Like technology, in general, artificial intelligence that includes automation and machine learning or hyper-automation should be considered morally neutral. At least, that is our hypothesis today.
Let us all assume that AI, like any technology, can be used in an ethically responsible manner or, conversely, in a way that causes harm to others, whether voluntary or not.
AI, A TECHNOLOGY SO DIFFERENT FROM THE OTHERS?
Some technologies work directly for the good of societies, such as progress in the field of medicine or the environment. Societal impacts are beneficial, observable and above all voluntary. We must understand that this is, above all, because the mission of activities in these areas serves an honorable cause. Other technologies work necessarily for bad things like war. Others for the ideal of maximizing goods. Under no circumstances is the technology itself put in the dock. So why does AI seem to be guilty today?
We need to see how AI is a particular technology. Because it’s smart? But then, how smart is it? One could answer that it is because it is able to predict events. Or because it is able to make decisions. To be autonomous then, shall we say. That’s when fears arise. Something autonomous, able to make decisions will have ethical choices among the choices to be made. And where man encounters dilemmas, the machine decides. This is because in its ability to choose, the machine has no moral doubt that can hold it back.
WHY TALK ABOUT AI BIAS?
But before we go so far in our fears and assumptions, we must start from the fact that we are responsible for the direction that a particular technology can take. In our direction, we judge the impact of our technology on society by analyzing the positive or negative consequences it entails.
For example, such progress in medical imaging allows us to intervene and predict with much more subtlety. We have a positive, calculable impact. But these progress in imaging are not limited to the medical field. The intelligent pixel opens up a huge field of research whose facial recognition is one of the application. It is also known that applications using this facial recognition are sometimes biased (intentionally or not) related to skin color that may result in racial selection.
Thus, algorithmic progress are not related to a particular field, they are free and accessible to any good programmer. It is difficult in these circumstances to assess the consequences of a particular algorithm. How can we not fear this new intelligence?
The threat is real and the fears are justified, but we can act. Because, just as we learn to be responsible for our planet, we will become responsible for this machine intelligence. The advantage of fears is that they take us out of our comfort zone and push us to think about how to defend ourselves. They make us evolve.
I proposed that AI Ethics is an invitation to humanity to take stock, re-examine itself and choose what it truly wishes to become.
– Matthew James Bailey, Inventing World 3.0, p.237
But why not just enforce the laws we know? Well, because these laws don’t exist. And what do we find upstream of the laws? Ethics or the evaluation of moral values. That is why we are questioning this ethic of AI today? Opinions differ and the ground is very fertile. There is no current consensus on the approach we need to have in AI ethics.
Some theories advocate the plurality of values, others focus on the universality of certain values. This AI ethic is in progress. Observatories are being created to put forward debates in the ethics of artificial intelligence, such as OBVIA and the AlgoraLab in Quebec, or Turing Institute and many others. The mission of these new forms of ecosystems is to promote the adoption of responsible AI.
My doctoral researches in AI ethics suggest that we can work internationally to build standards and that we will have no choice but to adopt an evolutionary, and why not, agile methodology.
Original article in French: [ANALYSE] L’intelligence artificielle fait-elle preuve de morale? – CScience IA
Ansgar Koene, Algorithmic Bias, Addressing Growing Concerns, IEEE Technology and Society Magazine, juin 2017
Bernard Marr, The intelligent revolution, Transforming your business with IA, ed. Bernard Marr, 2020
Harini Suresh, John V. Guttag, A Framework for Understanding Unintended Consequences of Machine Learning, Cornell University, février 2020.
La Commission Européenne, Libre blanc: Intelligence artificielle, Une approche européenne axée sur l’excellence et la confiance, Bruxelles, PDF, 2020
Matthew James Bailey, Inventing world 3.0, Evolutionary Ethics for Artificial Intelligence, ed. Matthew James Bailey, 2020
Pascal Bornet, Ian Barkin, Jochen Wirtz, Intelligent automation, welcome to the world of hyperautomation, ed. Pascal Bornet, octobre 2020