Azure AI
AI, or artificial intelligence, is technology that enables computers and machines to imitate human behaviours and capabilities. Microsoft provides a number of services in Azure that utilise AI.
Microsoft's Six Guiding Principles for Using AI
Development using AI at Microsoft is governed by six guiding principles, fairness, reliability and safety, privacy and security, inclusiveness, transparency, and accountability. These principles are outlined below.
Where systems utilise AI, they should treat everyone fairly, without bias that creates an unfair advantage or disadvantage to specific groups of individuals. This bias could include but is not limited to gender, ethnicity, ability, or disability.
Software that incorporates AI should undergo rigorous testing to ensure that it functions as desired in different use cases. Any limitations of the system should be outlined to users, so that they are aware when using the system.
AI based systems use large volumes of data for the training of the system. This data may contain personal details, so it is important to keep it secure to maintain privacy, as well as the integrity of the data. Security of the system must also be maintained once the finished product is available to end users.
AI systems should be inclusive of everyone, regardless of physical ability, gender, sexual orientation, ethnicity, or other factors.
Developers should make end users aware of exactly how and why AI is used in a system, as well as making sure that it is understandable in the way that it works. Also, if there are any limitations of the system, end users should be made aware of these.
Designers and developers of AI-based systems should be accountable for those systems and work within a framework of principles and governance to ensure that the end result meets defined ethical and legal standards.
Azure AI Services
Machine learning is the way in which we train a computer model to make predictions and draw conclusions from data. It is the foundation for most AI systems. The Azure Machine Learning service provides a means to create, manage, and publish machine learning models.
Useful Links
Computer Vision is a part of AI that can handle the processing of visual elements, by applying a machine learning model to items such as images and videos.
Azure AI Vision is a service that includes pre-built, as well as customisable, computer vision models, which can be used to analyse images in a number of different ways. It can extract text from images using Optical Character Recognition (OCR), generate image captions and descriptions, detect common objects in images, and tag visual features contained in images.
For more specific analysis of the human face, the Azure AI Face service includes pre-built algorithms to detect, recognise and analyse them.
Useful Links
Natural Language Processing, or NLP for short, is a part of AI that can be used to understand both written and spoken language, as well as provide responses to it.
The Azure AI Language service can be used to both understand and analyse text. It can analyse the sentiment, identify key phrases, summarise text and understand conversational language.
The Azure AI Speech service provides a means to convert text to speech and speech to text using speech recognition and synthesis. It offers pre-built and customisable machine learning models for various tasks including audio transcription, speaker identification and the creation of custom voices.
Useful Links
Document intelligence is an area of AI that provides support for the processing of text in order to make sense of the information contained within.
Azure AI Document Intelligence is a service that includes the capability to analyse documents and forms using both pre-built and customisable machine learning models.
Useful Links
Knowledge mining is the act of extracting information from large volumes of data, which is often unstructured.
Azure AI Search can be used to build user-managed indexes of data, which are searchable. To do this it utilises other Azure AI services for image processing, content extraction and natural language processing.
Useful Links
Generative AI is an area of AI that can be used to generate original content. Applications that utilise Generative AI accept natural language input and then provide a response in any one of a number of different formats, including natural language, images, and code.
The Azure OpenAI service can be used to deploy, customise, and host large language models.