AI-900 Study Cards
What are the NIST AI Risk Management framework steps?
1. Identify potential harms 2. Measure the presence of the harms 3. Mitigate the harms at multiple layers 4. Operate the solution responsibly
What are the four layers for harm mitigation techniques?
1. Model 2. Safety System 3. Metaprompt and grounding 4. User experience
What type of workload do you use for identifying a fraudulent credit card payment?
Anomaly detection
What are the four key elements of training a regression model?
1. Split the training data to create a dataset to train the model and a validation set 2. use an algorithm to fit the training data to a model - often a linear regression 3. use the validation data you held back to test the model 4. compare the known actual labels in the validation dataset to the labels that the model predicted to see accuracy
What is opinion mining?
Aspect-based sentiment analysis More granular information about the opinions related to aspects - looks at subject and opinion tied to a sentiment Part of Text Analytics
What is semantic segmentation?
A form of ML classification that seeks to classify each pixel in an image by by what class / object it belongs to.
What is classification?
A form of supervised machine learning in which the a categorization or class can be binary (is or isnt) or multiclass (one of these things)
What kind of information can you put in a knowledge base?
A web page, existing document FAQ, or manually entered Q&A pairs Cannot directly use image or audio file to import a knowledge base
Which service can you use to train an image classification model?
Azure AI Custom Vision
What is Azure AI Language?
Azure AI Language is a cloud-based service that provides Natural Language Processing (NLP) features for understanding and analyzing text.
What can be used as deployment targets once a model's training is complete?
Azure container instance or AKS cluster
Which AI service can be integrated into chat applications and generate content in the form of text?
Azure OpenAI only service capable of generating text that can be used in chat applications to create conversational experiences
Which of the following meets the transparency principle for responsible AI? A. Ensure that all visuals have an associated text that can be read by a screen reader. B. Enable autoscaling to ensure that a service scales based on demand. C. Provide documentation to help developers debug code. D. Ensure that a training dataset is representative of the population.
C. Provide documentation to help developers debug code.
What is language detection?
Capability of Azure AI Language to identify the language in which text is written. You can submit multiple documents at a time for analysis will extract - Language name ISO 639 code a confidence score
What type of workload do you use for identifying handwritten letters?
Computer vision
What type of workload is used for determining whether a photo contains a person or not?
Computer vision
What type of workload type is used for an automated chat to answer questions?
Conversational AI
_____ use plugins to provide end users with the ability to get help with common tasks from a generative AI model
Copilots
Which three capabilities are examples of image generation features for a generative AI model?
Create variations of an image Edit an image Create new image Cannot animate static images
Which genAI model is used to generate images based on natural language prompts?
DALL-E
You are exploring solutions to improve the document search and indexing service for employees. You need an artificial intelligence (AI) search solution that will include searching text in various types of documents, such as images. Which type of AI workload is this?
Data mining These workloads focus on searching and indexing data
What does Language Detection do?
Detects which language and input text is written in Part of Text Analytics
Which feature of the Azure AI Language service includes functionality that returns links to external websites to disambiguate terms identified in a text?
Entity recognition includes the entity linking functionality that returns links to external websites to disambiguate terms (entities) identified in a text.
_____ is used to generate additional features?
Feature engineering
In a regression machine learning algorithm, how are features and labels handled in a validation dataset?
Features are used to generate predictions for the label, which is compared to the actual label values There is no direct comparison of features or labels between the validation and training datasets.
What does Sentiment analysis do?
Find out about what people think about a brand or topic Assigns sentiment labels like positive, negative, netural. Also provides a confidence score Part of Text Analytics
As per the NIST AI Risk Management Framework, what is the first stage to consider when developing a responsible generative AI solution?
Identify the potential harms
You are designing an AI system that empowers everyone, including people who have hearing, visual, and other impairments. This is an example of which Microsoft guiding principle for responsible AI?
Inclusiveness
Which two specialized domain models are supported by Azure AI Vision when categorizing an image?
Landmarks and celebrities
Which three values are returned by the language detection feature of the Azure AI Language service in Azure? Select all answers that apply.
Language Name Score ISO 6391 Code
What is the tree of life for Machine Learning?
Machine learning - supervised machine learning ---regression ---classification -------- binary classification -------- multiclass classification - unsupervised machine learning - clustering
What type of workload do you use for predicting next month's toy sales?
Machine learning (regression)
What are the features in Azure AI Language?
NER Entity linking PII detection Language detection Sentiment analysis and opinion mining Summarization Key phrase extraction
What type of workload do you use for predicting the sentiment of a social media post?
NLP
Which workload would be used for extracting key phrases from text to identify key phrases?
NLP
What is Text Analytics and what 5 things can it perform?
NLP service for text mining and text analysis Sentiment analysis Ppinion mining Key phrase extraction Language Detection Named Entity Recognition (NER)
What is the confidence score returned by the Azure AI Language detection service of natural language processing (NLP) for an unknown language name?
NaN Score values range between 0 and 1, but NaN is "not a number"
What is NER?
Named Entity Recognition Detects words or phrases and applies a semantic type (people, place, event, diagnoses) PII is a subset of NER Part of Text Analytics
Which Azure AI Service for Language feature allows you to analyze written articles to extract information and concepts, such as people and locations, for classification purposes?
Named Entity Recognition Named entity recognition can identify and categorize entities in unstructured text, such as people, places, organizations, and quantities
______ used to extract dates, quantities, and locations from text
Named Entity Recognition (NER)
What type of workload is used for determining whether a review is positive or negative?
Natural Language Processing
Which type of artificial intelligence (AI) workload provides the ability to generate bounding boxes that identify the locations of different types of vehicles in an image?
Object detection
Which artificial intelligence (AI) technique should be used to extract the name of a store from a photograph displaying the store front?
Optical character recognition (OCR) OCR provides the ability to detect and read text in images.
What is entity recognition and linking?
Provide Azure AI Language with unstructured text and it will return a list of entities in the text that it recognizes. An entity is an item of a particular type or a category. For recognized entities, the service returns a URL for a relevant Wikipedia article.
What is key phrase extraction?
Quickly identifies the main concepts in a text Part of Text Analytics
The handling of usual or missing values is a consideration for which microsoft pillar?
Reliability and safety Systems needs to respond safely to edge cases
At which layer can you apply content filters to suppress prompts and responses for a responsible generative AI solution?
Safety System The safety system layer includes platform-level configurations and capabilities that help mitigate harm, like content filters
Which type of artificial intelligence (AI) workload provides the ability to classify individual pixels in an image depending on the object that they represent?
Semantic segmentation
What does Whisper do?
Series of models that can transcribe and translate speech to text
What does Embeddings do?
Set of models that can convert text into numerical vector form to facilitate text similarity
What can used to identify constraints and styles for the responses of a generative AI model?
System messages They can be used to set the context for the model by describing expectations so the model knows how to respond to prompts
What is the purpose of knowledge mining?
To make large amounts of data searchable
Which principle of responsible artificial intelligence (AI) raises awareness about the limitations of AI-based solutions?
Transparency Provides clarity regarding the purpose of AI solutions, the way they work, as well as their limitations
What do GPT4 and GPT3.5 do?
Understand and generate natural language and code
What are system messages?
Used in prompt engineering, they set the context to the model by explaining expectations and constraints to guide the model's responses
What is grounding data?
Used in prompt engineering, where you include contextual data in the prompt so the model can use it to generate an appropriate example, like attaching an email that you want it to summarize
You need to use Azure Machine Learning to train a regression model. What should you create in Machine Learning studio?
a job
What do you need to create before you can access Machine Learning studio?
a workspace
Which responsible AI principle? Implementing processes to ensure that decisions made by AI systems can be overridden by humans
accountability The people who design and deploy AI systems must be accountable for how their systems operate. AI systems are not the final authority on any decision that impacts people's lives and that humans maintain meaningful control over otherwise highly autonomous AI systems.
Which principle of responsible artificial intelligence (AI) ensures that an AI system meets any legal and ethical standards it must abide by?
accountability ensures that AI systems are designed to meet any ethical and legal standards that are applicable
What is clustering?
an example of unsupervised learning, which groups data points that have similar characteristics without relying on training and validating label predictions. analyzes unlabeled data to find similarities in the data
What does the fairness principle entail?
applied to AI system to ensure that users of the systems are treated fairly. Azure ML can tell you how each feature can influence a model's prediction for bias
Which three supervised machine learning models can you train by using automated machine learning (automated ML) in the Azure Machine Learning studio?
classification, regression, and time-series forecasting are all supervised machine learning models
Which type of machine learning algorithm finds the optimal way to split a dataset into groups without relying on training and validating label predictions?
clustering A clustering algorithm is the most common type of unsupervised learning groups data points that have similar characteristics without relying on training and validating label predictions. groups them into discrete clusters
Which workload would be used for identifying objects in landscape images?
computer vision
Which artificial intelligence (AI) technique serves as the foundation for modern image classification solutions?
deep learning Modern image classification solutions are based on deep learning techniques.
What is the purpose of a validation dataset used for as part of the development of a machine learning model?
evaluating the trained model
When using the Face Detect API of the Azure AI Face service, which feature helps identify whether a human face has glasses or headwear?
face attributes
What is Azure AI Custom Vision?
image recognition service that lets you build, deploy, and improve your own image identifier models image identifier applies labels to images, according to their visual characteristics. Each label represents a classification or object. Custom Vision allows you to specify your own labels and train custom models to detect them.
Which principle of responsible artificial intelligence (AI) has the objective of ensuring that AI solutions benefit all parts of society regardless of gender or ethnicity?
inclusiveness
Which type of artificial intelligence (AI) workload has the primary purpose of making large amounts of data searchable?
knowledge mining purpose is to make large amounts of data searchable
In a regression machine learning algorithm, what are the characteristics of features and labels in a training dataset?
known feature and label values
You need to identify numerical values that represent the probability of humans developing diabetes based on age and body fat percentage. Which type of machine learning model should you use?
logistic regression Multiple linear regression models a relationship between two or more features and a single label.
What are the types of AI workloads?
machine learning (ML) deep learning (DL), knowledge mining natural language processing (NLP) computer vision
Which workload would be used to predict whether consumers are likely to buy a product based on previous services?
machine learning (regression)
What are classification models?
make decisions or assign items into categories classification models output Boolean values like true or false, or categorical decisions like apple, banana, cherry
Which feature of the Azure AI Translator service is available only to Custom Translator?
model training with a dictionary Model training with a dictionary can be used with Custom Translator when you do not have enough parallel sentences to meet the 10,000 minimum requirements
Bounding a box that indicated the location of a vehicle in an image is an example of what?
object detection
Which analytical task of the Azure AI Vision service returns bounding box coordinates?
object detection
Which responsible AI principle? Provide consumers with information and control over the collection, use, and storage of their data
privacy and security AI systems must comply with privacy laws that require transparency about the collection, use, and storage of data and mandate that consumers have appropriate controls to choose how their data is used
What is supervised vs unsupervised learning?
supervised uses data with labels unsupervised uses data that does not have labels
What does Azure AI Translator service do?
supports text-to-text translation in more than 60 languages
What is Azure AI Face?
provides AI algorithms that detect, recognize, and analyze human faces in images
Predicting rainfall for a specific geographical location is an example of which type of machine learning?
regression bc predicting a numeric value
What are features in regression?
relationship between variables in the data that represent characteristics (like the x axis, where labels are the y axis we are trying to predict the value from)
A company is currently developing driverless agriculture vehicles to help harvest crops. The vehicles will be deployed alongside people working in the crop fields, and as such, the company will need to carry out robust testing. Which principle of responsible artificial intelligence (AI) is most important in this case?
reliability and safety
Which responsible AI principle? When developing AI for self-driving cars, the microsoft ____ principle should be applied to ensure consistent operation of the system during unexpected circumstances
reliability and safety AI systems need to operate reliably, safely, and consistently under normal circumstances and in unexpected conditions
Which responsible AI principle? Ensure that AI systems operate as they were originally designed, respond to unanticipated conditions, and resist harmful manipulation
reliability and safety AI systems operate reliably, safely, and consistently under normal circumstances and in unexpected conditions. These systems should be able to operate as they were originally designed, respond safely to unanticipated conditions, and resist harmful manipulation.
Which Azure AI Service for Language feature can be used to analyze online user reviews to identify whether users view a product positively or negatively?
sentiment analysis Provides sentiment labels, such as negative, neutral, and positive, based on a confidence score from text analysis
What capabilities does Azure Speech service have?
speech to text (+realtime) text to speech batch transcription for large audio files translation (only for speech to speech or speech to text, NOT text to speech) language identification speaker recognition pronunciation assessment intent recognition
What should you do after preparing a dataset and before training the machine learning model?
split data into training and validation datasets
Which type of translation does the Azure AI Translator service support?
text-to-text
Which two Azure AI Services features can be used to enable both text-to-text and speech-to-text between multiple languages? Each correct answer presents part of the solution.
the Speech service the Translator service
A company is using machine learning to predict house prices based on appropriate house attributes. For the machine learning model, which attribute is the label?
the price of the house label is what you are trying to predict