Ramon Amaro: Machine Learning, Surveillance and the Politics of Visibility
- Amaro talks about skin colour and facial recognition software. He uses the example of Joy Buolamwini and her work ‘The Aspire Mirror’ where she used multiple software libraries for the facial recognition. Buolamwini encountered a probelm during testing, the software couldn’t detect her face due to her skin tone. This led me to think about the issues with recognition software and bias. My ‘Is it cake or is it real?’ project uses Google’s teachable machine software to recognise the objects presented to the camera. However, I am in control of the data that the machine uses. So inevitably there is a bias as I pick what images the machine trains with.
- If I had more time with this project I could have made it possible for other people to upload their own images to reduce bias as the machine would be learning from multiple sources rather than one.