Conclusion

• Pages: 127-128

127

CONCLUSION

AI has the potential to address challenges facing our world, stretching





divisions, which exacerbate existing forms of inequality.

AI is a heterogeneous set of very powerful and very useful techniques for

discovering complex connections, but that does not make it an intelligent

subject. The appearance of intelligence does not make a model intelligent,

          

with the specialists who design them, use them and are able to interpret their

results, like a doctor does when making a diagnosis.

The opacity of AI systems, rather than engendering trust, erodes the trust

humans have in machines, especially as machines move towards autonomous

decision-making (even more so if they apply self-learning). But ethics in AI

should not be limited to the development of transparent and fair systems.

Sustainable AI solutions must take into account other human dimensions

such as the personal, social, legal, organisational and even institutional.

The proposition of Trustworthy AI is a distraction that can cover up the

hidden work of underpaid employees, blur responsibilities and justify the

      

           

AI that can be trusted in areas previously reserved for human judgement,

merely by meeting certain established prerequisites, leaves humanity out of

the equation. The most important human activity, “how should I act to be



Coeckelbergh notes:

Moreover, other sorts of questions also need to be answered in policy proposals: not

only what should be done, but also why it should be done, when it should be done, how

much should be done, by whom it should be done, and what the nature, extent, and

urgency of the problem are. (Coeckelbergh, 2020b)

Some authors propose Virtue Ethics for AI. Not as a genius combination

of algorithmic requirements, but for those involved in AI research, develop-

ment, deployment and exploitation. Decision-makers in the life cycle of AI

systems must be virtue-ethicists, i.e. people with habits of acting appropri-