A Brief Discussion on The Importance of Ethics with Artificial Intelligence

Artificial Intelligence (AI) is a rapidly evolving area in Information Technology. The promise of creating machines, systems, and applications capable of higher order decision making or “thinking” is an exciting for techies and non techies alike. 

My intent in starting this discussion is not to assault or impugn the concept of AI. I recognize that progress in this evolution must continue. What I do question is how we go about this, specifically with regard to ethics. 

Why such interest in ethics and AI? In short - if we create machines capable of decision making without the ability to tell right from wrong, we invite consequences that could be catastrophic. Worse than this, once the AI genie is out of the bottle, fixing the ethical bugs will be challenging if not impossible, depending on what the AI is being used for. 

Hollywood has long grappled with this concept. Go back even further, Mary Shelley’s Frankenstein touched on this concept. Man-made creations, from Skynet in the Terminator series, to Ultron in the Avengers franchise, turning against humanity has dire consequences. These consequences are even worse when the AI systems and machines are armed and can outsmart humans.

In this discussion, we will look at the complexities of this subject by first defining the concepts of ethics and ethos, then looking at potential applications of AI (and things that could go wrong). We will also make some recommendations on how to frame these important issues as the technology continues to develop.

What do Ethics and Ethos Mean?

In the most basic of terms, ethics is the discipline of what is good and bad, or right and wrong. Ethos speaks to character attributes which signify the ethics of a person, group, or institution. These qualities form the basis of human reasoning and intellect; they greatly influence our decision making abilities. 

While they may differ based on cultural norms, all societies have ethics. In practice they can be influenced by religious views, but in theory they should not be.

Given these definitions and parameters, one would expect an intelligent being - artificial or otherwise - would have to possess a basic ethical framework to make rational and ethical decisions. Decisions without basic ethics - without basic reasoning - will likely be flawed. The three example applications of AI that we will look at here will help in illustrating this.

Example 1 - AI in Medicine

A potential application of AI is in the medical field, to assist doctors in analyzing patient records and data to make treatment recommendations.  AI enables higher end 

analytics with greater efficiency. Theoretically, AI could make treatment recommendations to doctors faster, and doctors could use this information to treat more patients faster. The net result of this would be better outcomes from diagnosis and procedures - thereby saving lives.  In medical research, the potential would exist to solve medical mysteries faster. 

Example 2 - AI in Manufacturing

A second example of an AI application is in a manufacturing plant. An AI application could help analyze production methods to avoid waste in the production of a given product. Waste avoidance reduces manufacturing costs (increasing profits and profit margins). Taking it one step further - an AI application could make its recommendations directly to an automated assembly line without human intervention, saving time and materials.

Example 3 - AI in a Military Targeting System

An AI system could be used by the military to assist with targeting hostile actors, both in identifying and tracking targets, all the way through attacking and eliminating them. It can also be used to help deconflict targets and avoid friendly fire.

Ethical Implications in these Scenarios:

In Medicine: The fundamental ethical rule in medicine is “do no harm”. Without this, incorrect medical decisions could be made (adversely affecting patient care). Clearly in medicine, an ethical framework is required.

In Manufacturing: if the only motive in the manufacturing example is cost savings, the AI system could bypass other product requirements, such as those needed for safety.  An ethical business would only develop a safe product for consumers. Absent an ethical compass, an AI application might not.

In Military Applications: By far the most alarming example, given the nature of what the military does. Military ethics are governed by law to avoid civilian casualties - something that an unethical AI targeting system might disregard. If not properly calibrated to highly value human life, an AI system could potentially target innocent people, who could be killed if the system is not questioned.

Recommendations and Observations

We need to solve the ethical side of these AI scenarios prior to putting the systems in place.

The ethical dilemmas involved in AI are just as (if not more) complex than the technical challenges.

This is a broader discussion than just a technical one. It needs to be examined appropriately. This cannot be a discussion with a bunch of engineers and programmers locked up in a room. We need to engage people who really understand ethics, business people, legal professionals, and others who can help shape what ethical requirements are needed for a thinking machine.

Until we can ensure that AI applications and systems can operate ethically - there will always be a need for a human being to be a part of the decision making process. You cannot move to a fully automated system without an ethical compass without creating risks of really bad outcomes.

Technology advancements are moving ahead at an accelerating rate. In under 20 years, we went from dial up internet to the mobile web in the palm of our hands with smartphones. AI is on a similar trajectory. I am a nerd at heart - I find AI just as exciting as the next nerd. It has the potential to do great things for humanity. But we have to temper our enthusiasm for technology with a healthy dose of ethics. This is one of the times where slowing down will be a good thing in the long run.

Individual humans and often our societies are flawed. There is no such thing as perfection when it comes to people. People and societies often miss the mark when it comes to right and wrong, and we probably always will. While we are flawed, we aspire to be ethical - and it is these aspirations that make us truly human. We know right from wrong, even when we digress from what is right. As we look to the future of AI, we must decide whether or not we will embed our own ethical flaws into a system which may not be able to make the same judgement calls that we do.

Thanks for reading my thoughts on this! I invite you to keep this dialogue going.