Domain 1 - AI Problem Definition

complete
1. Mean squared error and mean absolute error are measures of prediction error.
2. When selecting an algorithm, people should know whether a(n) numeric prediction or a
categorization of items is required.
3. Inputs to an AI model are known as features.
4. Different algorithms solve different problems in an AI solution.
5. All efforts to reach compliance should be documented.
6. People should understand the ethical and legal concerns associated with sensitive data and
know how to handle that data appropriately.
7. Intelligence is the ability to learn, adapt to, and understand concepts and use this knowledge
to control the happenings in an environment.
8. Classifying data helps determine what data is best for solving an AI-related problem.
9. An AI expert needs enough business acumen to interpret the requirements for a project, the
desired outcome of said project, the data that will be used, and how a(n) model is best built
and optimised for a solution.
10. Scalability allows an AI system to adjust the resources needed to accommodate the
processing demands placed on a system.
11. A business expert should know what business constraints apply to a possible AI solution, as
these create limitations for an AI solution.
12. People should ensure that gathered data does not violate any privacy laws, security laws, or
industry regulations.
13. A lack of transparency could breed suspicion, lead to data misuse, or result in flawed
decision-making based on misunderstood results.
14. When building an AI model, Algorithm expertise is less critical than expertise regarding the
problems being solved.
15. A security team should keep up-to-date on threat intelligence.
16. When possible, a(n) AI ethics document should explain how data is collected.
17. An implementation expert needs to know how to engineer and process data in an AI model.
18. When necessary, one or more individuals should look over data to ensure it is not biased,
which skews the results of an AI solution.
19. Before using AI, one should determine the desired outcome.
20. Infrastructure security comes in two primary forms: network security and physical security.
21. Data can be stored securely on-premises, in the cloud, or both.
22. Compliance may require an organization to publish documents regarding how data is
handled.
23. Once AI risks are identified, one should assess the impact and likelihood of each risk.

MCQ
1. What is artificial intelligence (AI)?
Programmed behaviours of intelligence that can be interpreted as human intelligence.
2. What is linear regression?
A type of regression that uses continuous variables as outcomes.
3. Why should users take steps to reduce bias in AI models?
Bias can lead to inaccurate results that exclude certain groups.
4. What is feature scaling?
A process that determines what features will be used in an AI project.
5. What does the principle of least privilege state?
People should have access to what they need to perform their job functions and no more.
6. What is the purpose of area under the receiver operating characteristic curve (AUC-ROC)?
It means how well an AI model can distinguish between classes.
7. What do many AI solutions require to develop those solutions best?
Knowledge of a programming language.
8. For network security, what should happen to data stored on a server?
It should be encrypted.
9. As part of ethics and data gathering, what ideas should be assumed?
No data is part of the public domain unless proven otherwise.
10. What happens to an F1 score if the precision or recall metric is low?
The F1 score will be low.
11. Which two statements about AI inputs and outputs are true?
Outputs can be numeric value or a classification.
Features are expected values for an AI model to predict.
Outputs should be determined before inputs.
Features cannot include text
Features are often numerical
12. Match data type to its description:
Unlabelled data  has no target values
Labelled data  Paired with output targets
13. What are three downsides of AI:
AI can be expensive.
AI results can be wrong due to problems with algorithms or data.
AI cannot be scaled.
AI cannot manage analytics.
AI can lead to less privacy.
14. What are 3 actions one should take to ensure legal requirements are met when using AI?
Consult legal expert.
Review existing data governance policies.
Research applicable laws and regulations.
Store data in the cloud.
Store data in multiple locations.
15. What are three uses of AI?
Managing employees.
Managing costumer service inquiries.
Analyzing information.
Creating relationships with costumers.
Automating tasks.
16. Match the AI problem to its description:
Unsupervised problems  find hidden patterns and structures in data
Classification problems involve categorizing the results
Regression problems  evaluate the relationship between an independent variable and
dependent variable
reinforcement problems  involve AI performing actions to maximize a reward
17. What is Role-Based Access Control (RBAC)
A control that assigns people to roles that exist in groups with designated permissions.
18. What should business expertise start with?
Understanding what problem AI will be used to solve.
19. Besides permissions, what should an organization implement to protect its on-premises
data?
Access to the data should be audited.
20. How can a user help mitigate bias from affecting AI results?
Perform a bias assessment of data.
21. Match the metric to its description:
Accuracy  the percentage of predictions an AI model gets right in an outcome.
Precision  the percentage of true positive predictions among all positive predictions in an
outcome.
Recall  the percentage of the true positive predictions among all actual positives in an
outcome.
22. How can one help ensure people interpret AI results correctly?
Be prepared to explain the AI algorithm in user-friendly terms.
23. What are three upsides to AI?
AI can manage creative business needs
AI is free of cost
AI can help improve organizational decision making
AI can help reduce human error
AI is scalable
24. What does an implementation expert need to know to implement an AI solution? Choose 2
answers.
Business constraints.
Ai techniques
AI algorithms
AI costs
Business problems
25. When should a business consider buying or developing AI solutions? Choose 2 answers.
When the business believes the AI will provide high return on investment
When the business believes the AI will replace employees
When the business believes the AI will create new products for the business
When the business believes the AI will improve business processes
When the business believes the AI will analyze historical data often
26. How can one decide who should have access to AI results? Choose 3 answers.
Consider the purpose of the results
Evaluate the type of algorithm used
 Consider the cost of the solution
Identify stakeholders
consider the context of the solution
27. What should be included in a transparency document? Choose 2 answers.
Cost of the AI solutiom
How collected data can help solve an AI problem
A reason for collecting data
A copy of the AI algorithm
Instructions for how to access collected data
28. What is the purpose of a subject-matter expert when building an AI model? Choose two
answers.
To provide financial expertise
To provide expertise on the problems being solved
To provide technical expertise
To provide expertise on data security and privacy
To provide expertise on creating AI algorithms
29. What is necessary for effective and ethical AI implementation? Choose 3 answers
ongoing dialogue
transparency
user information
Algorithm changes
Clear communication
30. What do items should organisations create to minimise risks associated with AI?
A risk prediction AI model
A risk register
A risk mitigation strategy
An AI generated list of risks
a risk management fund
31. What does R-squared measure?
The accuracy of an AI model
the number of dependent variables
the correlation between dependent and independent variables
the number of independent variables
32. Match the risk identification method to its description:
risk management  Identifies potential issues with an AI model, such as the likelihood that
a model will fall to certain types of attacks.
threat modelling  involves assessing potential threats to end AI model and vulnerabilities
in a system.
33. When gathering data, how can people help ensure the data is accurate?
Change data when it seems inaccurate.
34. Arrange the steps an AI expert should take when creating an AI solution in the correct order:
Decide which algorithms are best for an AI solution  Position 3
transform data to build an AI model  position 2
understand the requirements of the AI project  position 1
35. How can one help ensure AI solutions are not used out of context?
 Clearly communicate the scope and limitations of AI solutions
36. What is logistic regression?
A type of regression that uses ordinal variables as outcomes
37. Match the algorithm selection guideline to its description:
Empowering users  when possible users should have some degree of control over
selecting an AI algorithm
transparency and explainability  AI algorithms should provide an optimal balance between
accuracy and interpretability
Ethics and fairness  AI algorithms should promote impartiality and avoid bias.
Accessibility  AI algorithms should provide solutions that people with varying degrees of
technical proficiency can use
38. How can users ensure network storing data is secure? Choose 3 answers.
The network should be behind a firewall
The network should be a public connection
The network should be segmented
The network should be acquired with an intrusion detection or prevention system
The network should only be accessible for a limited amount of time