RESPONSIBLE AI

摘要

Since the Cambridge Analytica case broke, the discussion on how to use Artificial Intelligence (AI) ethically has intensified. Whilst beforehand, discussion was centred on fears of automation and possible loss of jobs, the discussion now is about the uncontrolled use of personal data and the consequences of bias.We know our online shopping and browsing histories are used to recommend to us new products. The fact, however, that data about Facebook Likes paired with a personality test can lead to predictive models that can be used to micro-target voters has been a shock to most. But there are ethical concerns in other areas too.Machine learning uses historical data. The resulting models that are used to make predictions on new data can only replicate the statistical distributions contained in that historical data. Bias in the original data will make its way into the model. Research has shown that today's facial recognition systems can be racially biased suggesting that these systems have been trained with data that is race and gender biased.