AI Dialogues: exploring the significance of training AI models

It's AI bingo at the moment! Everyone's talking about it - but did you know that the quality and quantity of training datasets are crucial to the accuracy and effectiveness of machine learning models? The more diverse and representative the data, the better the AI model can perform in terms of nuance. Did you ever wonder how to ensure that AI discerns what Americans would wear to a wedding? And what people would wear in Asia, Europe or Africa, taking into account the specificities of each country? In this enlightening AI dialogue episode of our Globally Speaking podcast, Melanie Peterson - Program Director for Train AI at RWS - joins Vasagi Kothandapani to discuss how she tackled one of her biggest AI challenges: a rating and labelling data project for western-style womenswear using resources in the APAC region only, where the raters were 80% male. Melanie unpacks the intricacies of her mission, shedding light on how she explains what she does to her family and friends, why a rocket scientist or a chef might have been involved in one of Train AI's latest data training projects, and what exactly we mean by data collection, data creation, data annotation and sentiment analysis. Get ready for a captivating exploration into the heart of AI.