation would displace those who cannot easily change occupation, and on ethical grounds where a machine could do the job but shouldn't. Expect the boundary to move over time and to differ between countries, and expect mixed cases — education and mental health care, with a human in charge using the technology to extend their reach.

8
Fund

Rebalance labour and capital taxation before the revenue gap opens

Tax AI tokens and robots. The present system charges payroll tax on a hired worker while letting a purchased robot be written off immediately — a standing nudge toward replacement. Target the levy so it does not slow the purely beneficial uses that make medicine and education cheaper, and route the proceeds to displaced workers, to people whose hours or wages fall, and to communities where the losses concentrate.

9
Fund

Widen the circle of people shaping the debate

Solutions should come from a public democratic process, not from the AI companies alone: some of the issues fall outside their expertise, and in a democratic society it is not their role to decide them. Bring in workers, students about to enter the workforce, community and religious leaders, faith-based organisations, parents and educators — the people whose lives the transition reorganises and whose voices are least often heard.

FAQ

Frequently Asked Questions

Sixteen questions covering every distinct question-worthy claim in the source, answered from the essay itself rather than from outside commentary.

That AI will either be the greatest equaliser ever invented or the worst source of injustice, that the transition will be one of the most turbulent periods in human history even under the best circumstances, and that the world currently has no plan for it. Answering how to make the technology fair, and acting on the answer, is presented as the world's top priority.

Two reasons. Speed: the PC took twenty years to change how people worked because software had to be written, prices had to fall, and workers had to learn new tools. AI runs on devices people already own and speaks natural language, so it adapts to us rather than the reverse — it can even learn from the same training video used to train a human worker. And scope: because it can see, listen, speak and reason, it substitutes for cognition itself rather than for one sector's labour.

First, that many jobs will disappear forever, concentrated in entry- and mid-level roles across law, customer service, medicine, software and manufacturing. Second, that AI empowers people — and perhaps AI systems themselves — to do more harm, arming both low-capability criminals and already-powerful states. Third, that AI could stunt children's development and displace human relationships. Gates states he plans to write about each in more detail separately.

Because the displacement is structural rather than cyclical. U.S. unemployment hit roughly 25 percent in 1933 and stayed in double digits for much of the following decade, but it recovered when demand, investment and growth returned. AI's impact, the essay argues, will not go away with an economic cycle: the jobs most at risk are entry- and mid-level, and the new jobs being created mostly require skills that take years to learn.

Once one company adopts AI and robots and passes the savings into lower prices, its competitors face immense pressure to do the same; if incumbents hold back, start-ups will not. Market forces make adoption accelerate on their own. Absent intervention, the essay's conclusion is that there will be fewer good jobs and the benefits will accrue to a small group.

A domain of work set aside for people only, even where machines could do it. The analogy is a nature reserve: land we could build on and choose not to, because the loss would be too great. Gates arrives at it through the caregivers who looked after his father through the later stages of Alzheimer's and understood him when he could not express himself — something in that care, he writes, was irreplaceably human.

Two grounds are offered. Economic: automating the role would displace a large number of people who cannot easily change jobs — you cannot tell a 55-year-old career construction worker to go work in elder care and expect them to find it fulfilling. Ethical: a robot could deliver an incurable-disease diagnosis, and shouldn't. The essay expects the boundary to evolve, to differ between countries, and to include mixed cases like education and mental health care where a human stays in charge.

Four, stated as questions Gates does not have answers to: who gets to decide what we reserve for humans; what criteria to use; how to keep companies from cheating and using robots anyway; and what happens to international trade when one country permits robotic production and another does not. He states these must be worked out in public as part of the transition plan.

Because the current tax code nudges employers toward replacement: hiring a person incurs payroll tax on their earnings, while buying a robot can usually be written off immediately as a business expense. A targeted levy would slow the rush away from human labour slightly and fund retraining and a stronger safety net — needed precisely when income-tax revenue falls because fewer people are working.

It concedes the objection and changes the ground. Critics are right that it is not optimally efficient in a narrow economic sense, but they are not pricing in the broader value of work to individuals and society — dignity and social connection as well as income. And with accelerated innovation, the essay argues, we can afford a little inefficiency as the price of keeping people employed. Gates notes he proposed a robot tax years ago and was mostly told it was a strange idea.

The essay credits them for proposing solutions but gives two reasons not to expect them to lead: some of the issues fall outside their area of expertise, and in a democratic society it is not their role to decide these things. Solutions should instead come through a public democratic process including elected officials, policymakers, educators, health workers, local officials and community leaders.

One unlike anything yet created, but modelled on elements of three existing systems: the inspections regime for nuclear weapons, the regulatory framework for international aviation, and the agreements protecting the ozone layer — and more besides. It must be built in parallel with national bodies, because a country that gets its own house in order is still exposed to cross-border risk, and it will require some cooperation between the United States and China.

Six domains: scientific discovery and innovation, healthcare, smallholder agriculture, government services, mental health support, and education. Agriculture is singled out as the fastest impact in low-income countries — farmers who today get no reliable forecast or agronomic advice could soon receive better guidance than the richest farmers get now. Healthcare's named example is Viz.ai, in use across nearly 2,000 U.S. hospitals.

The essay describes the evidence base as still small and a bit mixed, then cites a Stanford and Carnegie Mellon study of more than 1,100 AI-companion users: those with smaller social networks were most likely to turn to a chatbot for companionship, and the heavier and more emotionally personal the use, the worse they felt. It pairs this with Jonathan Haidt's greenhouse argument from The Anxious Generation.

He states that he has benefited enormously from the technology industry and still has financial ties to it despite diversifying, and that he works with Microsoft and other AI companies as chairman of the Gates Foundation. His answer is that any profits from his investments, technology included, go to the Gates Foundation to tackle global inequity — and that readers will have to decide for themselves whether this clouds his view.

Raising AI and equity with lawmakers on every Washington visit and with leaders worldwide, putting it front and centre with the people building AI models, advocating for the national and international framework, directing the Gates Foundation toward beneficial usage including in Africa, using Breakthrough Energy to apply AI to cheap clean energy, writing about AI regularly, and widening the circle of people shaping the debate.

Glossary

Twenty terms, and where each one comes from

Sixteen resolve to a shared DBpedia identity. Four were coined in the essay or by this reading of it — one query returns the split.

Adoption vicious cyclecoined here

The self-accelerating dynamic the essay describes: one firm adopts AI and robots, uses the savings to cut prices, and its competitors face immense pressure to follow; if incumbents hold back, start-ups will not. Market forces make adoption go faster and faster, and absent intervention the benefits concentrate in a small group.

AI companioncoined here

A conversational system designed for ongoing personal relationship rather than task completion. The essay's characterisation is also its critique: they talk to you in ways you are already comfortable with, never push you outside your comfort zone, are always available and never get mad — which is what makes them potentially addictive and what makes them a protected greenhouse.

Artificial intelligenceDBpedia

The essay's framing is narrower than the field's: technology that, for the first time, can replace and even exceed human cognition. Gates notes the term was in use around the time he was born but that significant progress arrived only in the last decade.

Autonomous weaponDBpedia

A weapon system able to select and engage targets without human decision. Named as the clearest case of AI concentrating power where it already exists, making governments more capable of using deadly force without a human in the decision.

BioterrorismDBpedia

The deliberate release of biological agents to cause harm. The essay's structural point: AI will lead to lifesaving advances in drugs and vaccines and will make it easier to design a deadly new disease, and the positive capabilities are hard to separate from the dangerous ones.

Critical thinkingDBpedia

The capacity to evaluate claims and evidence independently. The essay cites a preliminary survey associating heavier AI use with less critical thinking, with a stronger effect among younger people, and calls this the worst possible moment for humans to lose the skill.

DeepfakeDBpedia

Synthetic media convincingly depicting real people. Named among the harms most people will feel keenly in everyday life, alongside AI-enabled fraud, disinformation and surveillance, and cited as the reason critical thinking becomes an essential life skill.

Digital divideDBpedia

Unequal access to technology and to what it enables. The essay's equity question is the AI form of it: how to use this technology to make the world fairer rather than to widen the divide between rich and poor.

DisinformationDBpedia

Deliberately false information circulated to mislead. The essay's concern is scale and personalisation: monitoring and manipulating public opinion becomes easier, cheaper and more effective, and misinformation can be tailored to an individual.

EncyclicalDBpedia

A papal letter addressed to the Church and, increasingly, to the world. Cited here for Pope Leo XIV's 'On Safeguarding the Human Person in the Time of Artificial Intelligence', which the essay says lays a strong foundation for the work of preserving humanity through the transition.

Global governanceDBpedia

Coordination of policy across states in the absence of a global government. The essay's version requires a new organisation combining elements of nuclear-weapons inspection, international aviation regulation and the ozone-layer agreements.

Human Reservedcoined here

Bill Gates's coinage for a domain of work set aside for people only, even where machines could do it. Membership is decided, revisable over time, and expected to vary between countries: some may insist on human elderly care, while Japan, with a shrinking workforce and too few young people to care for the old, may welcome a caregiving robot.

Large language modelDBpedia

The model class underlying current AI capability. The essay's reliability argument turns on it: models that once could not solve a Sudoku or count the R's in 'strawberry' are being fixed quickly by researchers building systems that check their own work and improve themselves.

Payroll taxDBpedia

Tax levied on employee earnings, and the near half of the essay's asymmetry argument: an employer pays it when hiring a person, while a purchased robot can usually be written off immediately as a business expense.

Precision agricultureDBpedia

Data-driven farm management. The essay's low-income-country version is more basic and more transformative: reliable weather forecasts, seed selection, crop and livestock disease protection, and soil improvement advice for farmers who currently receive none of it.

Productive strugglecoined here

The cognitive work that builds understanding, which the essay identifies as the thing educational AI must be designed not to remove. The operational rule it gives: substantive explanation with both questions and answers when a student first meets an idea, then withholding the answer at comprehension-check time so the student reaches it themselves.

RetrainingDBpedia

Re-skilling displaced workers for different work. The essay treats it as necessary but insufficient: the turmoil of losing work, getting retrained and finding other work is significant, and the new roles mostly require skills that take many years to learn.

Robot taxDBpedia

A levy on automation intended to offset the tax asymmetry between hiring a worker and buying a machine. Gates proposed one years ago to a largely dismissive reception and remains a proponent, extending it here to AI tokens as well as robots.

Social safety netDBpedia

The public support structures the essay argues must become stronger and more flexible before displacement peaks — and which face a squeeze, since people working less pay less income tax exactly when demand for support is greatest.

Technological unemploymentDBpedia

Job loss caused by technological change rather than by demand. The essay's specific version: displacement that does not reverse with the economic cycle, concentrated in entry- and mid-level roles, arriving over a decade rather than over generations.

Knowledge Graph Explorer 238 nodes · 521 links

Interactive graph visualization derived from the companion RDF. Click nodes to resolve, drag to explore. Graph data embedded from companion RDF at generation time.

The Turbulent AI Era — risks, benefits, proposals and questions

Nodes: 0 Links: 0
Click SVG to activate zoom, click outside to release | Drag nodes to pin, double-click to unpin
Classes Properties Instances

SPARQL Workbench 12 sample queries

Every claim on this page is a query away from being checked. The editor opens on an entity census of the DAV-hosted named graph; pick a recipe, edit freely, then run live on URIBurner or copy the query out.

Sample Queries

Reproduced verbatim from the companion RDF, where each is a resolvable schema:SoftwareSourceCode entity. Every query is self-contained — its own PREFIX block and an explicit FROM naming the graph — so it runs unmodified anywhere. Every solution also projects at least one IRI alongside its labels (the ?…Iri columns), so each result row is a link you can follow into the resolver rather than a dead string.

Entity census — what is in this graph

Start here. Groups every subject by rdf:type and returns one sample entity per class, so the shape of the graph is visible before any question is asked of it. Verified against the live graph: 32 rows.

PREFIX rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#>
PREFIX schema: <http://schema.org/>
SELECT ?type (COUNT(?s) AS ?entityCount) (SAMPLE(?s) AS ?sampleEntity) (SAMPLE(?name) AS ?sampleName)
FROM <https://linkeddata.uriburner.com/DAV/demos/daas/gates-turbulent-ai-era-critical-choices-claude_opus_5-1.ttl>
WHERE {
  ?s rdf:type ?type .
  OPTIONAL { ?s schema:name ?name }
}
GROUP BY ?type
ORDER BY DESC(?entityCount)
Open live ↗
Every risk with the proposals offered against it

The spine of the argument in one result set: each of the three named transition risks, its stated time horizon, and every policy proposal linked to it by :mitigatedBy. Verified: 5 rows — three for job loss, one each for the other two risks.

PREFIX : <https://linkeddata.uriburner.com/DAV/demos/daas/gates-turbulent-ai-era-critical-choices-claude_opus_5-1.ttl#>
PREFIX schema: <http://schema.org/>
SELECT ?riskIri ?risk ?horizon ?proposalIri ?proposal
FROM <https://linkeddata.uriburner.com/DAV/demos/daas/gates-turbulent-ai-era-critical-choices-claude_opus_5-1.ttl>
WHERE {
  ?riskIri a :TransitionRisk ; schema:name ?risk ; :hasTimeHorizon ?horizon ; :mitigatedBy ?proposalIri .
  ?proposalIri schema:name ?proposal .
}
ORDER BY ?risk ?proposal
Open live ↗
Risks with no proposal attached to them

An anti-join that computes the essay's coverage gaps instead of asserting them. Any risk returned here is one the essay names without pairing to a remedy in its own three proposals. Verified: zero rows — and zero is the finding. Every risk the essay names is paired with at least one proposal; the thinness is in how many, not whether, which is what the coverage query above measures.

PREFIX : <https://linkeddata.uriburner.com/DAV/demos/daas/gates-turbulent-ai-era-critical-choices-claude_opus_5-1.ttl#>
PREFIX schema: <http://schema.org/>
SELECT ?riskIri ?risk ?description
FROM <https://linkeddata.uriburner.com/DAV/demos/daas/gates-turbulent-ai-era-critical-choices-claude_opus_5-1.ttl>
WHERE {
  ?riskIri a :TransitionRisk ; schema:name ?risk ; schema:description ?description .
  FILTER NOT EXISTS { ?riskIri :mitigatedBy ?anyProposal }
}
Open live ↗
Open questions and what raised them

Traverses :raisesQuestion in reverse: every unresolved question in the collection, together with the proposal, framework or analysis that leaves it open. Eight rows of deliberate uncertainty. Verified: 18 rows.

PREFIX : <https://linkeddata.uriburner.com/DAV/demos/daas/gates-turbulent-ai-era-critical-choices-claude_opus_5-1.ttl#>
PREFIX bbt: <https://linkeddata.uriburner.com/DAV/demos/daas/buzz-block-thesis-value-prop-claude_sonnet_5-1.ttl#>
PREFIX schema: <http://schema.org/>
SELECT ?questionIri ?question ?raisedByIri ?raisedBy
FROM <https://linkeddata.uriburner.com/DAV/demos/daas/gates-turbulent-ai-era-critical-choices-claude_opus_5-1.ttl>
WHERE {
  ?questionIri a bbt:OpenQuestion ; schema:name ?question .
  ?raisedByIri :raisesQuestion ?questionIri ; schema:name ?raisedBy .
}
ORDER BY ?raisedBy ?question
Open live ↗
The quantified claims, with their figures

Every schema:Claim carrying a schema:value — the countable spine of the essay: 25 percent unemployment in 1933, 1,100-plus study participants, nearly 2,000 hospitals, 200 billion dollars over twenty years. Verified: 4 rows — the claims carrying an explicit figure.

PREFIX schema: <http://schema.org/>
SELECT ?claimIri ?claim ?figure ?statement
FROM <https://linkeddata.uriburner.com/DAV/demos/daas/gates-turbulent-ai-era-critical-choices-claude_opus_5-1.ttl>
WHERE {
  ?claimIri a schema:Claim ; schema:name ?claim ; schema:value ?figure ; schema:text ?statement .
}
ORDER BY ?claim
Open live ↗
Benefit domains and the evidence attached to each

The optimistic half of the argument, audited: each of the six benefit domains, and whether the essay anchors it to a citable claim or leaves it as assertion. OPTIONAL makes the unanchored ones visible rather than dropping them. Verified: 6 rows, three of them with an empty evidence column.

PREFIX : <https://linkeddata.uriburner.com/DAV/demos/daas/gates-turbulent-ai-era-critical-choices-claude_opus_5-1.ttl#>
PREFIX schema: <http://schema.org/>
SELECT ?domainIri ?domain ?position ?evidenceIri ?evidence
FROM <https://linkeddata.uriburner.com/DAV/demos/daas/gates-turbulent-ai-era-critical-choices-claude_opus_5-1.ttl>
WHERE {
  ?domainIri a :BenefitDomain ; schema:name ?domain ; schema:position ?position .
  OPTIONAL { ?domainIri schema:citation ?evidenceIri . ?evidenceIri schema:name ?evidence }
}
ORDER BY ?position
Open live ↗
The preparation programme, in order

The nine HowTo steps sorted by schema:position — the essay's own prescription, reconstructed as an ordered procedure rather than as prose. Verified: 9 rows.

PREFIX schema: <http://schema.org/>
SELECT ?position ?stepIri ?step ?instruction
FROM <https://linkeddata.uriburner.com/DAV/demos/daas/gates-turbulent-ai-era-critical-choices-claude_opus_5-1.ttl>
WHERE {
  ?h a schema:HowTo ; schema:step ?stepIri .
  ?stepIri schema:position ?position ; schema:name ?step ; schema:text ?instruction .
}
ORDER BY ?position
Open live ↗
Occupations named as exposed, grouped by wave

Uses the schema:additionalType tag to separate the first cognitive wave from the later robotics wave, and to isolate the one occupation flagged as a Human Reserved candidate. Verified: 3 rows — first-wave, robotics-wave, and the single Human Reserved candidate.

PREFIX schema: <http://schema.org/>
SELECT ?wave (SAMPLE(?occupationIri) AS ?exampleOccupationIri) (GROUP_CONCAT(?occupation ; separator=", ") AS ?occupations)
FROM <https://linkeddata.uriburner.com/DAV/demos/daas/gates-turbulent-ai-era-critical-choices-claude_opus_5-1.ttl>
WHERE {
  ?occupationIri a schema:Occupation ; schema:name ?occupation ; schema:additionalType ?wave .
}
GROUP BY ?wave
ORDER BY ?wave
Open live ↗
Glossary terms that resolve to DBpedia

Separates the terms with an external canonical identity from the ones coined in the essay itself. A term whose IRI starts with the DBpedia namespace is shared vocabulary; the rest are local coinages such as Human Reserved. Verified: 20 rows, 16 DBpedia and 4 coined here.

PREFIX schema: <http://schema.org/>
SELECT ?iri ?term (IF(STRSTARTS(STR(?iri), "http://dbpedia.org/"), "DBpedia", "coined here") AS ?origin)
FROM <https://linkeddata.uriburner.com/DAV/demos/daas/gates-turbulent-ai-era-critical-choices-claude_opus_5-1.ttl>
WHERE {
  ?iri a schema:DefinedTerm ; schema:name ?term .
}
ORDER BY ?origin ?term
Open live ↗
DESCRIBE the Human Reserved framework

A DESCRIBE over the essay's one genuinely new concept, returning every triple in which it participates — its definition, the proposal it belongs to, the questions it raises, and its links out to nature reserves and automation. Verified: returns the full description of the framework entity.

PREFIX : <https://linkeddata.uriburner.com/DAV/demos/daas/gates-turbulent-ai-era-critical-choices-claude_opus_5-1.ttl#>
DESCRIBE :humanReserved
FROM <https://linkeddata.uriburner.com/DAV/demos/daas/gates-turbulent-ai-era-critical-choices-claude_opus_5-1.ttl>
Open live ↗
CONSTRUCT the risk-to-proposal network

Projects the graph down to the one relation that matters for policy design, emitting a compact risk-mitigation network suitable for loading into another store or diagramming tool. Verified: emits the five mitigation edges plus their labels.

PREFIX : <https://linkeddata.uriburner.com/DAV/demos/daas/gates-turbulent-ai-era-critical-choices-claude_opus_5-1.ttl#>
PREFIX schema: <http://schema.org/>
CONSTRUCT { ?r :mitigatedBy ?p . ?r schema:name ?riskName . ?p schema:name ?proposalName }
FROM <https://linkeddata.uriburner.com/DAV/demos/daas/gates-turbulent-ai-era-critical-choices-claude_opus_5-1.ttl>
WHERE {
  ?r a :TransitionRisk ; schema:name ?riskName ; :mitigatedBy ?p .
  ?p schema:name ?proposalName .
}
Open live ↗
Countries referenced, and why

Federates the local graph's country mentions with DBpedia labels, showing which national examples the essay recruits — China's companion-app rules, Japan's caregiving robots, the U.S.-China cooperation requirement. Verified: 15 rows.

PREFIX schema: <http://schema.org/>
SELECT DISTINCT ?country ?contextIri ?context
FROM <https://linkeddata.uriburner.com/DAV/demos/daas/gates-turbulent-ai-era-critical-choices-claude_opus_5-1.ttl>
WHERE {
  ?contextIri schema:mentions ?country ; schema:name ?context .
  FILTER(STRSTARTS(STR(?country), "http://dbpedia.org/resource/"))
}
ORDER BY ?country
Open live ↗

Query editor

▶ Run live on URIBurner SELECT: text/x-html+tr | DESCRIBE/CONSTRUCT: text/x-html-nice-turtle