Synthetic data - What is it we are adding to our training data, actually? By Heléne Ström and Ericka Johnson of Fair AI Data

​Following up our previous Lunch and Learn and the discussion about synthetic data we welcome Heléne Ström and Ericka Johnson from Fair AI Data to Nordic Innovation house.

​Synthetic data can augment and amplify training data for AI algorithms - but this is not as straightforward as it may appear. Fair AI Data, a spin-off from Swedish research that explores missing edge cases (mode collapse) and the production of intersectional hallucinations in synthetic tabular data, engages a concern for the complexities of representation in synthetic data to produce tools that clarify important relationships in data, ensure their continued representation in synthetic data sets, and document synthetic datasets according to open science standards, allowing for transparent synthetic datasets that are trustworthy and accurate.

​Bring your lunch (or eat beforehand), join the session, and let us start the weekend with curious discussions about how to tackle transparent synthetic datasets that are trustworthy and accurate.