The Artificial Intelligence Action Summit kicks off Monday in Paris. Machine learning applications are already ubiquitous. How does it work? This videographic answers. VIDEOGRAPHIC
Category
🗞
NewsTranscript
00:00Machine learning applications are ubiquitous.
00:11We are using them on a daily basis, often without realising it.
00:14On your smartphone alone, there are already countless applications.
00:18When you log on, browse for products while shopping online, check emails, type a text
00:23message or plan a route.
00:25So what is machine learning and how does it work under the hood?
00:29It is a subfield of artificial intelligence that gives computers the ability to learn
00:33without explicitly being programmed, by enabling them to learn from data and experience.
00:39Sometimes an algorithm can learn a task under supervision.
00:42This involves using labelled data to train a model to make predictions or decisions.
00:48It is used on online shopping to predict a user's preferences based on historical data
00:53such as purchase history, browsing behaviour and product attributes.
00:58It can train a model to recommend products accordingly.
01:01Other times algorithms can learn without being supervised.
01:05This involves identifying patterns and relationships in unstructured data to create structure and
01:10categorisation.
01:11It is used to cluster different products based on attributes such as price, category, brand
01:17and descriptions, without labelled examples or guidance.
01:21The goal is to find hidden structures and patterns and identify similarities or differences
01:26between products.
01:27Finally, sometimes algorithms learn from their own performance and previous experiences as
01:32they adjust their behaviour to improve the decision making process.
01:36This can be used to train a model to recommend products that maximise user engagement or
01:40satisfaction by recommending the right products to the right users.
01:44Machine learning has a wide array of applications but it comes with ethical implications we
01:49need to consider surrounding issues related to bias, fairness, accountability, privacy
01:54and transparency.
01:57www.microsoft.com
01:58www.microsoft.com
01:59www.microsoft.com
02:00www.microsoft.com