With the fast progression in AI, its role in healthcare is bound to be ubiquitous in near future. However, creating AI software that is unbiased and trained ethically remains a challenge. Any AI is as good as the data it is trained on. In the recent years, we have seen several examples of how blindly training an AI on data available on the internet can create extremely biased and almost dangerous AI. Recent examples being,  a chatbot trained by a large software company on tweets showing racist behaviour to melanoma detection softwares not working correctly on dark-skinned people. 

In this session, we will go through different risks involved with AI and how an organisation can introduce concrete measures from the get-go to reduce the bias in data. We will particularly look into AI-driven radiology systems and translation systems in healthcare. 

Similarly, it is important to understand the fact that ethical AI is a two-way street. We will also look into how the society and the regulatory bodies should play a role for AI to be constructed in an ethical way.