Source-linked AI summary
Artificial Intelligence and Life in 2030: The One Hundred Year Study on Artificial Intelligence
Peter Stone, Rodney Brooks, Erik Brynjolfsson, Ryan Calo, Oren Etzioni, Greg Hager, Julia Hirschberg, Shivaram Kalyanakrishnan, Ece Kamar, Sarit Kraus, Kevin Leyton-Brown, David Parkes, William Press, AnnaLee Saxenian, Julie Shah, Milind Tambe, Astro Teller
TL;DR
The report addresses how AI is progressing and affecting people, communities, and society, especially in a typical North American city by 2030. It uses a recurring, domain-based assessment of AI’s development and societal impacts. It concludes that useful applications are likely to expand, while policy and deployment choices will shape their benefits, risks, and distribution.
Problem
The report addresses the need for an accurate assessment of AI’s current state, potential advances, and effects across society and major urban domains.
Method
A multidisciplinary Study Panel analyzes AI progress and anticipated developments across eight domains in a typical North American city through 2030.
Results
The panel expects useful AI applications to expand through 2030, alongside workforce disruption and ethical and social challenges.
Takeaways & Limitations
Near-term application design and policy decisions should balance innovation with safety, reliability, fairness, privacy, and broadly shared benefits.
Takeaways & Limitations
The report focuses on a typical North American city and excludes military applications from this initial report.
Abstract
from arXiv · showhide
In September 2016, Stanford's "One Hundred Year Study on Artificial Intelligence" project (AI100) issued the first report of its planned long-term periodic assessment of artificial intelligence (AI) and its impact on society. It was written by a panel of 17 study authors, each of whom is deeply rooted in AI research, chaired by Peter Stone of the University of Texas at Austin. The report, entitled "Artificial Intelligence and Life in 2030," examines eight domains of typical urban settings on which AI is likely to have impact over the coming years: transportation, home and service robots, healthcare, education, public safety and security, low-resource communities, employment and workplace, and entertainment. It aims to provide the general public with a scientifically and technologically accurate portrayal of the current state of AI and its potential and to help guide decisions in industry and governments, as well as to inform research and development in the field. The charge for this report was given to the panel by the AI100 Standing Committee, chaired by Barbara Grosz of Harvard University.
STUDY PANEL
The report was produced by a multidisciplinary Study Panel and supported by the Standing Committee and contributors.
- The report was published by Stanford University in 2016 under a Creative Commons Attribution-NoDerivatives license.
- The Standing Committee acknowledged the panel, its chair Peter Stone, and contributors who supported the report’s development and editing.
EXECUTIVE SUMMARY
The report assesses how AI may affect everyday life in a typical North American city through 2030. It describes substantial potential benefits alongside workforce disruption, ethical challenges, and the need for informed policy.
- AI is already changing everyday life through specialized technologies such as computer vision, deep learning, speech understanding, natural language processing, and reasoning systems.
- AI applications are expected to expand in areas including self-driving cars, healthcare diagnostics, targeted treatments, elder care, agriculture, factories, and fulfillment centers.
- The report examines eight domains and reviews their past progress while anticipating developments over the next fifteen years.
- The panel found no cause for concern that AI is an imminent threat to humankind, while expecting increasingly useful applications with potentially profound positive impacts through 2030.
- AI may augment or replace human labor, creating economic and social challenges that make broadly shared benefits and near-term policy choices important.
OVERVIEW
The report presents AI as specialized technology already reshaping daily life, while emphasizing domain-specific challenges, ethical risks, and the importance of steering deployment responsibly. It also describes current research trends toward deep learning, human-aware systems, and renewed traditional AI methods.
- AI applications are portrayed as improving health, safety, productivity, driving, learning, and the ability to extend and enhance people’s lives.
- Future human-machine relationships are expected to become more nuanced, fluid, and personalized as systems adapt to individual personalities and goals.
- The report focuses on eight domains in a typical North American city where AI already has or is projected to have substantial impact.
- AI’s benefits coexist with ethical and social issues, including privacy concerns and job displacement, while research, development, and regulation will shape how benefits and risks are distributed.
- AI systems remain specialized, with each application requiring years of focused research and careful, unique construction.
- The report recommends technical expertise in government and research on fairness, security, privacy, and societal impacts.
- AI-related governance raises questions about responsibility, discrimination, financial cheating, and the distribution of gains across multiple regulatory agencies.
- AI RESEARCH TRENDS: Deep learning has advanced perception and reinforcement learning, including AlphaGo’s success through self-play and reinforcement learning.
SECTION II: AI BY DOMAIN
The report identifies eight domains for assessing AI’s effects, with transportation emerging as an early setting where people must trust AI for critical tasks. Rapid progress in autonomous driving could reshape urban life while raising safety, ethical, legal, and access concerns.
- The study organizes AI’s societal effects into eight domains, including transportation, healthcare, education, employment, and entertainment.
- Transportation: Transportation is likely to be among the first domains requiring public trust in AI reliability and safety for a critical task.
- Transportation: From 2004–2012, rapid progress in sensing and machine learning moved autonomous vehicles from failed challenges toward deployment on city streets.
- Transportation: Google’s vehicles were completely autonomous, while Tesla’s semi-autonomous cars required engaged human drivers who could take over when needed.
- Transportation: Broad acceptance of self-driving cars remains uncertain because the required improvement and the cognitive demands of human–vehicle collaboration are not well understood.
- Transportation: Autonomous transportation raises unresolved security, ethical, legal, and responsibility questions, including testing across road conditions and liability after accidents.
- Transportation: Dynamic pricing can ration transportation access by willingness-to-pay, creating fairness concerns when high-demand services become unavailable to some groups.
Ethical questions arise
AI-enabled transportation depends on sensing, connectivity, optimization, and increasingly connected infrastructure. These developments create practical benefits but also expose unresolved challenges involving standardization, privacy, safety, regulation, and equitable access.
- Large-scale transportation data and connectivity support real-time traffic prediction, route calculation, ridesharing, and other machine-learning applications.
- City infrastructure has adopted sensing and optimization more slowly than individual vehicles, with no standardization across cities and high coordination costs.
- Connected transportation infrastructure could improve reliability and efficiency through car-to-car communication, multi-agent coordination, collaboration, and planning.
- Drones and connected infrastructure increase concerns about privacy, private-data safety, and the pace and scope of future transportation advances.
- On-demand transportation uses location and reputation modeling to match drivers with passengers, while dynamic pricing rations access and raises regulatory and safety issues.
- By 2030, humans are expected to partner with self-driving cars and drones in training, execution, evaluation, communication, and coordination.
HOME/SERVICE ROBOTS
Home and service robots are becoming safer and more capable through advances in AI, sensing, embedded computing, cloud systems, and mechanical design. Nevertheless, high costs, hardware constraints, and limited demand are expected to keep applications narrowly defined.
- Over the next fifteen years, mechanical and AI advances are expected to increase the safe, reliable use and utility of home robots.
- Special-purpose robots may deliver packages, clean offices, and enhance security, but technical constraints and high costs will limit commercial applications.
- Home-robot applications have expanded slowly because real homes challenge mobility, hardware remains difficult to build, and few applications attract enough purchasing demand.
- Cloud systems can accelerate software releases and aggregate household data for machine-learning improvements to already deployed robots.
- Advances in speech understanding and image labeling enabled by deep learning are expected to improve robots’ interactions with people at home.
- Low-cost 3D sensors and improved embedded processors are expected to speed development and enable more onboard AI in robots.
- Low-cost robot arms have reached hundreds of research labs, supporting manipulation research that may become applicable in homes around 2025.
HEALTHCARE
AI could improve healthcare through data-driven diagnosis, treatment, monitoring, and clinical assistance, but adoption depends on trust, better interaction, and regulatory and structural change.
- HEALTHCARE: Healthcare AI could improve outcomes and quality of life through clinical decision support, monitoring, coaching, automated care devices, and system management.These applications require trust from doctors, nurses, and patients, alongside removal of policy, regulatory, and commercial obstacles.
- Healthcare analytics: Large-scale clinical data could support finer-grained diagnosis and treatment for individual patients and populations.Relevant data comes from monitoring devices, mobile apps, electronic health records, and surgical or hospital robots.
- The clinical setting: Healthcare delivery remains structurally ill-suited to absorb rapid advances, with poor EHR implementation, substandard interfaces, and regulatory barriers limiting analytics.These problems have eroded clinicians’ confidence and left the promise of EHR-based AI largely unrealized.
- The clinical setting: Automated image interpretation is gaining momentum, but near-term systems are more likely to triage or check images than replace radiologists.Deep neural networks have been trained on linked scans, radiological reports, and patient records to produce basic findings with high reliability.
- The clinical setting: The da Vinci system became standard care in multiple laparoscopic procedures and is used in nearly three quarters of a million procedures annually.Its physical and data platforms have supported new instrumentation, image fusion, biomarkers, and competing robotic-surgery ecosystems.
The problem in medicine
Medical AI faces a difficult recognition problem: useful systems must support fine-grained, high-stakes judgments while fitting into complex human and organizational workflows.
- The problem in medicine: Hospital automation has been less successful than surgical robotics, with delivery robots adopted by few hospitals despite practical systems in other service industries.Robots may deliver goods to the correct room while people perform the final placement.
- The problem in medicine: Healthcare robotics is increasingly generating data for quantifying performance, identifying deficiencies and errors, and developing predictive analytics.Surgical, delivery, and patient-care platforms are building data-oriented capabilities on top of semi-automation.
- The problem in medicine: New healthcare applications combine mobile, social, wearable, and environmental data to provide monitoring, recommendations, and personalized health management.Examples include alerts for caregivers, exercise coaching, and support for people with multiple conditions or treatment interactions.
- The problem in medicine: The aging population creates opportunities for home health monitoring, social support, transportation, assistive devices, and help with daily living.The number of elderly people in the United States was projected to grow by over 50% over the following fifteen years.
Personalized rehabilitation
Personalized rehabilitation and home-based assistance could extend independent living through monitoring, adaptive support, and assistive technologies, while education AI expands personalized learning under evidence and resource constraints.
- Personalized rehabilitation: In-home monitoring and mobile applications could detect behavioral changes, alert caregivers, and recommend activities supporting mental and physical health.Personalized management may also address complexities from multiple co-morbid conditions and treatment interactions.
- Personalized rehabilitation: Personalized rehabilitation and in-home therapy could reduce the need for hospital or care-facility stays.Intelligent walkers, wheelchairs, and exoskeletons could extend the activities available to infirm individuals.
- Personalized rehabilitation: These innovations introduce privacy questions within relationships among patients, families, and caregivers.They also create challenges for accommodating an increasingly active and engaged population beyond retirement.
- EDUCATION: Educational AI is expected to assist human teachers through improved interaction in classrooms and homes and through more sophisticated immersive virtual-reality learning.The report emphasizes that human teachers remain part of quality education.
- EDUCATION: Schools and universities have been slow to adopt AI because of limited funds and insufficient evidence that these technologies improve learning objectives.Robotics kits and other tools therefore require compelling evidence of improved academic performance to become widespread.
- EDUCATION: Intelligent tutoring systems and online learning tools support individualized mastery, automated assessment, and much larger classes.Examples include Bayesian Knowledge Tracing for personalized sequencing and AI-enabled courses with class sizes of a few tens of thousands.
While formal education
AI is expected to blur formal education and self-paced learning through adaptive systems, online education, digital media, translation, and professional retraining. These changes may broaden educational access while also creating social and developmental concerns.
- While formal education: Adaptive learning systems are expected to become a core part of higher education as institutions serve more students while containing costs.The panel links this shift to faster student progression and broader use of individualized learning.
- MOOCs and other forms: MOOCs and other online education formats are expected to become part of education at multiple levels, including K-12.
- MOOCs and other forms: Digital and audio texts, smarter reading devices, and machine translation are expected to make educational materials more accessible and affordable.Translation services will increasingly combine human translators with automatic methods to improve speed and affordability for school systems.
- MOOCs and other forms: Online learning systems will expand opportunities for adults and working professionals to gain skills, change fields, and pursue online degrees or certifications.
- MOOCs and other forms: Online resources may improve access to quality education in countries where broad populations face difficulty obtaining it, if people have suitable access tools.
- MOOCs and other forms: Educational apps and online resources can support international educational programs, but electronic-only social contact and some technologies may create adverse effects.The passage also notes reported benefits for autistic children interacting with AI systems.
- LOW-RESOURCE COMMUNITIES: AI research has historically underfocused on low-resource populations, while funders have tended to prioritize commercially applicable work.
- LOW-RESOURCE COMMUNITIES: AI scheduling and planning techniques have been used to distribute surplus food before it spoils to food banks, community centers, and individuals.
PUBLIC SAFETY AND SECURITY
AI is already used for public safety and security and is expected to become more prominent in North American cities by 2030. The report emphasizes both operational gains and risks involving bias, privacy, surveillance, and public trust.
- By 2030, North American cities are expected to rely heavily on AI for surveillance, anomaly detection, drones, and predictive policing.The panel stresses that public trust will be crucial because policing applications may become overbearing or pervasive.
- AI analytics has succeeded in detecting white-collar crime such as credit-card fraud, while machine learning is also affecting cybersecurity.
- Existing cameras are generally better at solving crimes than preventing them because event identification remains weak and video streams exceed available human attention.
- Improved video classification and automated anomaly detection could strengthen crime prevention and prosecution, potentially including evidence of police malpractice.These improvements could also lead to more widespread surveillance.
- Machine learning enhances predictions about where and when crimes may occur and who may commit them, while well-deployed tools may reduce bias in human decision-making.The report also notes the risk of unjustly targeting innocent people.
- AI social-network analysis may help identify radicalization risks and detect disruptive-event plans, but crowd and security applications remain active areas of development.
- Security agencies are expected to increase AI use for efficiency and efficacy, including vision, speech, and gait analysis for detecting possible deception or criminal behavior.The TSA’s DARMS project illustrates risk-based airport screening, but datasets may reproduce prior bias.
EMPLOYMENT AND WORKPLACE
AI is expected to reshape employment gradually by replacing tasks, creating jobs, and changing organizational scale, while its broader distributional effects remain difficult to predict. The report also connects AI-enabled work and entertainment to questions of access, trust, and concentration of influence.
- Assessing AI’s current employment effects is difficult because recession, globalization, and non-AI digital technologies have also reshaped production and jobs.
- AI is expected to replace tasks rather than entire jobs in the near term while also creating new kinds of employment.The report says new jobs may be harder to imagine than existing jobs likely to be lost.
- AI may reduce the need for large organizations when scalability no longer requires adding human labor across locations or management hierarchies.
- AI can create jobs by increasing the importance of some tasks and enabling new modes of interaction, markets, and participation.Examples include app stores, Airbnb, and TaskRabbit.
- AI-driven labor substitution may lower the cost of goods and services, but job loss is more salient to affected people than diffuse economic gains.
- A rapid replacement of all human jobs within one generation is described as highly unlikely, while gradual AI diffusion across employment sectors is expected.
- Education, retraining, new goods and services, expanded social services, or basic income may mitigate longer-term employment effects.The report notes that Switzerland and Finland have considered such measures.
- AI-enabled media can micro-analyze and micro-serve content to individuals, raising concerns about concentrated control over ideas and online experiences.
AI POLICY, NOW AND IN THE FUTURE
The report expects AI to advance gradually but warns that failures, unequal access, bias, privacy concerns, and weakened human capabilities can shape its social effects. It recommends stronger expertise, research, and interdisciplinary policy work to guide deployment.
- AI development is expected to proceed gradually, although small technical improvements can sometimes produce novel, game-changing applications.
- AI systems’ mistakes may trigger user backlash and reduce trust, especially when embedded in critical tasks such as driving.
- When machines take over tasks, people’s ability to perform those tasks may weaken, even though humans and AI retain complementary abilities.
- Early exposure to AI may improve children’s interactions with these systems while widening generational differences in perceptions of AI’s social influence.
- Unequal access to AI, computation, and data may widen existing inequalities of opportunity by improving the abilities and efficiency of those with access.
- Designer and user biases in data and systems can deepen social biases and create fairness problems for groups such as women and people with accents.
- Pervasive AI surveillance raises privacy concerns, while disagreements about bias, privacy, and wealth distribution are expected to resist quick resolution.
- The panel recommends building government AI expertise, removing barriers to fairness and security research, and increasing interdisciplinary funding on societal impacts.These recommendations are presented as responses to individual and societal concerns about rapidly evolving AI.
AI technologies. Private
AI applications raise context-dependent legal, privacy, accountability, and labor questions as they enter products, services, and public systems. Existing regulatory structures do not comprehensively address these issues, while AI’s employment effects remain uneven.
- Policy and legal considerations: AI governance remains fragmented because applications fall under different regulators and rules, while comprehensive treatment is unlikely in the near term.The report notes that autonomous vehicles, medical diagnoses, trading, and tax advice can each face different legal regimes.
- Privacy: Privacy risks arise when AI predictions use proxies correlated with sensitive traits, even when race or sexual orientation are not directly supplied.The report identifies credit-risk and recidivism prediction as examples requiring careful design, testing, and deployment.
- Liability: AI applications can create criminal-liability questions about whom to hold accountable when their behavior would constitute a crime if performed by a human.The issue centers on criminal law’s concern with intended harms and mens rea.
- Labor: As AI substitutes for human roles, some jobs will disappear and others will emerge, but labor-market gains and losses are unlikely to be distributed evenly.Demand, employment, and wages may decline for people whose skills become less needed, with outcomes also depending on government policy.
- Public finance: Municipalities may lose revenue from speeding and parking tickets if autonomous vehicles reduce or eliminate such violations.Government bodies facing budget pressures may respond with legislation that slows or alters AI deployment.
Like other technologies,
The report argues that AI policy should balance innovation, privacy, accountability, and equitable distribution of benefits. Because future effects are uncertain, governance should rely on transparency and enforcement while being continually reassessed.
- Guidelines for the future: Overly restrictive regulation could stifle innovation or relocate it to other jurisdictions, while misunderstanding AI could fuel opposition to beneficial technologies.The report presents both outcomes as counterproductive consequences of poorly calibrated policy.
- Innovation policy: Strict, detailed privacy rules can encourage corporate compliance mentality, whereas ambiguous goals paired with transparency and enforcement better support proactive privacy protection.The comparison covered privacy regulation in European countries, the United States, and Germany.
- Innovation policy: Broad mandates and transparency can build professional privacy processes, stakeholder engagement, and outside accountability.Civil society groups and media can become credible enforcers, increasing privacy’s salience to corporate boards.
- Guidelines for the future: AI regulation should encourage creativity and equitable sharing of benefits rather than concentrating power and benefits among a fortunate few.The report calls for informed debate about steering AI in ways that enrich society while preserving innovation.
- Guidelines for the future: AI policy must be continually re-evaluated because future technologies and their societal effects cannot be foreseen with perfect clarity.The report ties reassessment to observed societal challenges and evidence from fielded systems.
APPENDIX I: A SHORT HISTORY OF AI
The report’s history appendix explains that AI’s development has been shaped by changing technical perspectives and remains incomplete. It emphasizes that data-intensive, big-data approaches have not solved every problem.
- APPENDIX I: A SHORT HISTORY OF AI: The appendix is primarily based on Nilsson’s history and reflects the field’s current emphasis on data-intensive methods and big data.It explicitly states that a complete and fully balanced history lies beyond the appendix’s scope.
- Scope: The appendix cautions that the prevailing big-data focus has not yet demonstrated itself to be a solution to all AI problems.This is presented as a scope and perspective limitation rather than a rejection of data-intensive methods.
- Origins: AI began as an effort to make machines simulate aspects of intelligence, with the field formally named at the 1956 Dartmouth workshop.The workshop was organized by John McCarthy, who is credited with the first use of the term artificial intelligence.
The field of Artificial
AI developed from symbolic reasoning and early computational ideas into data-driven learning, robotics, and other specialized subfields. The field’s history includes an AI winter and a later resurgence enabled by better real-world sensing, actuation, and data.
- The field of Artificial Intelligence: Early AI pursued formal models, symbolic reasoning, heuristic search, knowledge representation, computer vision, and expert systems.These efforts ranged from theorem proving and general problem solving to specialized repositories for chemistry and medical diagnosis.
- AI winter: The AI winter followed limited practical success, insufficient grounding in environmental signals and data, and overreliance on Boolean logic.The report describes funding and interest declining in the mid-1980s after these shortcomings became apparent.
- Resurgence: A 1990s resurgence treated Good Old-Fashioned AI as inadequate for end-to-end systems and emphasized solving tasks from the ground up.Cheaper, more reliable sensing and actuation hardware made systems driven by real-world data more feasible.
- Traditional subareas: Traditional AI subareas include search and planning, knowledge representation and reasoning, machine learning, multi-agent systems, robotics, and human-robot interaction.The report notes that some areas are currently hotter than others but may re-emerge in the future.
- Machine Learning: Machine learning enables systems to improve task performance by observing relevant data, supporting applications from recommendations and speech recognition to fraud detection and image understanding.The report identifies machine learning as a key contributor to AI’s recent surge and the scaling of services such as e-commerce.