Dr. Haniyeh Mahmoudian (DataRobot): Who should be involved in AI ethics?
In AI We Trust?
00:00
The Process of Mitigating Bias in AI
The data can be biased whether it's historical biases that us as human have and now it's embedded in the data. In order for us to be able to intervene there are three stages: pre-processing, processing and post-processing. The focus is actually during the trainings part so when we are training the way that it's set up is for model to be rewarded when it's getting the right answer but what we try to do in the in-processing step is actually enforce the aspect that the model should find patterns that are fed.
Play episode from 21:18
Transcript
