Ethics in AI and DL — Who is Accountable?
Ethics cannot be discussed or practiced as a subject on its own nor can it be discussed and applied universally. Ethics can only be discussed and applied in relation to your philosophy or world view, this makes is an extraordinarily complex subject to address in a universal application because there are such divergent philosophies, so the challenge in AI and DL is how do you address ethics in a society that is not homogeneous and does not share the same basic philosophy. Even age plays a roll a teenage coder does not have a sophisticated understanding of ethics and morality, relying on them to police themselves is not going to be effective at all.
There is an additional aspect that needs to be considered, that is freedom of speech and what that means in terms of what can be said.
The biggest challenges however are not the questions of ethics but are in fact the question of accountability. If there is a standard of ethics that should be upheld, who is going to ensure that they are upheld?
We can discuss, teach and agree on a complex ethical standard that covers all of the areas of concern like the misuse of technology for nefarious reasons such as spreading fake information with the express intention of creating an alternative reality so that you can take advantage of the new reality. This kind of abuse of technology is real and present in the world today and clearly there is not mechanism in place to control what is happening.
The debate over net neutrality coupled with freedom of speech and a lack of ability to enforce ethical principles makes AI an open target for those who want to use it for less than ethically acceptable uses. The amount of false information that has been generated and spread concerning elections in different countries as well as information regarding CoViD-19 with no consequences to the originators of the false information makes AI a haven for misuse.
The most concerning of the technologies is GANs where deep fakes can be created that are undetectable. Evidence can be created to either convict or acquit someone of a crime, a political or influential leader in business could be “caught on tape” saying or doing something that boosts or destroys their career.
Teaching ethics is not going to minimise the risk of technology being abused, rather the debate should be around some form of accountability for the abuse of the technology, this is once again raising a series of complex questions that will require international cooperation and some very creative ways to measure the use and abuse of technology.
The bottom line is you cannot legislate morality, but you do need to have clearly defined consequences for your actions, this could be a peer group enforcement or a legislated set of standards that can be enforced globally.
There are advantages to both methods and the final best practice might be a combination of peer and legislated principles that can be flexible enough to deal with the newest developments in technology.
