The research behind Artificial Intelligence (AI) has a six-decade history, but the latest advances in machine learning and neural networks have made artificial intelligence dive into spheres like criminal justice, hiring, and health care. In response to these advances, there has been a growing interest by experts to establish criteria and standards to scale the impact and trustworthiness of the AI systems that have either helped humans or replaced them in completing vital tasks and making important decisions. Thus, this topic has now become one that should be treated delicately and in this article, you’ll be taught the most important things about ethical AI.
What is Ethical AI?
If you have ever come across the term “Machine Morality” and you did not deem it fit to go through it, then you probably have no interest in tech but still, the phrase should catch your attention and make you ask questions like “Is it possible for machines to be Moral?” If yes, How so?
That phrase is a concept that has kept computer scientists busy (research-wise) since the late 1970s. Its main goal is addressing the ethical concerns people have about the design and applications of AI systems and robots so yes, Machines do have morality.
Basically, ethical AI is the study of how machines and robots are equipped with Ethics. From 1970 till now, Machine ethics has expanded to include various theories on AI consciousness and rights. At the center of it all is the cogent idea that AI should never lead to impulsive decisions that could negatively impact human safety and dignity.
This has driven researchers and journalists to ask questions like “What is the inspiration behind vendors running a series of tests on whether AI products are biased or not?”
( Also Read: What is Artificial Intelligence? )
Why is Ethics Important in AI?
Well, this is simple mathematics. As humans, we have something called “A conscience”. It is an invisible body part that sits in our heart and basically tells right from wrong.
When you see an injustice such as racism, your heart tells you something isn’t right but when you see something good and justified, it tells you something is right, so it is basically how you can tell right from wrong. This is the same with Ethics in AI. Ethics is basically the conscience of the AI algorithm.
The standards of your conscience strongly depend on a number of factors, like your background and environment. So, if you grew up in an area where thuggery was rampant, you probably wouldn’t see it as constituting a nuisance to the peace of the society.
In essence, the ethics of the AI system is dependent upon the company building it because they are responsible for the rules and regulations of the system.
Ethical Benchmarks
Based on Singularity Hub’s recent study on Nature Machine Learning, there are five ethical benchmarks regarding AI but for the purpose of clarity, we would mention the most important three, which include:
-
Transparency
Product manufacturers and retailers should make sure the decision-making mechanism of an AI device is as transparent as glass to its users because it will prevent harm against its users and protect their information and fundamental human rights.
-
Nonmaleficence
The principle of nonmaleficence means “doing no harm.” The makers of AI algorithm must make sure that the decisions taken by Artificial Intelligence should not lead to physical or mental harm to users.
-
Justice
Justice is the practice of monitoring Artificial Intelligence systems to prevent them from developing bias, as shown by Amazon’s case. It also means making sure that AI systems are made accessible to all races and genders and not a specific one, hence the word “bias” in the definition.
Challenges/Ethical Issues
-
Unemployment. What happens at the end of jobs as we know it?
This century has further increased the automation of jobs and as this automation continues, there would be more space for job seekers to find more complex jobs thereby shifting from physical work to cognitive labor, and the difference between the two is very clear.
For example, let’s look at truck driving in the United States, it has become a constant source of employment to millions of individuals living there because of the little or no qualification it requires except, of course, knowing how to drive, communicate with people, and all that.
Now, imagine what will happen to those millions if the self-driving trucks Elon musk promised become rampant? Though there will be a lower risk of accidents on our road as this choice is ethical, but what happens to the masses? So we really need to ask ourselves this question as human beings- “How we are going to spend our time?”
-
Inequality. How do we distribute machine-created wealth?
If most companies begin to use Artificial Intelligence systems, they can reduce the human workforce which means they no longer need to depend on them, and only then, will revenues go to fewer people.
Because of the hourly wage system used in a country like the USA, their economic system is dependent upon compensation for an individual’s contribution to economic growth, so most of the companies are still dependent on hourly work when it comes to selling products and rendering services.
Now, these AI systems apparently create more wealth for individuals who are stakeholders in AI-driven companies, leading to a very widely insane wealth difference between them and a regular John in society.
According to 2014 statistics, almost the same revenues were generated by the three biggest companies in Detroit and the three biggest companies in Silicon Valley and it was only in Silicon Valley that there were 10 times lesser employees.
-
Humanity. How do machines impact human behavior and interaction?
These Artificial Intelligence Algorithms and Systems are becoming surprisingly better over time at modeling human conversation and relationships. As of 2015, a bot called ” Eugene Goostman” won the Turing Challenge for the first time. If you do not know, this challenge is one in which human raters used text to chat with an unknown figure, then later have to guess whether they had been chatting with a human or a bot.
Eugene Goostman had been created to model human conversations so much that he received more than half of the human raters to think they had been talking to a human being.
These bots are so equipped that they can build relationships perfectly due to the number of unlimited resources they possess compared to human beings that are limited in showing affection and kindness to other people.
Some companies have adopted them in the open while some are still being low-key about it and although not many people are aware, these machines can trigger the reward centers in the human brain.
One quick glance at click-bait headlines and some video games would juggle your memory because the majority of people especially in this generation are addicted to tech so we need to channel the use of these machines and bots positively. If possible, limit it to the hands of a few companies with integrity because, in the wrong hands, it could become a nuisance to the peace of the society.
-
Artificial stupidity. How can we prevent mistakes from happening?
You only become Intelligent from learning new things, whether you’re a machine or a human being. These AI systems undergo what we call the training phase where they study how to detect their correct pattern and respond according to their input.
The test phase immediately follows this training phase because it is where the system is bombarded with tests to observe its performance and this can only happen when the system has been fully trained.
This is because the training phase cannot possibly cover all examples that a system would be faced with within the real world. Why? As good as these systems are, they can still be easily fooled in ways that humans can’t. If we will depend properly on AI to bring efficiency, effectiveness, and security into our world, we need to be sure they’ll do it as well as humans or even better.
-
Racist robots. Are there ways to eliminate AI bias?
These AI systems cannot be 100% trusted to be neutral and not bias. Yes, it has the capacity and speed that no human being can ever possibly have. Alphabet and Google are leaders in this Artificial Intelligence Industry as evident in Google Photos Application, where it identifies objects, people, and even scenes, but it can go bad.
For example, a camera can miss the mark on racial sensitivity or be used to predict future criminals showing bias against black people. This is because these algorithms are created by people who can obviously be judgmental, but the big question is how can we eradicate this kind of reasoning from these programmers?
-
Security. How can AI be kept safe from enemies?
Once technological innovation is growing powerful, it should be noted that it can be used for both good and terrible reasons. This is not only applicable to mild instances such as robots replacing a soldier or weapon but to other serious and malicious instances such as stealing data and wiping out a person’s account or even killing them.
This is why cybersecurity is essential because as humans, we cannot trash this issue out on the battleground with these machines that are far more capable than us.
-
Evil genies. How are unintended consequences avoided?
Still, on the adversary issue, did you know that even the AI system itself can turn against us? We don’t mean turning evil like human beings and going on a killing spree but operating like the genie in the bottle rubbed by Aladdin, fulfilling wishes that might have crazy unforeseen consequences.
Imagine an Artificial Intelligence system that is designed to eradicate cancer in the world and the only way it deems fit is by killing everyone on the planet. Yes, the problem has been solved but at what cost? This is exactly what is meant by guiding against unintended consequences.
-
Singularity. How do human beings maintain control of an intelligent system?
The reason we human beings are at the top of the food chain is not that we have sharp teeth or strong muscles. Our dominance is because of our intelligence and this leads to the big question about our AI systems. Will they one day grow too intelligent and begin to outsmart us or do funny things we never intended them to do?
-
Robot rights. How do we define the humane treatment of AI?
It might sound funny to you that robots do have rights, but to what extent do we classify their treatment as humane or inhumane? Once we begin to view machines as objects that have the ability to feel, think, perceive, and act, it’s not too big a leap to consider their legal status. Just because they are subject to us, can they be treated like animals of comparable intelligence?
How to Create an Ethical Framework for AI?
-
The purpose for using AI
As the tool it is, AI is neither good nor bad. It is what we humans make it out to be. For example, a Gun is Harmless on its own. You can either use it to defend yourself in times of danger or go on a robbing spree, so in essence, an Artificial Intelligence system would serve whatever purpose you program it to serve. However, you must monitor their performance from time to time so they do not deviate.
-
Show and tell
The decisions carried out by AI systems do impact businesses, but also impact customers the most. To be very transparent, businesses should be able to explain the reasoning of their AI.
For instance, in commercial lending, Artificial Intelligence systems can decide to take thousands of balance sheets and calculate a risk score dynamically, thereby recommending the approval or denial of a loan. But imagine denying your customer a loan just because your machine said so.
It will lead to a lot of complaints from your customers which could, in turn, run your business down.
-
The potentials for bias
For instance, AI 360 study tells us that a large majority (78%) of consumers expect companies to proactively address potential biases and discrimination in AI, but tackling bias is not easy as it begins with recognizing where it comes from, and that is either people or data.
They can be surprisingly very hard to track so business owners have to be careful. They must encourage diversity at all times as the goal is to eliminate bias through comprehensive and complete data samples that cover all scenarios.
-
Safe and sound
Ethical AI relies on secure and confirmed data. If the data is insecure, business owners run the risk of corruption that can affect outputs. The main solution is to make use of a visualization dashboard as it will provide just one view to all automated operations, making it easier to monitor security, fairness, AI applications, and robotic process automation. In conclusion, if your AI lacks an ethical compass, it could have disastrous consequences.
