A Simple and Repeatable Example

We wish to introduce the topic of why AI needs ethical thinking. However, before we dive in, a disclaimer is required. This video and blog do not seek to offend, nor are they religious in nature. We simply want to perform a search engine query and analyse the result.

Our query is, “Jesus”. If we now look at the results page, we are given the impression that Jesus was white.



Do you see the problem? I am shown images that might imply a hidden bias.

Please watch the video for the complete discussion and why academics think Jesus was of Iraqi descent and most likely not white.

What can we take away from the video?

If you fill up your databases with data and you are not very careful, you could bake in biases. Consequently, there could be unintended impacts on how you treat wider society. And here’s the rub, you could be doing this now and into the future. Like a lousy law could have negative long term consequences, our databases could disrupt over time.

It takes a considerable effort to gather data, build databases, create indexes, and train AI systems. Therefore, an ethical dimension is needed before going too far down the road to creating AI solutions. We should be asking if we are likely to cause harm. Similarly, in academia, we use a peer review process to keep research claims on the rails. We can take insights from our engineering experiences, especially in aviation, where ethics, discipline, and procedures are precisely why aeroplanes do not regularly fall out of the sky.

So, where are we with Data and AI Ethics? Is there not a need to ask ethical questions and hold people and organisations to account? Do we not need to move faster to have AI-based systems operating to ethical standards? Please remember we saw a simple and easy to reproduce example using search engines in the video. I am more worried about the impact of more complex and less easily identified systems, such as those used for medical diagnosis or criminal justice, where there could be bias against specific segments of society.