Synthetic data is cheaper, faster and possible
FIIF Event (AI in Industry series): “AI and the power of simulation” discussed the use of synthetic training data. In-house data, if available, needs to be collected, compiled, and annotated and still it may be biased and inadequate to model untypical use cases. Synthetic training data (i.e. computer-generated and realistic use case data) in most cases reduces the time needed for creating data.
Ouissem Aloui from AILiveSim pointed out in his presentation that there are industries, which already use synthetic data, because it is cheaper, for example mining and marine industry. Synthetic data is used for car crash tests, and for example for missile industry, synthetic data is the only option, because the use of real data is not possible.
The presentations of the event: ”AI and the power of simulation”, Antti Löytynoja, Mathworks;“Overview of Santtu project – simulation and AI towards autonomous heavy machinery”, Grzegorz Orzechovski, LUT; “Control oriented modeling with sub-system based learning for mobile working machines”, Jouni Mattila, Tampereen University; “How AI-centric Simulation Improves AI Model Performance: The AILiveSim Approach”, Ouissem Aloui, AILiveSim and “Simulation based Development” and Antti Mikkonen, Gofore.
This FIIF event was organized in co-operation with MathWorks and FIMA.
