Summaries > AI > Stanford > Stanford's Top AI Economist: The Next 10 Years Will Be the Best AND the Worst in ...

Stanford's Top Ai Economist: The Next 10 Years Will Be The Best And The Worst In History

https://www.youtube.com/watch?v=72duHF7iZiU

TLDR AI is already reshaping work, especially for younger workers in exposed fields, but it changes tasks more often than it eliminates whole occupations. Its biggest gains may arrive within three to five years, as businesses redesign work and people learn to use AI as a collaborator—while human judgment, creativity, domain knowledge, and connection remain valuable. The technology could create enormous wealth and improve health and living standards, but leaders must help workers adapt and prevent gains and power from concentrating in a few hands. AI’s future is still ours to shape.

Key Insights

1. Treat AI as a tool that changes tasks, not entire occupations

Start by identifying which tasks in a role AI can perform, rather than assuming it will eliminate the whole job. AI may handle activities such as reading medical images or drafting routine marketing material while people remain responsible for judgment, communication, and other duties. This distinction helps workers and employers pinpoint where roles may change and where human expertise still matters. Employment effects can vary across fields, and lower costs may even increase demand for some services.

2. Use AI to augment your work before trying to automate it

Look for ways AI can help you work faster, explore more ideas, or improve the quality of your output. Workers who use AI to augment their contributions may be better positioned than those who use it only to remove tasks from a process. For a practical starting point, ask an AI assistant to interview you about your work and recurring challenges, then suggest useful applications. Review its recommendations and keep responsibility for consequential decisions.

3. Build problem-framing and evaluation skills

As AI takes on more execution, people will increasingly need to define the right problem and judge whether the result is useful and accurate. Practice describing open-ended goals, setting constraints, and checking answers against evidence and your own expertise. Broad knowledge can help you recognize errors that a polished AI response might conceal. These skills are valuable alongside technical ability, especially when work requires judgment rather than simply following a procedure.

4. Combine technical ability with domain knowledge and people skills

The greatest value often comes from combining an understanding of AI with expertise in a particular field and the ability to work well with people. Technical tools can produce results, but domain knowledge helps determine whether those results fit the real situation. Communication and collaboration also help turn promising outputs into services or decisions that people can use. Develop all three areas rather than relying on technical skills alone.

5. Prepare for transitions through learning and flexible career planning

Workers in AI-exposed areas, including some coding, sales, marketing, and administrative roles, should expect tasks and required skills to change. Build transferable abilities such as project management, problem-solving, and evaluating AI-generated work, and look for opportunities to apply them in less-exposed or growing fields. Employers can support this transition by using AI to reduce routine work while training junior staff in broader responsibilities. Avoid assuming that every current job will remain unchanged—or that new opportunities will appear automatically.

6. Experiment with AI to create useful products and services

AI can make it cheaper to brainstorm, prototype, and test ideas, so use it to explore needs that are not yet being met. Try small experiments, gather feedback, and refine what works instead of waiting for a complete plan. Human improvisation, creativity, and the ability to notice problems remain important advantages when situations are uncertain. Entrepreneurship is one way people can help shape how AI creates value, though wider policies may also be needed to share its benefits.

7. Keep human judgment in charge of high-stakes decisions

Use AI as a collaborator, not as the final authority on decisions involving health, safety, finances, or other major consequences. Ask it to explain its reasoning, compare alternatives, and identify uncertainties, then verify important claims with reliable sources or qualified professionals. AI can accelerate research and planning, but it can also produce errors or miss context. Retaining human oversight makes its speed and breadth more useful and responsible.

8. Help steer AI toward broadly shared benefits

AI could create substantial productivity gains and improve access to valuable services, but those gains may not reach everyone automatically. Support workplace training, education, and practical pathways for displaced workers as technology changes the demand for skills. Businesses, policymakers, and communities can also consider how to prevent economic and political power from becoming too concentrated. The choices people make now will help determine whether AI contributes to shared prosperity or deepens existing inequalities.

Questions & Answers

What employment effects of AI has research found so far?

Research found about 16% lower employment among workers up to age 25 in occupations most exposed to AI. Coding, call centers, and some sales and marketing tasks are especially affected, while employment is growing in less-exposed fields such as home health care. Workers who use AI to augment their work fare better than those who mainly use it to automate tasks.

Does AI eliminate entire occupations?

AI primarily affects tasks rather than eliminating whole occupations. For example, AI can read medical images, but it cannot perform all the other duties of a radiologist. It can reduce staffing in some areas, while lower costs can also increase demand and employment in others.

When will AI produce significant productivity gains?

The speaker expects meaningful productivity gains in roughly three to five years and believes the effects could become transformative by 2030. Businesses need time to redesign processes, reskill workers, and develop new products, so the change will be fast relative to past technologies but not instantaneous.

How should workers prepare for a future with AI agents?

As AI agents take on more execution, people will need to define the right problems, scope open-ended work, and judge whether AI results meet their needs. Combining technical skills with people skills and domain knowledge will be especially valuable.

Which jobs and workers are especially exposed to AI?

Junior software engineering, mid-level marketing management, paralegal work, coding, call centers, and some sales and marketing tasks are particularly exposed. Skilled trades may be harder for machines to replace, though that advantage may not last indefinitely.

What should companies and society do about AI-driven job displacement?

Companies should reskill workers and continue developing junior employees rather than eliminate the pipeline that produces future senior workers. Society should invest in education and retraining, help people transition into new work, and plan for disruption instead of trying to preserve every existing job.

Will AI reduce the number of radiologists?

Not necessarily. Despite predictions that AI would make radiologists unnecessary, there are now more radiologists than ever and the field faces a shortage. AI may make imaging more efficient and affordable, which can increase demand and allow more people to receive care.

What human abilities may become more valuable as AI advances?

Human agency, creativity, improvisation, judgment, and authentic connection may become more valuable. Liberal arts and broad academic knowledge can also help people develop taste and assess whether AI-generated answers are correct.

What is the main risk of AI-driven wealth creation?

The gains could concentrate in a small number of companies or governments, concentrating political power as well as wealth. The goal should be shared prosperity, with many people able to participate and hold a stake in society.

What policies could help distribute AI’s economic gains?

Entrepreneurship and broader participation in creating AI-powered products and services can help. If those measures are insufficient, the speaker says options such as universal basic income and progressive income or wealth taxes should remain available.

How might AI change economic measurement?

Traditional GDP misses much of the value of free goods such as Wikipedia, YouTube, and ChatGPT. The speaker’s team developed GDP Benefits (GDPB), which estimates value by asking how much people would need to be paid to stop using a good or service.

How does the speaker recommend that newcomers use AI?

Ask an AI to interview you about your work, family, and recent problems, then suggest practical ways it could help. Treat AI as a collaborator: evaluate its suggestions and override them when needed, especially for life-or-death decisions.

What are the major promises and dangers of AI?

AI could create substantial wealth, improve productivity, extend lives, and help cure diseases. Serious dangers include engineered viruses, manipulation, concentration of power, and lethal autonomous weapons.

How can people influence AI’s future?

People have agency: they can decide what they want to use AI for, build useful products and services, and work with companies and policymakers to steer the technology toward shared prosperity while reducing catastrophic risks.

Summary of Timestamps

AI’s effects on employment are already appearing, especially among younger workers in highly exposed fields. Research cited in the conversation found employment about 16% lower for workers up to age 25 in occupations most exposed to AI, including coding, call centers, and some sales and marketing tasks. The impact is uneven: employment is growing in less-exposed areas such as home health care.

AI changes tasks more often than it eliminates entire occupations. For example, AI can read medical images, but it cannot perform every responsibility of a radiologist. Workers who use AI to augment their abilities may fare better than those whose work is primarily automated, and lower costs can sometimes increase demand enough to create more work.

AI’s business impact may arrive faster than the effects of past technologies, but organizations still need time to redesign processes, retrain workers, and create new products. The speaker expects meaningful productivity gains within roughly three to five years and sees a major opportunity for businesses to use AI more effectively.

As AI agents take on more execution, people will increasingly need to define the right problems and judge whether the results are useful. Skills such as scoping open-ended challenges, evaluating answers, and combining technical ability with judgment will become increasingly important.

The greatest value comes from combining technical skills with people skills and knowledge of a specific field. As AI handles more routine tasks, broad education may become more important because it helps people assess whether an answer is accurate, relevant, and appropriate.

Entry-level software engineering, mid-level marketing management, and paralegal work are among the roles facing significant exposure. Companies that stop hiring junior workers risk weakening the pipeline that develops future experts, so education, retraining, and thoughtful workplace practices will be essential.

Technological change destroys some jobs while creating others, so societies should help people transition rather than trying to preserve every existing role. The speaker warns that failing to support workers affected by AI could provoke a backlash similar to the reaction to globalization.

Radiology shows why AI’s employment effects can be counterintuitive. Despite predictions that AI would make radiologists unnecessary, the number of radiologists has grown and the field faces a shortage. More efficient, less expensive imaging may increase demand and allow more people to receive care.

The speaker believes AI could have a broader economic impact than the Industrial Revolution: earlier machines amplified physical strength, while AI can augment human thinking across many kinds of work. As intelligence becomes more widely available, human agency, authentic connection, and creative judgment may become more valuable.

Humanities and the arts may gain importance in a world of greater abundance. Philosophy, music, art, and history can help people develop taste, interpretation, and appreciation—qualities that matter when AI can generate or retrieve information but people must decide what is meaningful.

Many cognitive jobs may be automated, while skilled trades could be harder to replace—at least for now. Still, history suggests people often underestimate the new jobs and opportunities that technology will create, making entrepreneurship and innovation important sources of future work.

A distinctly human strength is improvisation: people can respond creatively when circumstances change or something unexpected happens. AI may excel at defined tasks, but human flexibility can help address situations that were not anticipated in advance.

AI could generate enormous wealth, but the gains might concentrate in a small number of companies or governments, increasing political as well as economic power. The goal should be shared prosperity, with broad opportunities for people to participate and benefit.

People can help shape the future by using AI to identify needs, develop products and services, and test ideas. Because AI makes brainstorming and experimentation less expensive, more people can act entrepreneurially instead of limiting themselves to following instructions.

Entrepreneurship may not be enough to prevent extreme concentration of wealth, so policies such as universal basic income or progressive income and wealth taxes could serve as safeguards. The speaker argues that a society with sharply unequal access to safety and services would be worse for everyone.

AI agents could transform companies and markets, which are themselves systems for processing information. The speaker expects many basic needs to become nearly free, while scarce goods and social status remain important; a key challenge will be directing status competition toward beneficial achievements.

Traditional GDP misses some of the value people receive from free services such as Wikipedia, YouTube, and ChatGPT. The speaker’s team developed GDP Benefits, or GDPB, which estimates value by asking how much someone would need to be paid to give up a product or service.

The speaker uses tools such as Claude and ChatGPT for research and everyday planning. For newcomers, they recommend asking an AI to learn about a person’s work and recurring challenges, then suggest practical ways to help—while keeping human judgment in control of important decisions.

AI has made the speaker’s research faster: discussions that once took weeks can sometimes happen in minutes. This efficiency helps researchers tackle more interesting problems, although the broader economic effects of AI have not yet reached the technology’s full potential.

The next decade could bring major advances in wealth, health, and disease treatment, but it also carries serious risks, including engineered viruses, manipulation, concentrated power, and autonomous weapons. The speaker describes feeling both excited and concerned about what lies ahead.

The closing message is that people have agency over AI’s future. Rather than focusing only on what AI might do to society, people should decide what they want to use it for and help steer it toward shared prosperity while working to prevent catastrophic outcomes.