Midterm
Define robotics.
*A programmable machine that can complete a task *must have: a mechanical aspect, electrical components, code
What are the consequences of ethical risks
*AI creates many ethical risks that need to be mitigated *AI scales - risks affect & can harm a large number of individuals *These ethical risks create significant reputational, regulatory, and legal risks that are costly in terms of money, resources, and money to address and loss of reputation and consumer trust *Some techniques for mitigating AI ethical risks can be -Legally compliant but ethically risky -Ethically sound but illegal AI
What do we mean by AI for not bad? What is the distinction with AI for good?
*AI for good: To create positive social impact *AI for bad: -Avoid ethical pitfalls in one's pursuit of one's goals -risk mitigation
What are some ways in which CAs have been classified?
*Based on Underlying Technology- rule based or ML based -rule based: operate based off a specific set of rules; Can only respond to very specific commands *Based on Purpose- informational vs transactional *Based on Context and Mode of communication- -context: general vs domain specific -mode: text based vs speech *Based on Embodiment- physical vs virtual
Why is it not advisable to focus on consumer ethical beliefs (perceptions) as the basis for the organization's AI ethics program?
*Consumers' perceptions are too coarse-grained for the fine-grained problems you are facing *Your problems are ones that your consumers have not even thought about yet *Consumers are looking for ethical leadership, and a mere appeal to the sentiment of the day does not meet the bar *The approach will alienate both those who are not particularly concerned aboutthe ethical risks of AI within your organization and those who are, leading to a lackof compliance and turnover, respectively.
Why are corporate codes of conduct and current regulations not adequate to cover AI ethical risks?
*Corporate codes of conduct cover employees' behavior, but AI ethical risks are not realized because of bad behavior *They result from: -Not thinking through the consequences -Not monitoring AI "in the wild" -Not knowing what one should be on the lookout for when developing or procuring AI
What are the components of a basic chatbot architecture?
*Entities: thing the user is talking about *Intents: the question the user asks the CA *Responses: answer th CA provides
Three reasons why ethics are misperceived as subjective:
*Ethics is subjective because people disagree about what is right and wrong *Science delivers us truth. Ethics isn't science so it doesn't deliver us truth *Ethics requires an authority figure to say what's right or wrong; otherwise it is subjective.
What is social response theory and how does it relate to anthropomorphism?
*Humans respond socially to things that display human-like characteristics and provide social cues (e.g., speech, interactivity, etc.) *relates because the more anthropomorphic, the more/better humans will respond
What is anthropomorphism and why do we do it?
*Imbue CAs with human-like characteristics, motivations, attitudes, intentions, & behaviorsSo that they can: (a) present a more compelling experience (b) mimic human-human relationships to build rapport, trust, etc., (c) better perform their objective *we do it to make the bots seem less cold
What do users see as the main challenges with CAs? How are these challenges addressed through design features?
*Natural Language Understanding *Unable to Handle Complex Issues *Unhelpful *Lack of Personal Experience/Empathy *addressed through: -anthropomorphism- give CA's human-like features so they can present a more compelling experience, mimic human relationships to build trust, and better perform their object -conversational design- use this to understand intent to fulfill requests to provide social cues to facilitate interactions and promote certain behaviors
What makes a CA ethical?
*To be ethically justified: (1) the individual's freedom of choice should be preserved (2) the intention behind the design should be transparent (3) the goal-oriented intention of the interference should be justified
What are the objectives of conversation design?
*To understand intent to fulfill request *To provide social cues to facilitate interaction and promote certain behaviors
What factors have fueled the growth of CAs?
*advances in AI *chatbot platforms -ex: after launching its new messenger platform, more than 100k bots appeared on Facebook within the first year
What are the conclusions of prior automation studies? Why is there a resurgence of automation anxiety?
*automation substitutes, compliments, and also alters jobs *past concerns can't settle arguments about the future because: -greatly improved computing power -Ai including robotics -machine learning *raises the possibility of replacing labor on a scale not previously observed which increases automation anxiety
What kinds of jobs are more likely to be automated/substituted by AI?
*codable routine tasks are being substituted by AI *what's less/not substituted- non routine abstract tasks and manual tasks requiring situational adaptation
What is the difference between content and structure for developing AI ethics programs?
*content- reflect a deep understanding of all the ethical risks we are trying to avoid -explainable AI -fair & unbiased AI -privacy *Structure- formal mechanisms for identifying and mitigating ethical risks -governance structure- Policies, Processes, Roles, and responsibilities
What elements of conversations do we design?
*conversation style- how something is being communicated -Expression of emotions -Variability in syntax and word choice -Politeness *content- elements that relate to the literal meaning of a message -Self-disclosure to promote reciprocity -Conversation repair strategies (apology, compensation, providing options) *adaptivity-verbal adaptation of the CA to the users -Tone awareness, facial expressions -Emotional Intelligence
What are some of the main challenges of AI and the current approaches to address them?
*data availability (volume, quality, labeling) *robustness -brittle- when the environment changes and the algorithm can't perform correctly -brittleness is exploited by adversaries *computational load is costly *interpretability- can't explain the results *difficulty generalizing tasks even if they're similar *approaches: transformers, turning data and models, federated learning, unsupervised reinforcement learning to mitigate need for labeling
What is the impact on the economy? What are the expected short and long term impacts of AI on labor markets and jobs?
*effects : -efficiency gains through automation -extending lifespan of old capital through better detection and identification of potential breakdown -innovation *job loss, job creation, job reconfiguration, different skills/reskilling *short term pain for long term gain
What are the main organizational and societal challenges that need to be addressed?
*employment- it will get rid of jobs but also create new ones *bias- make sure we don't create biased algorithms *filter bubbles- AI is good at pattern recognition; tiktok only shows videos they think you'll like so you won't be shown things in other peoples' perspectives *missing human touch/connection *autonomous weapons- bombing locations/ people using facial recognition *environmental impact- server farms take up the size of multiple football fields and use a lot of energy to cool off the servers *automated attacks- DDOS; AI is getting better at their wording in phishing attacks
What are the expected impacts on the economy in terms of productivity and economic growth?
*estimated additional global economic activity of around $13 trillion by 2030 *about 16% higher cumulative GDP compared with today *1.2% additional GDP growth/year
What does polarization of labor mean? Why does it occur?
*hollowing out the middle class because demand and wages for top (cognitive) and manual (bottom) jobs remain high while middle jobs decrease
What are some of the main benefits of robotics?
*improved safety- can operate in hazardous conditions *improved efficiency and productivity- operate 24/7, no fatigue *enhanced precision *greater flexibility- can be repurposed
Differentiate between intended or anticipated, and unintended risks
*intended risks are intentional like using AI to hack someone through phishing emails whereas unintended were not intended to cause harm like dat
What is physical appearance anthropomorphism and what is behavioral anthropomorphism?
*physical appearance- human identity; name, gender, voice, personality, smile, race *behavioral- verbal vs non-verbal
What are the strengths and limitations of human intelligence?
*strengths: emotionally intelligent, intuitive, culturally sensitive, adaptable, creative, judgement *limitations: slower, error prone, subject to fatigue, heuristics & biases, limitations in information processing
What are the strengths and limitations of machine intelligence?
*strengths: fast, accurate, consistently rational *limitations: no emotional intelligence, not intuitive, not culturally sensitive
What factors explain why aggregate economic impacts will take time to manifest?
*takes time to manifest because: there needs to be significant adoption, complementary business and process capabilities are required, and current applications focus on improving existing processes-impact may be invisible outside of efficiency gains
Discuss the disparate effects that AI will have on countries, organizations, and workers. What are some of the factors that explain them?
*the effects are based on AI readiness and other levers *the gap will get wider *factors: strong digital database, more digitized, computing power, specialized talent, expertise in analytics, innovative uses, etc.
How does social response theory relate to ethical vs. unethical anthropomorphism?
*there are CAs being designed to exploit our deeply ingrained, human reactions to human appearances and behavior
Walmart JetBlack Case: Understand why Jetblack, while being sound from a technological perspective, failed.
*they tried to mass distribute luxury brands even though the whole point of luxury items is for them to be scarce *they didn't make sure their business ideas fit into their business portfolio *ur CC should bring synergy to the core business *ur CC must be contingent with ur business positioning
What are the short term transitional costs and negative externalities that accompany a transition to an AI enabled economy?
*transition and implementation costs: displaced workforce (ex: severance pay, PTO), implementing solutions (ex: system integration fees), hiring new workers, continuous upgrade of skills *negative externalities: skilling and reskilling to reenter the workforce, loss of domestic consumption (ex: unemployed individuals consume less), loss of economic contribution (unemployed individuals don't produce economic output), unemployment benefits
What outcomes do CAs impact?
*user/customer: -need satisfaction -improved decision making -efficiency -personalized service *team -coordination -communication -interaction -cohesion *firm -sales -cost reduction -brand reputation -effectiveness *employee/job -change in job structure -hybridization -substitution -turnover
What are some examples of beneficial bots?
- Content Moderation - Chatbots
What are the main benefits of CAs to users?
- No user training is needed- user efficiency- available 24/7
What are the main benefits of CAs to organizations?
- Reduce Operational Costs-Empower Human Agents- Chats can be reviewed for Improvement
Based on AI consumption and AI production landscape, describe the different categories of companies
1) AI giants- end to end integration across the AI stacks; can produce AI for internal/external use (ex: Google) 2) AI creators- can produce/customize AI for clients but rely on AI giants for basics 3) AI powered operators- leverage AI in day to day operations using AI giants & internal capabilities to produce AI (ex: Facebook) 4) AI traders/integrators- purchase and sell off the shelf AI solutions or use cases, adding commercial/marketing efforts (bundling, branding) or support integration without changing anything 5) AI takers- leverage only or mostly off the shelf solutions
Describe the main types of robots
1) AMRs- (autonomous mobile robots) mobile solutions that require limited human input to do their job -move around and make decisions in real time -sensors and cameras help them ingest surroundings 2) AGVs- (automated guided vehicle) portable robot that follows along marked lines or wires, or uses radio waves, vision cameras, magnets, etc. 3) Articulated Robots- robotic arms meant to emulate human arms -2-10 joints -good for welding, packing, etc 4) Humanoids- robots that perform human-centric functions and often look like humans -use the same technology as AMR so falls under that category 5) Cobots- (augmenting robots) function alongside or directly with humans -used to eliminate manual, dangerous, or strenuous tasks 6) Hybrids-when the robots mentioned above are combined -ex: AMR might be combined w/ a robotic arm
Discuss the main CA biases that have been identified and ways to mitigate them
1) Anthropomorphism Bias -ex: Propagating gender stereotypes like using women assistants -Can mitigate by creating genderless voice assistants and having more gender representation in assistants 2) Natural Language Understanding Bias -Not understanding certain accents -Can mitigate by having more representative and inclusive training data 3) Response Generation Bias -Generated by word embedding & translation -Examples: Pronouns used for certain professions 4) offensive dialog- -Generated because of learning from social interactions on social media (e.g. Tay) -Can mitigate by: censoring and limit the amount of influence users can have over the bots' behavior
What are the big three AI ethical challenges? What is the risk of focusing only on these three)?
1) Bias 2) Lack of Explainability 3) Privacy *if you focus on only these 2 you'll overlook other challenges
What has led to the explosive growth of AI applications?
1) explosion of data 2) computing power 3) new algorithms (especially neural networks and deep learning)
What are some modes of Humans-AI collaboration? What factors would influence which one you use for a specific situation?
1) human in the loop- human makes decisions and AI provides decision support 2) human in the loop for exceptions- most decisions are automated; human handles only exceptions 3) human on the loop- AI makes decisions but human reviews the outcomes and makes adjustments 4) human out of the loop- machine makes decisions; human sets new constraints and objectives and monitors *factors that influence: routineness of decision, predictability of environment, # and complexity of external factors, volume of decisions, explainability
What are the four forms of intelligence and how do they help us understand job replacement and how they evolve over time?
1) mechanical -minimal degree of learning/adaptation -precise, consistent, efficient -skills that require limited training -repetitive, consistent, simple tasks -ex: mcdonalds kiosk -where we are now 2) analytical -learns and adapts based on data -logical analytical rule-based learning -tech skills requiring training and expertise on data and analysis -ex: IBM chess player 3) intuitive -learns and adapts based on understanding -artificial neural network based -complex, chaotic, and idiosyncratic tasks -ex: alphago 4) empathetic -learn and adapt empathetically based on experience -emotion recognition -high touch serves; tasks that require empathy/emotional labor -ex: chatbots communicating with customers and learning from it *how they evolve over time -> look at that little graph pic in the slides
What are the main types of AI based on ability? What are the benefits/pitfalls?
1) narrow AI- dedicated for one task *also known as weak or specialized AI *brittle- doesn't perform well outside their skill *what we've developed so far *ex: siri, alexa, spam, Netflix recommendations 2) general AI- performs like a human *a universal algorithm for learning and acting in any environment *also known as strong AI *doesn't have brittle boundaries 3) super AI- more intelligent than a human -------------------------------------- *job loss/creation, reskilling, takes a long time to adopt, costly, people aren't educated about it which is why they see it as a threat
Describe the four types of AI based on functionality
1) reactive- able to perceive and react to the world in front of it as it performs limited tasks, doesn't know past/future events, and reacts identically every time (ex: spam) 2) limited memory- stores past data/predictions to make new predictions and learns based on experience over time; machine learning (ex: self driving cars) 3) theory of mind- makes decisions based on perceptions of how others feel; is emotionally intelligent so they adjust behaviors based on emotions 4) Self-Awareness- able to operate with human-level consciousness and understand its own existence
What can we smooth the transition to an AI economy and address distributional challenges?
1) reskill the workforce through: education institutions, work programs, online resources, government initiatives 2) educate the workforce through: STEM, social, and problem solving skills 3) labor marketplace platforms- matching worker skills to jobs 4) fluid work arrangements 5) policy- like workers compensation
Describe the AI ecosystem
1st, AI technology is underpinned by enablers, which include physical infrastructure (e.g., chip technology) and data management and processing. 2nd, AI enablement supports the AI development environment, which encompasses platform technologies (e.g., AWS or Google TensorFlow) or other visualization software (e.g., Facets, TensorWatch, Tableau). 3rd, AI use cases developed in these environments can be deployed in conjunction with industry-specific applications to support businesses in optimal resource allocation or personalization. We consider AI consumption to be the use of analytical solutions in an industry application, thus turning the latent possibilities of AI into a specific output
How long does it typically take to reskill the workforce?
5-10 years
What are the ethical considerations in anthropomorphizing?
Dishonest anthropomorphism: CAs being designed to exploit our deeply ingrained, human reactions to human appearances and behavior.
What is social presence theory?
Feeling others being present in that conversation, improves the communication
What is AI?
Intelligence demonstrated by machines; thinking humanly and rationally
What is the uncanny valley and when does it occur?
as bots become more human-like, our reactions increase and then decrease because we start getting spooked by it
What are some of the challenges in Human-AI collaboration?
challenges: 1) algorithmic aversion- people have a bad attitude towards AI because of low trust so they ignore/avoid algorithmic advice 2) new leadership skills needed- align human and machine, create inclusive and healthy team environment and team processes 3) training to understand how AI works and how to create human AI teams for better outcomes
What are the market trends (in general) in terms of CAs?
chatbot market in 2024 is predicted to be $1.34 billion; by 2027 it's $1.95 billion
Define social bots
computer algorithms that automatically produce content and interact with social media users
Define CAs.
computer programs designed to communicate with humans using natural language in form of a conversation
What is the substitution and augmentation discussion about?
computers substitute for workers in performing routine, modifiable tasks while amplifying the comparative advantage of workers in supplying problem solving skills, adaptability, and creativity *allows for less time crunching information and more on interpreting and applying it *there are challenges to substituting machines for workers in tasks requiring flexibility, judgement, and common sense remain immense
What are benefits of conversational commerce?
convenience, personalization, time saving
CAs present a significant opportunity for organizations. What is the benefit of shifting from "clicks" to "conversations?
drives the switch from technology-literate people to people-literate technology
Walmart JetBlack Case: What is conversational commerce?
e-commerce done via various means of conversation and using technology such as: speech recognition, speaker recognition, natural language processing and artificial intelligence
What is needed to harness the benefits of AI?
entrepreneurship, effort, restructuring
How does anthropomorphism relate to social presence theory?
increasing social presence will make the bot seem more anthropomorphic
When you designed a chatbot using dialogflow, which chat bot components did you have to specify and which of these were provided by dialogflow?
specify the intent, entities (some are provided already), response (some are provided for you)
Why are misperceptions about the nature of ethics a major obstacle to organizational buy-in to developing AI ethics programs?
they often put a stop to fruitful discussions and are an impediment to genuine organizational buy-in
On what dimensions do various types of robots vary?
vary in functionality, design, degree of autonomy (human controlled to fully autonomous)
What type of tasks are most suitable for AI at present?
well defined inputs, large data sets, defined goals, no reasoning needed, tolerance for error, no emotional intelligence needed, no common sense needed, predictable environment
