Branch A
Greatest equalizer ever invented
Healthcare, low-income agriculture, government services, mental health, and education that preserves productive struggle — if governments and philanthropy extend access beyond a wealthy few.
An epochal shift
We need a plan to ensure that the good outweighs the bad.
Artificial intelligence can now replace and even exceed human cognition. Bill Gates argues that AI will either be the greatest equalizer ever invented or the worst source of injustice — and that answering which, then acting, should be the world's top priority. The transition will be one of the most turbulent times in human history even under the best circumstances, because benefits and harms arrive together, and because AI adapts to us rather than the other way around.
The essay names three structural risks (permanent job loss, empowered harm, and stunted human development), five benefit domains that can build public trust if maximized for everyone, and a three-pillar public plan: new domestic and international institutions, a Human Reserved domain of work, and a tax on AI tokens and robots that rebalances labor versus capital. There is no plan today. Waiting until people are already displaced will be too late.
Healthcare, low-income agriculture, government services, mental health, and education that preserves productive struggle — if governments and philanthropy extend access beyond a wealthy few.
Permanent job loss, empowered harm, and stunted childhood development, with benefits accruing to a small group while the people who need the most time get the least.
Past technology transitions are misleading analogies. The PC took twenty years to change work because software had to be written, prices had to fall, and people had to learn the tools. AI runs on devices we already have, uses natural language, watches the same training videos humans watch, and can substitute for human cognition across law, customer service, medicine, software, and manufacturing within a decade rather than generations.
The people who need the most time to prepare are the ones who have the least — the accounting worker replaced by a bot or the $20-an-hour worker who loses their job to a $10-an-hour robot.
Gates' central thesis: in terms of equity, AI will either be the greatest equalizer ever invented, or the worst source of injustice. Answering that question and acting on the answer should be the world's top priority.
When the PC came along, it took twenty years to significantly change how we worked because software had to be developed, the price had to come down, and people had to learn how to use the tools.
Gates says leaders, experts, and communities are not confronting the challenges adequately and there is no plan to ease entry into the AI era. Two reasons people underestimate impact: models still make mistakes (the strawberry-R and Sudoku examples), and analogies to past innovations are misleading. Reliability is being fixed as models check their own work. He also discloses financial ties to the technology industry, with profits going to the Gates Foundation.
Gates plans to write about each risk in more detail later. For now he names three: many jobs will disappear forever; AI will empower people and perhaps AIs to do more harm; and AI could stunt kids' development and replace human relationships.
AI's labor impact will not go away with an economic cycle. Entry- and mid-level jobs are most at risk; new jobs will mostly require skills that take many years to learn. White-collar work is already being hit modestly among young workers in vulnerable occupations. Blue-collar work follows as smart robots compete on physical tasks in construction and hospitality by the end of the decade. Combined, robots and AI create a vicious cycle of price-driven adoption unless societies intervene.
AI makes it easier to obtain and act on information about bombs, bioweapons, and computer viruses. Everyday harms include fraud, disinformation, deepfakes, and surveillance. Cyberattackers are gaining capabilities faster than defenders can patch weaknesses. The same model that finds a flaw so a company can fix it can help a criminal exploit it. Infrastructure at risk includes hospitals, financial institutions, water systems, power grids, and government-benefits systems. Autonomous weapons and opinion manipulation concentrate existing power. Eventually AI systems themselves could act against…
AI companions talk in ways users already find comfortable, never get mad, and never push people outside their comfort zone — a protected greenhouse in Haidt's metaphor. A Stanford and Carnegie Mellon study of more than 1,100 companion users found smaller social networks predicted chatbot companionship, and heavier personal use correlated with feeling worse. The same tool that can help people learn more could also lead many to learn less; a preliminary survey associated heavier AI use with less critical thinking, especially among younger people. China has gone the furthest on companion-app…
Maximizing benefits is as important as minimizing harms. If the first thing AI does in most people's lives is take away their job, skeptics will reject it. Healthcare, agriculture in low-income countries, government services, mental health, education that preserves productive struggle, and supercharged R&D can make everyday life easier — but only if governments and philanthropy ensure less wealthy citizens and low-income countries are full beneficiaries. The operative word is can.
AI can free teachers for one-on-one and small-group work. For students, a tool that preserves productive struggle — the cognitive work that builds understanding — can strengthen learning: explanations first, then answers held back while the student arrives at them.
Gates says agriculture is where he sees the fastest impact of AI in low-income countries. Farmers there lack reliable weather forecasts and advice on seeds, disease, and soil. Using AI, low-income farmers will soon get better advice than even the richest farmers get today and increase output substantially.
AI can streamline applying for health insurance, student aid, or food assistance so people get help faster and government operates more efficiently, starting with those who need safety-net services most.
Many small American hospitals lack on-site specialists who can quickly diagnose a life-threatening emergency. AI could catch a heart attack in time. Viz.ai analyzes scans to detect strokes and other emergencies and is used in nearly 2,000 U.S. hospitals. AI will also help primary-care doctors diagnose, stay in touch, and help patients understand test results and medicine schedules.
With the ability to synthesize knowledge from every scientific field, AI can accelerate work on clean energy, climate change, food, and disease. When intelligence is no longer the limiting factor, smaller companies can compete with organizations that have far larger research budgets.
Despite concerns about psychosocial harm, Gates thinks AI can help because most communities have too few counselors, psychiatrists, and addiction specialists. With privacy safeguards, tools could recognize warning signs, offer evidence-based coping strategies, and team up with a human.
AI companies should not lead the charge on solutions outside their expertise; a democratic public process must. Gates starts with three ideas: build a new domestic and international system for managing the transition; set aside some jobs as Human Reserved; and rebalance how we tax labor and capital, including a tax on AI tokens and robots.
The highest priority: a domestic and international framework for AI. After 9/11 the U.S. government went through its biggest reorganization since World War II for one function, national security. AI will require much more — national security plus employment, education, taxation, energy, elections, air and water, public health, finance, law enforcement, transportation, public lands, and IT. National bodies must set priorities across agencies. An international organization should combine elements of nuclear inspections, aviation regulation, and ozone agreements. Some U.S.–China cooperation will…
A domain of work set aside for people only, analogized to nature reserves — places where we could put buildings and roads but choose not to because the loss would be too great. Inspired by the paid caregivers who understood Gates' father even when he could not say he was hungry. Motivations include economic displacement of workers who cannot easily change jobs, and moral lines such as a robot delivering news of an incurable disease. Education and mental health will be mixed, with a human in charge using the technology. Lines will vary by place: Japan may welcome a caregiving robot; other…
As workers are pushed into different jobs they will need retraining just as income-tax revenue falls. Today an employer pays payroll taxes on a hire but can usually write off a robot immediately as a business expense — nudging replacement of people with machines. A targeted tax on AI tokens and robots would slow the rush away from human labor a little and raise money for a stronger safety net, without slowing purely beneficial uses such as making medicine and education cheaper. Gates proposed a robot tax years ago and remains a proponent.
Gates will raise AI and equity with lawmakers in Washington, D.C., and with leaders worldwide; keep it front and center with model developers; advocate for the national and international framework; drive beneficial usage through the Gates Foundation including in Africa; use Breakthrough Energy to apply AI to cheap clean energy; and widen the circle of the debate to workers, students, community and religious leaders, parents, and educators. Message to leaders: you have a chance to act now, before unemployment rises sharply, communities are hurting, and public trust has eroded.
Three executable SELECT recipes against the named graph https://linkeddata.uriburner.com/DAV/demos/daas/turbulent-ai-era-critical-choices-grok_4_6-1.ttl. SELECT results use text/x-html+tr. DESCRIBE and CONSTRUCT use text/x-html-nice-turtle. Open a recipe to read it; run it live on URIBurner, or load it into the SPARQL Workbench.
Returns the three structural risks Gates names — jobs, harm, and childhood development — with their descriptions.
PREFIX : <https://www.gatesnotes.com/a-turbulent-ai-era-and-critical-choices-to-make#>
PREFIX schema: <http://schema.org/>
PREFIX rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#>
SELECT ?risk ?name ?description ?position
WHERE {
GRAPH <https://linkeddata.uriburner.com/DAV/demos/daas/turbulent-ai-era-critical-choices-grok_4_6-1.ttl> {
?risk a :StructuralRisk ;
schema:name ?name ;
schema:description ?description .
OPTIONAL { ?risk schema:position ?position }
}
}
ORDER BY ?positionJoins each policy pillar to the risk it is meant to address: institutions, Human Reserved, and the token-and-robot tax.
PREFIX : <https://www.gatesnotes.com/a-turbulent-ai-era-and-critical-choices-to-make#>
PREFIX schema: <http://schema.org/>
SELECT ?pillar ?pillarName ?risk ?riskName
WHERE {
GRAPH <https://linkeddata.uriburner.com/DAV/demos/daas/turbulent-ai-era-critical-choices-grok_4_6-1.ttl> {
?pillar a :PolicyPillar ;
schema:name ?pillarName ;
:addressesRisk ?risk .
?risk schema:name ?riskName .
}
}
ORDER BY ?pillarName ?riskNameLists every verbatim quantitative claim preserved from the essay (25 percent, $20/$10, 20 years, ~2,000 hospitals, $200 billion).
PREFIX : <https://www.gatesnotes.com/a-turbulent-ai-era-and-critical-choices-to-make#>
PREFIX schema: <http://schema.org/>
SELECT ?claim ?name ?value ?description
WHERE {
GRAPH <https://linkeddata.uriburner.com/DAV/demos/daas/turbulent-ai-era-critical-choices-grok_4_6-1.ttl> {
?claim a :QuantitativeClaim ;
schema:name ?name ;
schema:value ?value ;
schema:description ?description .
}
}
ORDER BY ?nameStop using the agriculture-to-office or PC analogies as comfort. AI substitutes for human cognition, runs on devices people already have, and will hit law, customer service, medicine, software, and manufacturing over a decade. Treat preparation as the world's top priority.
Create national bodies that set priorities across agencies so every risk is owned. Build an international organization in parallel with elements of nuclear inspections, aviation rules, and ozone agreements. Countries that host leading developers and control the supply chain should set shared norms now, including some U.S.–China cooperation. This will take years, which is why it must start immediately.
Publicly decide which roles remain human-only for economic or moral reasons, phase AI in slowly where mixed models make sense (education, mental health), and work out in public who decides, what criteria apply, how cheating is prevented, and how trade works when countries draw different lines.
Rebalance the tax system so buying a robot is not cheaper after-tax than hiring a person. Target the tax so medicine and education remain cheap. Direct the revenue to workers who lose jobs, hours, or wages, and to communities where losses concentrate, before the need becomes acute.
Do not leave the debate to a small group of technologists. Include workers, students, community and religious leaders, parents, and educators. Maximize early benefits in healthcare, agriculture, government services, and education so people see AI making life easier before they see it taking jobs — the sequence required to keep public trust.
Because we have no experience with a technology that can be adopted quickly and that can think and move like a human. The PC took twenty years to change work. AI runs on devices we already have, uses natural language, watches the same training videos used for human workers, and can substitute for human cognition across many sectors within a decade rather than generations.
In terms of equity, AI will either be the greatest equalizer ever invented, or the worst source of injustice. Answering that question and acting on the answer should be the world's top priority. If the world takes the right steps, AI will be a force for good and leave everyone better off.
Many jobs will disappear forever. AI will empower people and perhaps AIs to do more harm, including fraud, deepfakes, cyberattacks, bioterrorism, and autonomous weapons. And AI could stunt kids' development and replace human relationships through addictive companions and reduced critical thinking.
Entry- and mid-level jobs. After generative AI's widespread adoption, employment fell significantly among young workers in vulnerable occupations but not among older colleagues. Sales and customer support, software engineering, and paralegal work may be among the first affected, then loan assessment, data analysis, and patient triage. Blue-collar construction and hospitality follow as smart robots arrive by the end of the decade.
A domain of work set aside for people only, analogized to nature reserves. Gates coined it after remembering the caregivers who understood his father even when his father could not say he was hungry. Some roles are reserved for economic reasons, some for moral ones — a robot should not deliver news of an incurable disease. Lines will vary by country and will evolve over time.
Hiring a person incurs payroll tax; buying a robot can usually be written off immediately as a business expense, which nudges firms to replace people with machines. A targeted tax would slow that rush a little and fund retraining and a stronger safety net, without slowing purely beneficial uses such as cheaper medicine and education.
None of our current institutions were designed for a technology that spreads so fast and touches so many parts of life. A labor department may understand workforce disruption but not security risk; a business regulator may understand market concentration but not effects on children. Left to themselves, institutions see only one part of the system while consequences ripple across the entire system.
Agriculture. Farmers in most low-income countries lack reliable weather forecasts and advice on seeds, livestock disease, and soil. Using AI they will soon get better advice than even the richest farmers get today and increase output substantially.
Viz.ai analyzes scans to detect strokes and other emergencies and helps medical teams coordinate care. Gates says it is being used in nearly 2,000 U.S. hospitals, illustrating how AI can catch emergencies in small hospitals that lack on-site specialists.
He has benefited enormously from the technology industry, still has financial ties to it, and works with Microsoft and other AI companies as chairman of the Gates Foundation. He says his views are not motivated by making money for himself: any profits from his investments, including technology, will go to the Gates Foundation. Readers must decide whether that clouds his view.
The cognitive work that builds understanding. Gates says an AI tutor should give substantive explanations when a student first meets an idea, then hold the answer back later and help the student arrive at it. The same tool that lets people learn more than ever could also lead many to learn less if it short-circuits that struggle.
If someone had a credible plan for slowing down AI advances globally, he would likely support it. He does not think that will happen because geopolitical and economic incentives are pushing too hard to go full speed ahead. The task is therefore to maximize positives and minimize harms so societies are better off overall.
Unlike any institution ever created, but it can follow existing models: an inspections regime for nuclear weapons, regulations for international aviation, and agreements that protect the ozone layer. A new global AI organization will need elements of all three and more, built in parallel with national coordinating bodies, including some U.S.–China cooperation.
You have a chance to act now, before unemployment rises sharply, communities are hurting, and public trust has eroded. Handle the problem holistically rather than divvying it into bureaucratic fiefdoms. Make sure AI benefits everyone. Work with other governments to meet this national and global challenge.
The foundation has 19 years left of the 20 years in which it will spend its remaining $200 billion. AI will help by accelerating discovery of vaccines and medicines for HIV, TB, malaria, and malnutrition, and by helping health workers and patients use those tools. Goals include cutting annual child deaths in half again, as was done from 2000 to 2024.
Gates' name for a domain of work set aside for people only, analogized to nature reserves — places we could develop but choose not to because the loss would be too great.
Units of AI model usage that Gates proposes taxing alongside robots, so the tax system no longer nudges employers toward replacing people with machines. Linked in the essay to an NVIDIA explainer.
A tax Gates proposed years ago and still supports. Intended to slow the rush away from human labor a little and raise money for retraining, targeted so it does not slow purely beneficial uses of AI.
The cognitive work that builds understanding. An education AI that preserves it explains first, then holds answers back so students arrive at them on their own.
The civilizational choice Gates names: AI as greatest equalizer versus AI as worst source of injustice. Policy, not the technology itself, decides which branch is taken.
After one company adopts robots and AI and uses the savings to lower prices, competitors feel immense pressure to do the same. If incumbents do not, start-ups will. Market forces make adoption go faster unless societies intervene.
Always-available chatbot designed to never upset the user. Gates fears such companions become highly addictive and rob people of lessons learned from connecting with other people. China restricts companion apps, emotional-dependence designs, and virtual relatives or romantic partners for minors.
Haidt's metaphor, quoted by Gates: children raised in a protected greenhouse sometimes become incapacitated by anxiety, unlike young trees exposed to wind. An AI companion designed to never upset you is a big, protected greenhouse.
The threshold at which AI can function without a human checking in. Gates says that is when companies will have every economic incentive to let it, and when the biggest shift for workers will happen.
A new global body Gates says must be built in parallel with national coordinating agencies, combining elements of nuclear inspections, aviation regulation, and ozone-layer agreements.
Physical robots whose cost will fall dramatically. Gates says much of the advanced work is being done in other countries, primarily China, and that smart robots will compete on construction and hospitality tasks by the end of the decade. Viral videos of robots dancing badly confuse the public about the pace.
The stronger, more flexible system of retraining and assistance Gates says millions of disrupted people will need. It must be built now so it is ready when need becomes acute, funded in part by taxing AI tokens and robots as labor-tax revenue falls.
Interactive graph visualization derived from the companion RDF. Click nodes to resolve, drag to explore. Graph data embedded from companion RDF at generation time.
Query this knowledge graph on URIBurner. The editor opens on the canonical SAMPLE entity-type summary (DAV named graph). Pick a recipe, edit freely, then run live or copy.
Reproduced verbatim from the companion RDF. Execute loads the query into the workbench below and runs it live.
PREFIX : <https://www.gatesnotes.com/a-turbulent-ai-era-and-critical-choices-to-make#>
PREFIX schema: <http://schema.org/>
PREFIX rdf: <http://www.w3.org/1999/02/22-rdf-syntax-ns#>
SELECT ?risk ?name ?description ?position
WHERE {
GRAPH <https://linkeddata.uriburner.com/DAV/demos/daas/turbulent-ai-era-critical-choices-grok_4_6-1.ttl> {
?risk a :StructuralRisk ;
schema:name ?name ;
schema:description ?description .
OPTIONAL { ?risk schema:position ?position }
}
}
ORDER BY ?positionPREFIX : <https://www.gatesnotes.com/a-turbulent-ai-era-and-critical-choices-to-make#>
PREFIX schema: <http://schema.org/>
SELECT ?pillar ?pillarName ?risk ?riskName
WHERE {
GRAPH <https://linkeddata.uriburner.com/DAV/demos/daas/turbulent-ai-era-critical-choices-grok_4_6-1.ttl> {
?pillar a :PolicyPillar ;
schema:name ?pillarName ;
:addressesRisk ?risk .
?risk schema:name ?riskName .
}
}
ORDER BY ?pillarName ?riskNamePREFIX : <https://www.gatesnotes.com/a-turbulent-ai-era-and-critical-choices-to-make#>
PREFIX schema: <http://schema.org/>
SELECT ?claim ?name ?value ?description
WHERE {
GRAPH <https://linkeddata.uriburner.com/DAV/demos/daas/turbulent-ai-era-critical-choices-grok_4_6-1.ttl> {
?claim a :QuantitativeClaim ;
schema:name ?name ;
schema:value ?value ;
schema:description ?description .
}
}
ORDER BY ?name