May 4, 2021
You walk into a bank, and you have all the means to afford a mortgage to buy your dream house. After looking at your application the bank tells you they will not give you a loan. At this stage, what matters most to you?
1) To know if the decision was made by a steam engine calculator, by a human brain, or by an AI mathematical model, and to discuss exactly how each might have reached its "decision"?
Or
2) To follow a robust, simple, available process of appeal that corrects the decision? After all you are capable of paying.
Imagine now that you read an article online spreading misinformation about CoViD. What matters most to you:
1) To discover if the website owner employed a human or an advanced NLP model to write the article and to regulate the NLP model?
OR
2) To have the means to control and regulate the spread of misinformation and hold people accountable for publishing it?
We should spend more of our "ethics" and "regulation" resources on the consequences of actions and opportunities for redress. For example, many decisions are made with SQL-database statements, but do we need SQL regulation? Steam Engine ethics? The analogy isn't perfect but attention should re-calibrate to focus on the very real consequences and outcomes to those affected.
Attention paid to the victims of any injustice is well deserved. And resources (people, money, time, etc) dedicated to the fascinating details of "how does AI/human-brain/steam-engine work?" may come at the cost of resources dedicated to the less glamorous but more practical, and largely technology agnostic, problems of "what systems do we put in place to prevent and correct injustices?".
Precisely because getting the regulation and ethics right matters: where do you think we should focus attention? When faced with any specific AI-ethics or AI-regulation situation, here is simple check: ask yourself "do I care mostly about the technology that is being used, or about what is being done?"
I believe that focusing on what is being done and systems and processes to correct injustices in largely technology agnostic ways, will in the end result in better more trustworthy human-AI interactions and systems. I also think it will have larger positive impact in people's lives.