Stephen Hawkings, Elon Musk, Bill Gates, Rachel Thomas, and other brilliant minds show caution when speaking of the benefits of generalized AI where systems can handle any task that intelligent humans could perform, and most likely beat us at each of them.
Here are 5 listed cautionary scenarios:
- Photo by Matan Segev on Pexels.com
(1) Autonomous weapons with an individual or government that doesn’t value human life
(2) AI used to target and manipulate clusters of humans, based on personality makeup datasets, to spread propaganda for an economic or political agenda of the AI owners.
(3) Invasion of privacy and government-based social credit policies turn to social oppression whereby mass social behavior is monitored, socially graded, and consequently rewarded or punished.
(4) Without specific ethical rules written in code, misalignment between human goals and AI may result in task efficiency at the risk of ignoring basic human needs and rights.
(5) AI with extensive human data may discriminate individuals based upon their “social credit score”
MIT has just announced a new internal college that will provide an interdisciplinary environment for the study of applied AI ethics.
e.a.r.t.h. is curious as to how it might participate in furthering AI ethics research in order to mitigate an existential risk. For now, this short article provides alert to the potential risks, and one solution that we see helpful for humanity: public awareness of the AI risk. We may be interested in starting exploring options to help AI ethics. We will start by having all e.a.r.t.h. staff and participants to complete a certificate of AI ethics and governance curriculum. This way, all members will be able to communicate most effectively while developing a strategy.
Just discovered a link of AI ethics resources that could help us here, written by cofounder of Fast.AI, Rachel Thomas (@math_rachel).
My machine learning is turned on. One must.