Render Leanpub

manuscript/02a-Avoiding_Harm-Overall_quiz.md

@@ -0,0 +1,55 @@
+
+{quiz, id: avoiding_harm_overall, attempts: 10}
+
+## Avoiding Harm - Overall Quiz
+
+? Which of the following is TRUE?
+
+a) Unintended harm is alright because it wasn't intended
+b) AI systems are unbiased and should be used to address human biases
+C) Evaluate new AI tools for biases throughout its development and use
+d) AI tools are accessible and useful in an equitable sense
+
+
+? What is one potential issue with using AI systems that lack interpretability?
+
+A) It can be unclear why the system made a certain decision
+b) The systems always behave as expected
+c) There are no risks associated with using them
+d) They cannot be trusted for any applications
+
+? According to the material, what should be considered at every stage of AI system development and use?
+
+a) Funding sources
+b) Model usage
+C) Ethical implications
+d) Marketing strategies
+
+? What can help reduce issues with faulty AI responses?
+
+a) Never questioning the accuracy of AI responses
+b) Relying completely on AI tools without human review
+c) Assuming AI tools reflect current best practices
+D) Cross-checking responses from multiple AI tools
+
+? Why is transparency important when using AI systems?
+
+a) It builds user trust in the systems
+b) It facilitates public monitoring and feedback
+c) It enables better understanding of AI-assisted decisions
+D) All of the above
+
+? What should be done to mitigate algorithmic bias?
+
+a) Nothing, AI systems are inherently unbiased
+B) Train models on more diverse and inclusive data
+c) Disable safety checks that restrict harmful content
+d) Rely completely on AI systems for decision-making
+
+? What helps ensure user compliance to standards for AI use?
+a) Complex, frequently changing policies
+b) Generic training materials
+c) Self-education by users
+D) Relevant training content
+
+{/quiz}

manuscript/02b-Avoiding_Harm-concepts_quiz.md

@@ -0,0 +1,27 @@
+
+{quiz, id: societal_impact_quiz, attempts: 10}
+
+## Societal Impact Quiz
+
+? Which of the following is NOT considered a major ethical consideration discussed in the material?
+
+a) Security and privacy issues
+b) Climate impact
+C) Efficiency of the AI project
+d) Inadvertent harm
+
+? What does the concept of "AI paternalism" refer to?
+
+A) Trusting an AI system over human judgment or patient experiences in healthcare
+b) The use of AI systems as replacements for human parents
+c) Regulations restricting the development of advanced AI systems
+d) Biases inherent in AI systems against certain groups
+
+? According to the material, which of the following can help reduce potential harms from AI systems?
+
+a) Assuming AI systems will behave as expected in all situations
+b) Implementing new AI solutions quickly without extensive testing
+c) Avoiding transparency about AI system development
+D) Involving diverse perspectives in discussions of intended uses
+
+{/quiz}

manuscript/02ba-Avoiding_Harm-concepts_quiz.md

@@ -0,0 +1,28 @@
+
+{quiz, id: testing_training_harm_quiz, attempts: 10}
+
+## Effective Use of Data Quiz
+
+? Which of the following best describes the training data?
+
+A) Data used to initially train the machine learning model
+b) Data used to fine-tune and improve the machine learning model
+c) Data used to evaluate the performance of the machine learning model
+d) Data that is better determine how the model might work in other contexts
+
+? Why is it important to have independent training and testing datasets?
+
+a) To avoid overfitting the model to the training data
+b) To more accurately evaluate model performance
+c) To fine-tune the model using a validation set
+D) Both a and b
+
+? According to the passage, what is the preferable way to partition data for machine learning?
+
+a) Use all data for training only
+b) Use some data for training and the rest for testing
+C) Use some data for training, some for validation, and the rest for testing
+d) There is no preferred way, any partitioning is acceptable
+
+
+{/quiz}

manuscript/02c-Avoiding_Harm-algorithms_quiz.md

@@ -0,0 +1,32 @@
+
+{quiz, id: algorithms_quiz, attempts: 10}
+
+## Algorithms Quiz
+
+? What is one potential issue with using AI systems that lack interpretability?
+A) It can be unclear why the system made a certain decision
+b) The systems always behave as expected
+c) There are no risks associated with using them
+d) They cannot be trusted for any applications
+
+? According to the material, which of the following can help reduce issues with faulty AI responses?
+a) Never questioning the accuracy of AI responses
+b) Relying completely on AI tools without human review
+c) Assuming AI tools reflect current best practices
+D) Cross-checking responses from multiple AI tools
+
+? What does the term "hallucinate" refer to in the context of AI systems?
+
+a) AI systems having visions or dreams
+B) AI systems making up information based on artifacts of their algorithms
+c) AI systems malfunctioning due to hardware issues
+d) AI systems being trained on datasets containing hallucinogens
+
+? What recent real-world example demonstrates issues with faulty AI responses?
+
+a) AI systems spreading misinformation on social media
+b) An AI system beating the world chess champion
+C) A website banning AI-generated code that provided incorrect answers
+d) AI systems writing biased news articles
+
+{/quiz}

manuscript/02d-Avoiding_Harm-adherence_quiz.md

@@ -0,0 +1,24 @@
+
+{quiz, id: adherence_quiz, attempts: 10}
+
+## Adherence Quiz
+
+? According to the material, what is one benefit of starting slow when implementing AI systems?
+A) It allows time to understand how the systems work
+b) It eliminates any possible unexpected consequences
+c) It speeds up the rollout to more users
+d) There are no benefits to starting slow
+
+? What does the material recommend related to copyright when using AI systems?
+a) Assume AI-generated content does not violate any copyright
+B) Obtain permission to use copyrighted content used for training
+c) Never citing sources used by AI systems
+d) Waiting to address copyright issues until after system launch
+
+? According to the material, what can help ensure individuals comply with ethical standards for AI system use?
+a) Complex policies that frequently change
+b) Training materials not tailored to actual use cases
+c) Relying on users to educate themselves
+D) Education emphasizing relevance and fundamentals
+
+{/quiz}

manuscript/02e-Avoiding_Harm-consent_and_ai.md

@@ -0,0 +1,33 @@
+
+{quiz, id: consent_quiz, attempts: 10}
+
+## Consent and AI Quiz
+
+? Why is obtaining meaningful consent tricky with AI systems?
+
+a) Users readily understand all the risks of providing their data
+b) The risks associated with data collection are static and known
+c) AI systems anonymize all collected data
+D) The potential risks to users change dynamically over time
+? Which of the following are concerns with data collection and AI?
+
+a) Accountability
+b) Data persistence
+c) Data reuse
+D) All of the above
+
+
+? Which of the following technologies has been an especially debated topic in relation to consent in the EU?
+
+a) Speech recognition
+B) Live Facial recognition
+c) Object detection
+d) Machine translation
+
+? What is data spillover?
+
+A) A situation where data collection happens accidentally for other individuals in collecting data for intended people
+b) When datasets are too large that the AI system fails
+c) When datasets are too large that too many resources are used to support computing
+d) The deliberate collection of data not specified in an individual's consent agreement 
+{/quiz}

manuscript/02f-Avoiding_Harm-idare_and_ai.md

@@ -0,0 +1,31 @@
+
+{quiz, id: idare_and_ai_quiz, attempts: 10}
+
+## IDARE and AI Quiz
+
+? What does the acronym IDARE stand for?
+
+A) Inclusion, Diversity, Anti-Racism, and Equity
+b) Intelligence, Data, Algorithms, Research, Engineering
+c) Interfaces, Developers, Administrators, Regulators, End-users
+d) Identification, Development, Assessment, Regulation, Evaluation
+
+? Why can bias occur in AI systems?
+A) Humans and human-created data are inherently unbiased
+b) AI systems are intentionally designed to discriminate
+c) The training data for AI systems is too diverse in representation
+d) Bias only occurs when using unsupervised learning
+
+? What example demonstrates an AI system perpetuating existing biases?
+a) An AI system beating a human at chess
+b) AI-written articles discussing IDARE principles
+c) Algorithms that help flag medical images for further review based on patient reported pain
+D) Chatbots giving biased answers that lack diverse perspectives to users
+
+? What does the material recommend regarding using AI for employment decisions?
+a) AI should be fully relied upon to make unbiased choices
+b) Humans should have final say but AI can select a small number of resumes to evaluate without any potential bias issues
+C) AI is not appropriate for use in employment decisions with our current systems
+d) Guidelines are unnecessary as bias is no longer an issue
+
+{/quiz}

manuscript/02g-Avoiding_Harm-ethical_process_quiz.md

@@ -0,0 +1,25 @@
+
+{quiz, id: ethical_process_quiz, attempts: 10}
+
+## Ethical Process Quiz
+
+? When should ethical considerations occur according to the proposed framework?
+
+a) Only after an AI system is already in use
+b) At the beginning of during development only
+c) Just before an AI system is deployed
+D) Throughout the entire process of development and use
+
+? What is NOT a consideration when reflecting on AI use during the inception of an idea?
+a) What training datasets are a reasonable size
+b) What would the impact be if the AI system performed poorly
+c) Whether AI is actually appropriate for the use case
+D) How you could be more efficient and implement faster
+
+? What is needed after an AI system is launched according to the material?
+a) The system should not change after launch
+b) Occasional spot checks for issues
+c) Fixes only if users report problems
+D) Continual audit for unexpected and problematic behaviors
+
+{/quiz}

manuscript/03-Determining-AI-Needs-overall_quiz.md

@@ -0,0 +1,62 @@
+
+{quiz, id: needs_overall, attempts: 10}
+
+## Determining AI Needs - Overall Quiz
+
+? What are the two important qualities of an AI tool that make it good?
+
+a) Accuracy and computational complexity
+B) Accuracy and computational efficiency
+c) Speed and cost
+d) Usage and funding
+
+? What were the three AI tool customizations we discussed in this course?
+
+a) Data, algorithms, and hardware
+B) Knowledge, security, and interface
+c) Inputs, outputs, and models
+d) Servers, networks, and storage
+
+? What should you do first when planning an AI project?
+
+a) Find funding
+b) Assemble a team
+C) Define goals and use cases
+d) Start development
+
+? What size model is ideal for building a custom trained AI model?
+
+a) The biggest model that is available
+b) The newest version available
+C) Small as possible while still maintaining accuracy
+d) An private subscription model
+
+? Which was NOT a best practice for data security?
+
+a) Use authentication
+B) Allow broad data access
+c) Limit privileges
+d) Employ data agreements
+
+? What should you do before using or building AI tools for use with protected data?
+
+A) Consult legal experts
+b) Test extensively
+c) Accept terms of service
+d) Purchase usage licenses
+
+? Why is usability testing important?
+
+a) It identifies areas for improvement
+b) It gathers user opinions
+c) It identifies user needs
+D) All of the above
+
+? How might you decide what metrics are important for you to collect about your AI tool?
+
+a) Make up your own benchmarks to make your tool look good
+b) Use the most flashy metrics
+c) Copy your competitors regardless if it makes sense for your use case
+D) Evaluate your AI model against what types of "ground truth" you have available.
+
+{/quiz}

manuscript/03b-Determining-AI-Needs-components_quiz.md

@@ -0,0 +1,30 @@
+
+{quiz, id: ai_components, attempts: 10}
+
+## Components of AI tools
+
+? What is one way that AI models can become more accurate?
+a) Adding more irrelevant training data
+B) Adding more labeled, relevant training data
+c) Not adding any new training data
+d) Only training the model on text data
+
+? What enables modern AI models like large language models to be efficient?
+a) Using sequential processing algorithms
+b) Using transformer algorithms
+C) Using attention mechanism algorithms
+d) Using encoder algorithms
+
+? Which is NOT a core component of an AI system we discussed?
+a) Input
+b) Training data
+C) Parameters
+d) Output
+
+? Why can AI models sometimes give biased outputs?
+a) They don't have enough training data
+B) Their training data contains human biases
+c) Their algorithms are biased
+d) They don't use transformer algorithms
+
+{/quiz}

manuscript/03c-Determining-AI-Needs-determining-needs_quiz.md

@@ -0,0 +1,42 @@
+
+{quiz, id: ai_needs, attempts: 10}
+
+## Determining AI needs
+
+? What is an important first step when planning an AI project?
+a) Choosing an AI platform to use
+B) Defining the goals and use cases
+c) Assembling a development team
+d) Estimating required computing power
+
+? What does "customized knowledge" mean for an AI project?
+a) Restricting access to sensitive data
+b) Integrating with an existing interface
+C) Training models on specific domains
+d) Modifying algorithms for efficiency
+
+? Who should you definitely consult before starting an AI project that involves protected data?
+A) Legal experts and IRB
+b) Back-end developers
+c) Project manager
+d) User experience designer
+
+? What should be estimated to account for computing costs in an AI project?
+a) Number of users
+b) Size of input data
+c) Queries per user
+D) All of the above
+
+? As compared to using a prepackaged AI tool, building a custom solution from scratch will likely require?
+a) Less technical expertise
+b) Shorter development time frames
+c) Lower overall budget
+D) More technical expertise and longer time frames
+
+? What resource might you seek out if you need to build a fully customizable AI solutions including trained models and custom deployment?
+a) Cogniflow
+b) PrivateAI
+c) ChatGPT API
+D) Hugging Face
+
+{/quiz}