Although not an expert in any way on AI, I did work as a machine learning researcher once at an AI lab and have a lot of thoughts pertaining to this area. I personally think people grossly overestimate the power of AI in general, but particularly with regard to ethics. Of course, a lot of the problems in AI bias is in human bias so there is an argument that it would actually be more ethical to leave humans out of the equation, do a very hands-off version of unsupervised learning and have robots train robots because machines can’t be biased or problematic, can they? I find it interesting that Ge Wang argues against this approach in “Humans in the Loop: The Design of Interactive AI Systems” argues that AI is already too hands off and the solution is to incorporate humans in every step. Some great points were raised, especially given the fact that even amongst AI researchers there were lots of components in the neural networks and inference layers where even people in the field weren’t sure what exactly was happening there. It almost felt like a leap of faith, such as in recursion, where you just have to trust it’s working and judge it on whether or not it worked based on the results. Something interesting about Ge’s humans in the loop approach is it would not only add humanity to AI and improve its effectiveness, but it would allow people as engineers to better understand every step of how machines learn. Instead of these leaps of faith we can look under the hood and see where and why things end up messing up, like when an algorithm mistakes a muffin for a corgi. There’s also something I found poetic about entrusting humans to be more involved in these technologies not as engineers, but as creatures that are naturally capable of empathy, intelligence, and learning. It almost begs the question, why in a world where we still don’t have the cure for cancer are we spending so much money and resources on trying to make machines as smart or smarter than humans when there are SEVEN BILLION people in the world? Slowing down the development of algorithms to incorporate more human involvement would also still ultimately move technological advancement forward (as I mentioned before sure we have machine learning algorithms that work but barely anyone truly understands how).