A FUTURE THAT WORKS: AI, AUTOMATION, EMPLOYMENT, AND PRODUCTIVITY JAMES MANYIKA Extracts From McKinsey Global Institute Research, June 2017 CONFIDENTIAL AND PROPRIETARY Any use of this material without specific permission of McKinsey & Company is strictly prohibited
Amazing progress in AI and Automation McKinsey & Company 2
McKinsey & Company 3
2011 2016 Humans 26% errors 5% errors 3% errors SOURCE: Jeff Dean (Google Brain) McKinsey & Company 4
McKinsey & Company 5
McKinsey & Company 6
Why Now? Algorithms/techniques – Neural Networks, CNNs, 1 RNNs, Deep learning, Reinforcement Learning… Compute power – Silicon (CPUs, GPUs, Tus …); 2 Hyperscale compute capacity, cloud available … Data – 50 exabytes (2000), 300 exabytes (2007); 3 4.4 zettabytes (2013), 44 zettabytes (2020) … McKinsey & Company 7
Huge benefits to business, the economy and society McKinsey & Company 8
Machine learning has broad potential across industries and use cases Size of bubble indicates variety Agriculture Consumer Finance Manufacturing Pharmaceuticals Telecom of data (number of data types) Automotive Energy Health care Media Public/social Travel, transport, and logistics Volume Breadth and frequency of data 10 Lower priority Personalize Identify Personalize 9 fraudulent financial advertising Higher potential products transactions 8 Identify and navigate roads Discover new Personalize crops to 7 individual conditions consumer trends Optimize pricing Predict personalized and scheduling 6 health outcomes in real time Optimize 5 merchandising strategy Predictive 4 maintenance (energy) Case by case Predictive maintenance 3 (manufacturing) 2 Diagnose diseases Optimize clinical trials 1 0 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 Impact score McKinsey & Company 9
Good for business – Drives innovation, transformation and productivity ACCURACY OPTIMIZATION PREDICTION SCALABILITY THROUGHPUT CREATION DISCOVERY DECISIONS McKinsey & Company 10
Good for the economy - Automation can contribute to growth in GDP per capita FTE automation output (United States example, 2000 – 65) FTEs Millions Automation will be a 450 FTEs to achieve significant contributor projected GDP growth to the productivity boost needed to projected FTE Automation output GDP per capita growth 360 in earliest scenario FTE Automation output 270 in latest scenario FTEs required to maintain current GDP per capita 180 Projected FTE Assuming zero productivity growth, Historical FTE based on demographic trends, the 90 projected FTEs will be less than the FTEs required to maintain current level of GDP per capita 0 2000 05 10 15 20 25 30 35 40 45 50 55 60 65 2070 Year McKinsey & Company 11
What about jobs? McKinsey & Company 12
Our approach focuses on activities and capabilities of currently demonstrated technologies Occupations Activities (retail example) Capability requirements Social Retail sales- Greet 1 ▪ Social and emotional sensing people customers ▪ Social and emotional reasoning ▪ Emotional and social output ▪ etc Food and Answer questions about beverage service 2 products and services workers Cognitive ▪ Natural language Clean and maintain ▪ Recognizing known patterns / categories work areas ▪ Generating novel patterns / categories 3 Teachers ▪ Logical reasoning / problem solving ▪ Optimizing and planning Demonstrate product ▪ Creativity features ▪ Articulating/display output Health 4 ▪ Coordination with multiple agents practitioners ▪ etc Process sales and transactions ▪ ... Physical ▪ Sensory perception ▪ … ▪ ... ▪ Fine motor skills/dexterity ▪ … ▪ … ▪ Gross motor skills ▪ … ▪ Navigation ▪ Mobility ~2,000 activities assessed ▪ etc across all occupations ~800 occupations SOURCE: Expert interviews; McKinsey analysis McKinsey & Company 13
BASED ON Some activities have higher technical automation potential DEMONSTRATED Time spent on activities that can be automated by adapting currently demonstrated technology TECHNOLOGY % 81 69 64 26 20 18 9 Time spent 7 14 16 12 17 17 16 16 18 18 in all US occupations % Manage Expertise Interface Unpredictable Collect Collect Process Process Predictable Predictable physical data data data data physical physical Total wages 596 1,190 896 504 1,030 931 766 in United States, 2014 51% of US wages Most $ billion susceptible $2.7 trillion in wages activities McKinsey & Company 14
Some sectors have more automatable activities than others BASED ON DEMONSTRATED TECHNOLOGY Size of bubble indicates % of Ability to automate (%) time spent in US occupations 0 50 100 Automation potential Unpredictable Collect Process Predictable Sectors by activity type Manage Expertise Interface physical data data physical % Accommodation and food services 73 Most automatable 60 Manufacturing 60 Transportation and warehousing 57 Agriculture 53 Retail trade 51 Mining Other services 49 Construction 47 In the middle Utilities 44 44 Wholesale trade Finance and insurance 43 41 Arts, entertainment, and recreation Real estate 40 Administrative 39 Least automatable Health care and social assistances 36 Information 36 Professionals 35 Management 35 Educational services 27 McKinsey & Company 15
Employee weighted overall % of activities that can be All countries could be impacted by automation automated by adapting currently demonstrated technologies 45 – 47 47 – 49 49 – 51 <45 >51 No data Million FTE Automatability across economies $ trillion Employee weighted overall % of activities that can be automated Remaining countries China 100% = 1,156M FTEs $14.6 trillion Japan Big 5 in Europe United States India McKinsey & Company 16
A small percentage of occupations can be fully automated by adapting current technologies, but almost all occupations have some activities that could be automated 100 91 73 62 51 42 % of roles 34 26 (100% = 18 8 820 roles) 1 100 >90 >80 >70 >60 >50 >40 >30 >20 >10 >0% Percent of automation potential Sewing machine Stock clerks Bus drivers Fashion designers Psychiatrists Example Travel agents Nursing assistants Chief executives Legislators occupations operators Assembly line workers Dental lab technicians Web developers More occupations will have portions of their tasks automated e.g. While about 5% 60% of occupations could have of occupations could have close to 100% 30% of tasks automated, of tasks automated SOURCE: US Bureau of Labor Statistics; McKinsey Global Institute analysis McKinsey & Company 17
Automation potential spans from high to low wage occupations BASED ON DEMONSTRATED TECHNOLOGY Ability to technically automate Percentage of time on activities that can be automated by adapting currently demonstrated technology 100 File clerks 80 Landscaping and grounds-keeping workers 60 Chief executives 40 20 0 0 20 40 60 80 100 120 Hourly wage $ per hour McKinsey & Company 18
Several factors affect the pace and extent of AI and automation Technical Cost of Cost of labor Benefits Regulatory feasibility and developing and related including and social pace of and supply- and beyond factors breakthroughs deploying demand labor technologies dynamics substitution McKinsey & Company 19
In summary… McKinsey & Company 20
We’ve seen this before— but is this time different? Distribution of labor share by sector in the United States, 1840 – 2010 % 90 Rest of the economy 80 70 60 50 40 30 20 10 Manufacturing Agriculture 0 1840 50 60 70 80 90 1900 10 20 30 40 50 60 70 80 90 2000 2010 McKinsey & Company 21
So with huge benefits, some real challenges to address Benefits Challenges Faster innovation and business Jobs and wages transformation Skills and training Better performance, outcomes, Dislocation and Social and For quality, speed economic transitions businesses Overcome human limits; Solve Distributional issues and users new problems, create new Acceptance opportunities and innovations Safety, utility, quality of life Boost productivity growth, Transparency, openness GDP growth and prosperity and competition For Counter aging or shrinking Biases economies Other issues workforce Safety, Cybersecurity and society S olve “moonshot” problems Ethics (e.g., climate) McKinsey & Company 22
Recommend
More recommend
