Module 3b - Describe Core Solution and Mgmt Tools - AI Flashcards
What are the two basic approaches to AI?
Deep Learning - Neural Networks basically. Systems that model the human mind, enabling discovery, learning, and growth through experience
Machine Learning - uses existing data to train a model, to test the model, then apply the model to new data to forecast future behaviors/outcomes/trends
What is Azure Machine Learning?
A platform for making PREDICTIONS. Build your model, train and test them and find the one that gives the most accurate result, then give a RESTful endpoint for production use
You can use tasks in the Machine Learning UI to define tasks for
- Handling missing data
- Obtaining data
- Splitting data into TRAINING vs TESTING data sets
- Pipelines that define where/when to run experiments
- Deployment of your selected algorithm as an API
When would you use a customized Azure Machine Learning implementation over an Azure Cognitive Services one?
When your data is private/proprietary and when you need to update your model over time.
Azure Machine Learning can be used to customize Prediction and Recommendation models when the data is private or proprietary. Usually we’d use Cognitive Personalizer for Prediction/Recommendation but not when data is proprietary.
Also when the model must be updated over time; we can’t do that if using a Cognitive Services out-of-the-box model. So a custom model through ML is more appropriate
What is Azure Cognitive Services?
Prebuilt machine learning models that enable apps to see/hear/speak/understand, and possibly reason.
Azure Cognitive Services doesn’t require a Data Science background in order to use and can be accessed via API (T/F)?
True
Azure Cognitive Services requires you to also train models with your data, just as Azure Machine Learning does (T/F)?
False. Cognitive services already has pre-built models you can use right out of the box, which you can use your live data on to get predictions right away.
What are the FOUR (4) Categories of Azure Cognitive Services?
Language Services - Processing Natural Language, like evaluating sentiment, recognizing what users want or ask for, etc.
Speech Services - Transcribing basically (convert Speech to Text, Text to Speech), as well as Translation Services (verification and recognition)
Vision - Recognition and identification capabilities with images, videos and other VISUAL content
Decision - Recommendations, contend moderation, and Time Series Data anomaly detection
What is Azure Bot Service?
The service and the Bot Framework let you create virtual agents that understand/reply to questions like a human.
Azure Bot Service can be applied to a wide variety of Use Cases in the same manner as Cognitive Services (T/F)
False. This is a SPECIFIC USE CASE: creating virtual agents for intelligent communication. Bot Services use other Cognitive Services for understanding
What kind of tasks do we assign to Bot Services?
Simple, repetitive tasks that don’t necessarily require human intervention:
- Reservation systems
- Gathering profile/purchase/form fill-out information
- FAQs
What’s an example of Azure Bot Service using Azure Cognitive Services?
Multilingual Virtual Agents
When a Virtual Agent is interacting with a human who only speaks Mandarin; you’ll need to translate text to English then back again to Mandarin when you respond. This task is best suited for Azure Cognitive Translation Services
When should you choose Azure Machine Learning as your implementation for AI?
When you need complete control over the design and training of an algorithm using YOUR OWN data.
When you’re building AI to INTERACT with humans by using natural language, what Azure AI Service should you use?
Depends!
QnA Maker (found in Azure Marketplace) can be used to build, train and publish a sophisticated Bot service that uses FAQ pages, support sites, product manuals, Sharepoint docs or editorial content via RESTful API
Power Virtual Agents, which uses Microsoft’s Power Platform, uses prebuilt connectors for data and allows for customizable workflows with Power Automate.
Azure Bot Framework itself is best when the above two out-of-the-box solutions are too limited
When you’re building AI to analyze and understand images, videos or audio, or text translation, what Azure AI Service should you use?
Azure Cognitive Services.
Note these services are GENERAL PURPOSE. So many customers can use these right out of the box. The models have already been trained and tested, and the services can be inexpensively scaled
When you’re building AI to PREDICT BEHAVIOR or provide personalized RECOMMENDATIONS, what Azure AI Service should you use?
Azure Cognitive Services - PERSONALIZER Service.
Personalizer predicts behavior and provides relevant experiences based on usage patterns.
Note that Azure Machine Learning can be used to customize this behavior with your own data, but it also requires much more effort and expense