Artificial intelligence (AI) is in the process of becoming the
technology of the future. AI refers to the ability of a machine to
simulate human behaviour, such as reasoning, planning and
creativity, particularly though the use of learning algorithms.
Businesses are increasingly using AI to develop more automated,
personalized and customer-oriented services. AI also provides new
opportunities to reinforce and facilitate the detection and reduction
of risk and fraud and to enhance regulatory compliance.
However, the use of AI is fraught with risk, as illustrated by Deloitte
in one of its research studies1:
‰ Quality, quantity and relevance of the data used. The results of AI
systems are dependent on the quality and quantity of the data. If
the datasets used to build algorithms contain biases, the
generated algorithm is likely to reflect such biases as well, or
even amplify them.
‰ Operational opacity (black box, when referring to AI). Unlike
older generations of AI where systems made very clear, humandetermined decisions, new generations will rely on very complex
statistical methods based on thousands of parameters. All these
factors will make the final decision difficult to interpret, if not
impossible to explain, by humans.
‰ Possible malfunctions. Algorithms do not possess the conceptual
understanding and common sense that humans do, which are
necessary to assess radically new situations.
As pointed out in the latest reports of the World Economic Forum,
artificial intelligence is at a turning point. In the short term, it
seems important that the development of artificial intelligence
should meet minimum criteria for governance, ethics and risk
management. Such evaluation, according to Deloitte2, should focus
on proving the reliability of the algorithms used (in terms of
internal and external verification), the intelligibility of models and
the interactions between humans and intelligent algorithms.
It is therefore crucial that the duties of the Governance Committee
be reviewed to include this matter and to develop a code for the
use of artificial intelligence in order to assure shareholders and
interested parties (stakeholders) that the development and use
thereof involve placing humans at the heart of the machine,
guaranteeing the veracity, security and confidentiality of the data
that feed it, and regulating the algorithms so that they incorporate
diversity and overcome biases in the decision making process.