AI Exam Flashcards
General Purpose Technology
Affects the entire economy and can drastically change society
Problems with the line shaft system
Inefficient energy transmission (big energy waste)
2. All-or-nothing system
1. Machines could not be used without all of the power being on
2. The engine or shaft breaks, everything stops
3. The belts
1. Dangerous!
2. Blocked light
3. Lots of noise
4. Power distribution dictated the layout of the factory
5. Building size was limited
* All of this significantly limited the productivity of factories
The electrification of the factory & reasons for slow progress
At the start people replaced big steam engines with big electric engines in the same factory
* It took DECADES before the factory was redesigned to exploit the benefits of electricity
1. Move to group drive and smaller electric engines for groups of machines
2. And eventually unit drive: 1 machine = 1 engine (engine inside the machine)
* Reasons for slow progress:
* Habit/inertia
* Installed capital
The lag theory of General Purpose Tech
Electrification led to a big increase in productivity, but it took around 40 years for people to fully realize the gains
* The same (but shorter lag) happened with computers in the 70s-90s
* Leading economists argue the same is happening with AI today. Full exploitation of GPT requires organizational, psychological, social,
societal adjustments
* Tech exploitation is as much a human as a technical challenge
* This means that it usually takes a long time before a GPT impacts productivity stats
* (Another reason for why we need a behavioral science course on AI)
AI as General Purpose Tech: Key take-aways
- AI will change every industry & create huge benefits for mankind
- But it’s going to take time before we learn to fully benefit from AI
- Those who get there first will make fortunes
- The barriers are “human” more than technical
Effective integration of artificial and human intelligence is crucial to benefit from the strengths of each
“The New Diversity”
“The New Diversity”
- Effective integration of artificial and human intelligence is crucial to benefit from the strengths of each
- “Weak human + AI + good process” > “Strong human + AI + bad process”
Sequential approach
- AI screening to increase efficiency and effectiveness
What is the big picture
The objective should not merely be to automate tasks that humans are currently doing, but to automate tasks that humans are not currently doing
* To generate value with AI by imagining new uses for recording and matching capabilities
Logic behind CAPTCHA
Make image recognition difficult for machines
* Require more than purely representational skills (matching)
Representation
Recording and matching
Interpretation
Slice and explain
Slice and explain
Interpretation
Marketing is the nexus between
Supply & Demand
Customer Orientation
Decisions are made based on delivery of customer benefits (relative to competition), not product performance
Define Consumer AI
An ecosystem of three elements:
1) Data collection & storage (Listening)
2) Statistical & computational algorithms (Predicting)
3) Output systems (Producing & Interacting)
that enable products and services to perform tasks typically understood as requiring intelligence and autonomous decision making on behalf of consumers
AI Dystopias
Cautionary tales outlining the dehumanizing effect of AI on our existence (freedom, privacy, dignity, etc.) that, when circulated in popular culture, can stimulate anxieties towards Consumer AI
Technophilic Myths
stories that highlight how technology enables a better, more efficient, and fairer societ
Technophobic Myths
Stories that highlight the dark side of technology for business, society and individual
Designing Customer Experiences
- Optimize affective, cognitive and behavioral responses
- Optimize the customer experience across touchpoints
- Maximize positives, minimize negatives
- Exciting products, inspiring advertising, amazing buying environments, …
- Build customer relationships and brand affinity
Stages of customer experience
Undifferentiated, Competitive Position, Differentiated
1) Extract Commodities
2) Make Goods
3) Deliver Services
4) Stage Experiences
Problems with Listening: Data-Capture Experiences
Served vs exploited?
- Sociological Background: Surveillance Society
- Psychological Reaction: Loss of control
Problem with Producing: Delegation Experiences
Empowered vs replaced?
- Sociological background: Transhumanism
- Psychological concern: Loss of self-efficacy
What is the problem with Predicting: Classification Experiences
Understood vs Misunderstood?
- Sociological background: Unequal worlds
- Psychological reaction: Fear of mistakes/discrimination
Interacting: Social Experiences
Connected vs Alienated
- Sociological background: Humanized AI
- Psychological concern: Objectification/Dehumanization
Creative Vs Emotional Work
Creative Work
“With automation forecasted to eliminate many repetitive jobs, humans will move up the value chain and carry out more thoughtful, creative work”
“Why is [creativity] the most important skill for professionals today? … Today, basically anything that can be automated has been automated or soon will be automated, which cuts down on a lot of process-orientated tasks.”
Emotional Work
Jobs that require sensitivity, empathy, moral
Recording
capturing the data and storing it
Camera captures light on sensitive film
Slicing
taking elements of what is collected
Van Gogh’s painting with the town & galaxies
Matching
- using algorithms and machine learning to diagnose the information
Explaining
interpretation, understanding what makes something happen next
Causal inference
AI vs Human INtelligence
Artificial Intelligence - recording and matching helps us expand our known knowns from previously known unknowns
Human Intelligence - slicing and explaining helps us expand our known unknowns to unknown unknowns
Picasso and Socrates emphasize that we can answer the “why”, which machines cannot perform well
Limits of AI
We live in a causal world
AI representation provides intelligent outputs based on predictability for next word but there is no agenda, no mental models, still representation (recording and matching)
B.S. speech intended to persuade without regard for truth
Two Archetypes
Discovery as Epiphany
Light bulb goes off and make a connection
*Cognitive flexibility
- creativity as association process
- more holistic thinking enables more distal associations
- Outside the box
- Creativity as interpretation
Discovery as Serendipity
Chance favors the prepared mind, search costs involved
*Cognitive persistence
- Creativity as a stochastic process
- More extensive search enables more creative output
- Inside the box
- Creativity as representation
Insight vs. Effort: Communicating the Creative Process to New Products
Consumers hold associations between insight and arts, between effort and sciences
Consumers evaluate more favorably artistic and scientific products presented as the outcome of insight or effort
Problem with data driven decision making
Correlation does not equal causation (Ice Cream → Shark Attacks, Adidas, World’s Dirtiest Man)
Hard to measure ROI - Facebook Small Businesses vs. Apple
Selection Effects
Air plane example
Two trends in managerial decision making
Behavioral Science - More Aware of cognitive biases
Computer Science - Greater Ability to Store and Analyze Data
Problem with Data Driven Analytics
The name - should be decision driven analytics.
1) take data (scientists) at face value.
- Correlation is not causation
- Hard to measure ROI
- Selection Effects
2)Answer the wrong questions
AB test instead of historical data
Evolution of decision making
Intuitive Management; Big data analytics; decision driven analytics
Factual and couter factual questions
Prediction (Match) and Intervention (Explain)
Prediction vs intervention
In politics, the difference between persuadables and already strong supporters
Decision Driven Data Analytics
- Anchors on the decision to be mad
- Finds data for a purpose
- Starts from what is unknown
- Focus on action/impact
- Empowers Decision makers
consequences of human replacement depend on
context and technology
The psychology of human replacement (4 components)
Attribution
social comparison
perceived control
self awareness
Attribution (psych)
How do people infer causality (central to interpreting the process of human machine co creation). People care not only about outcomes but also the process leading to them. People value the opportunity to carry out tasks that are more central to their identity. Think about the cooking machine study. WHO COOKED DINNER?
- Internalizing positive outcomes is easier with human decision makers
because algorithms neglect individuals’ uniqueness (Longoni et al. 2019) - But we externalize negative outcomes regardless of the decision maker
- Uniqueness neglect vs. biased/unreliable
- Therefore, less positive response to algorithmic acceptance and
difference mitigated in the case of rejection
Social Comparison
Natural tendency to compare one’s performance ot that of others: upward vs downward. Relevence: Human replacement should be expected to trigger social comparisons
- Similarity to comparison target important antecedent of social
comparison - Robots induce less self-relevant comparisons, inducing less self-
threat - People thus have a psychological incentive to prefer robotic
replacement over human replacement - (But: Robotic replacement = more future concerns)
41
perceived control
Autonomy and control over one’s life are fundamental human desires
(Wertenbroch et al. 2020)
* Control is a situational appraisal
* Relevance: Automation takes over tasks we previously performed,
and this can create a perceived loss of control
- nicknaming brands and [products is a way to cope/
Self-Awareness
Increased sense of self-awareness
makes salient self-standards
* Relevance: Human enhancement expected to trigger greater self-
awareness and motivation to live up to self-standards
- Instagram study, appearance awareness made people more aware of their donation behavior.
Self-serving bias:
- Internal attribution of positive outcomes
- External attribution of negative outcomes
Symbolic consumption/Uniqueness motives
- Human labor associated with antecedents of symbolic value (e.g.,
scarcity, imperfections), and robotic labor to the opposite (e.g.,
standardization) - Human labor creates value for symbolic consumption because it
helps satisfy one’s need for uniqueness
tattoo study
- making a tattoo (symbolic)
- removing a tattoo (instrumental)
Stereotypes
Shared Beliefs/preconceptions about a person
- People hold believed about what machines can and cannot do
Algorithmic management
tends to objectify workers, decreases prosocial behavior
The morality of human replacement
attrubution, self expression, morality, dehumanization, stereotyep
