Science fiction authors were the first to detect and stretch to the maximum the ethical issues that could arise with Artificial Intelligence. What will happen if the robots develop feelings, they wondered over and over again? Their imagined responses have given us great works of fiction that now, while jokingly chatting with an AI chat, we can’t help but think. A few months ago, in fact, an engineer from a large technology company came to claim that he was completely convinced that a chatbot he worked on had become sentient (the company —and almost the entire world— denied it and granted him a leave with salary).
Leaving aside that question, which it seems we are not facing at the moment, Artificial Intelligence, in its current state, does present some ethical dilemmas . The most talked about, and the one that is being tackled the most, is that of the biases that a technology designed by humans and trained to learn from them can inherit. If the AI is developed by a fairly diverse group, its algorithm will be trained with certain biases and loopholes that the developers may not even be aware of.
One of the most extreme examples is when image recognition bots have made headlines by identifying a photograph of a black person with an animal, because they did not receive enough images of racialized people in their training. But there are also situations that can go unnoticed. When, for example, AI is used in recruiting , it can only serve to perpetuate a situation of inequality by selecting resumes similar to those of current workers and automatically discarding those that differ in some way.
The issue is so important and could prove so decisive in the lives of millions of people that in November 2021 the 193 Member States of the UNESCO General Conference adopted the Recommendation on the Ethics of Artificial Intelligence , the first agreement and normative instrument world on the subject. The objective is not to clip the wings of the AI, but, aware of its great potential to improve things, but also to make them worse, to guide it on the right path.
In addition to biases, AI raises other ethical dilemmas such as what it means for privacy ( apps that get your data by following us, for example), or for attribution of liability (who is responsible if an autonomous car causes an accident?).