Ethical Considerations And Bias in AI

Questions by Eirini Komninou in September 2018 (gpg: BCC7807286C460D2A2DB1EF376804AD4F89568FF)

For a start, how do you perceive the terms “ethical” and “unbiased” use of AI?
Do you consider it our personal responsibility to deepen our Machine Learning understanding, for ensuring algorithmic transparency?
How do we write AI that is useful, while avoiding bias?
How can we trust AI that comes from inherently biased data, and can be easily tricked (see adversarial attacks)?
Can we get to net neutrality, given machine learning is targeting us via different kinds of traffic?
Do biased models lead people to further dodgy content?
How does AI/ML intersect with data privacy?
What are the implications if we don’t have a diversity in AI skills, such as expertise in history, sociology, and psychology
Do we, software people, have any control over applying ethical and unbiased AI?
And finally, do we aim for fixing legacy code that’s already available or do we focus our energy towards studying how to design systems that are more auditable, more accountable, more transparent?