Automation

Preview

• Today: Automation
• Oct 19: Geography
• Oct 26: Appropriability + question time

Outline

• Productivity
• Adoption
• Specialization
• Employment

​

Automation

• Work comprises tasks
• Conventional view -- substitution: Machine replaces human in particular tasks
• Structured tasks => Easy to codify and automate

​

Automation: Advantages

• Machines better than humans
• Learning
• Absorptive capacity
• Memory: Fully retains, does not decay
• Scaleable: Freely transferable
• Performing task
• Precise, no random variation
• Need not rest – can work 24 hour shift
• Does not require lighting, heating https://www.youtube.com/watch?v=1aGgXkQCwDc

​

Automation: Costs

• Large set-up costs
• Counter example: Amazon Go store – no economies of scale in hardware (cameras, sensors, scanners)
• Requires skilled managers and engineers

Automation: Limitations

• “[I]t is comparatively easy to make computers exhibit adult level performance on intelligence tests or playing checkers, and difficult or impossible to give them the skills of a one-year-old when it comes to perception and mobility.” (Moravec (1988: 15))
• Autopilot: Cannot manoeuvre from/to gate https://www.youtube.com/watch?v=moUrQM9RyVk
• Abstract tasks: Require problem-solving, intuition, creativity, persuasion => high-end workers
• Manual tasks: Require situational adaptability, visual and language recognition, inter-personal interaction => low-end workers

​

​

Human back-up

• Chat bots: If cannot answer, then refer to human

Automation: Future

• Human back-up: Chat bots: If cannot answer, then refer to human
• Environmental control
• NUS autonomous bus: Fixed route
• Kiva warehouse: Streamlined layout; still need human to pick and assemble cartons https://www.youtube.com/watch?v=3UxZDJ1HiPE
• Machine learning: Not limited to codifiable routines
• Supervised machine learning -- pattern recognition, eg, reading Xrays for tumours; Train the algorithm
• Future: Unsupervised machine learning

Outline

• Productivity
• Adoption [student presentation]
• Specialization
• Employment

​

Kim et al., Home-tutoring services

1. Referring to Table 3, what does the coefficient of After reveal about student performance?
2. Explain why the authors did not simply estimate the effect of View on student performance by using ordinary least squares regression.
3. It has been theorized that users who expect relatively higher benefit would adopt a new technology earlier. What evidence do Kim et al. find for this hypothesis?
4. Discuss the possible network effects in use of the AI system and effects on student outcomes.

​

Law

• Unique factors in adoption of intelligent systems
• Legal responsibility
• Moral responsibility
• Autonomous vehicle – who is responsible for accident?
• Operator of vehicle
• Owner of vehicle
• Manufacturer of vehicle
• Software developer
• Software design choice
• In uncontrolled motion into a crowd of people, which way to turn?

Law

• Criminal liability – especially non-monetary penalties (detention, imprisonment, social stigma) maybe uninsurable
• Civil liability – insurable by employer or third party

Outline

• Productivity
• Adoption
• Specialization [student presentation]
• Employment

​

Complementarities

• Automation enables specialization without coordination costs => increases productivity in human task (Gong and Jie, 2022)
• Ride hail: Driver + map app
• Supermarket checkout: Cashier + self-pay machine
• Note
• Ride hail: Automated the more cognitively demanding tasks
• Checkout: Automated the task requiring more human interaction

Jie Gong and IPL Png, “Automation Enables Specialization”

1. In the textbook example of automation, machines replace human workers in their entire jobs. In automation-enabled specialization, machines replace humans only in particular tasks. Compare the effect of the two types of automation on employment.
2. What does Figure 3 say about the effect of scan-only checkout on productivity? Discuss whether the results are sensitive to outliers.
3. Referring to Table 4, discuss how payment in cash and the quantity of the customer’s purchases affect payment time.
4. Discuss whether the increase in productivity due to scan-only/self-pay system is illusory as it just shifts work to the consumer.

Outline

• Productivity
• Adoption
• Specialization
• Employment

​

Employment

• Commentators mainly focus on labour demand
• David Autor https://www.ted.com/talks/david_autor_will_automation_take_away_all_our_jobs

​

Polarization: Growth in jobs at both ends, shrinking middle

Labour supply

• Task-skill fit
• Activities that fit capabilities are motivating (Bandura 1999: 118, 160-161): Intrinsic motivation
• Fit increases experience of flow
• Individual perceives that activity is rewarding in itself (Barthelmas 2019; Nakamura 2009)
• In activities that are important to the person, flow highest when skills exceed task requirements (Engeser 2008)

Labour supply

• Deskilling
• High-skilled workers: Makes the job dull => reduces labour supply
• Low-skilled workers: Makes the job accessible to more workers (Ong and Png 2022)
• Ride hail drivers: Willing to pay equivalent of 5.1 percent points of platform commission to use map app – concentrated among those with poor geographical knowledge
• Supermarket cashiers: Willing to pay $85 per month to work at scan-only checkout counter

Ong and Png, “Deskilling”

• Cashier vignette experiment: Would you prefer
• Work at conventional counter at same pay
• Work at scan-only counter for $A more per month

Labour supply: Productivity