Article contributeur

Artificial intelligence: silver bullet in a Pandora’s box

DISCUSSION BOARD. What to expect from the accelerating progress of artificial intelligence (AI)? By Virgile Perret & Paul Dembinski

Keystone
As Etienne Perrot explains, «by developing algorithms based exclusively on computation and stochastic data, AI substitutes statistical correlations for (human) causal relationships».

L’Observatoire de la finance a lancé au printemps 2020 «From Virus to Vitamin». Ce conseil rassemble différents économistes suisses et étrangers. Depuis janvier, l’initiative s'est transformée en un groupe de discussion mensuelle sur l’économie et la finance. «L’Agefi» publie leur synthèse.

Question 25, February 2022. What to expect from the accelerating progress of artificial intelligence (AI) in manufacturing as well as in services and public administration? How to prepare for it, does AI deserve - and if yes why - specific regulation, guidance or taxation?

Advances in computing power, the availability of data and of new algorithms have led to rapid progress in the field of Artificial Intelligence (AI), «the single most influential human innovation in history» (Archana Sinha).

Deployed wisely, AI holds the promise of addressing some of the world’s most pressing challenges, but it may also have destabilising consequences on some key dimensions of economic and social life.

In manufacturing, AI promises to increase productivity by extending the capabilities of humans and by helping businesses achieve more efficiency, including through direct automation, predictive maintenance, reduced downtime, 24/7 production, etc. However, as Ezekiel Kwetchi Takam rightly points out, «this automation will be deployed at the expense of some human labour whose skills will be deemed irrelevant».

In public administration, the adoption of AI can contribute to better public services, for example by interacting with service users through virtual assistants or by enabling smarter analytical capabilities and better understanding of real-time processes. There is a risk, however, when using data, to amplify existing biases and produce discriminatory and unethical outcomes for different individuals. Moreover, as Etienne Perrot explains, «by developing algorithms based exclusively on computation and stochastic data, AI substitutes statistical correlations for (human) causal relationships. The lack of conflicting interpretations inherent in AI creates a human cost that can already be observed in predictive justice.»

Regulation is needed both on input (quality of data, bias avoidance, data ownership) and output (reasonability of results) rather than on software itself
Domingo Sugranyes

To mitigate these risks, Domingo Sugranyes argues, regulation is needed both «on input (quality of data, bias avoidance, data ownership, purpose of automated processes) and output (reasonability of results) rather than on software itself.» In addition, as pointed out by Etienne Perrot, «any regulation in AI must aim to keep the human factor and its responsibility at the heart of all economic, judicial and political decisions».

While AI can have positive impacts for humanity, it seems to have raised at least as many questions as it has answered, opening Pandora’s box. To ensure that the costs, benefits and risks generated by AI are equitably shared among citizens and stakeholders while respecting democratic values and human rights, public authorities have a crucial role to play to set up such a regulatory framework. One way to do so could be «to define a specific taxation of AI that will serve to:

1) Guarantee an unemployment income to those who could be called the «economic neglected by the technological evolution»;

2) Finance their training and their professional conversion.» (Ezekiel Kwetchi Takam)

“… the lack of conflicting interpretations inherent in AI creates a human cost…”

The main challenge of Artificial Intelligence is twofold: on the positive side, the optimisation of industrial and commercial processes (with the elimination of human intermediaries); on the negative side, on the one hand, the weakening of intra-systemic information and the concealment of risk, and on the other hand, the deshumanization of administrative and judicial relations. By developing algorithms based exclusively on computation and stochastic data, AI substitutes statistical correlations for (human) causal relationships. The lack of conflicting interpretations inherent in AI creates a human cost that can already be observed in predictive justice. Any regulation in this area must aim to keep the human factor and its responsibility at the heart of all economic, judicial and political decisions.

Étienne Perrot

Jesuit, Dr. Economics sciences, editorial board of the magazine Choisir (Geneva), editorial adviser of the journal Études (Paris).


“… it is urgent to define a specific taxation of AI…”

Artificial intelligence will increase the productivity in manufacturing while developing new areas of employment and expertise. Unfortunately, this automation will be deployed at the expense of some human labour whose skills will be deemed irrelevant for this new market. It is therefore urgent to define a specific taxation of AI that will serve to:

1.   Guarantee an unemployment income to those who could be called the “economic neglected by the technological evolution”

2.   Finance their training and their professional conversion.

In public administration, AI will contribute to develop a new "algorithmic governance”, defined by Müller-Birn and al. as a form of governance that integrates algorithmic systems. It is therefore important to regulate this new governance dynamic by involving citizens in the process: through democratic participation, the citizen must be the evaluator and co-constructor of this new form of public service.

Ezekiel Kwetchi Takam

PHD candidate in theological ethics of artificial intelligence at the University of Geneva.


“… there is a need to educate a capacity for discernment in digital environments…”

AI and big data are transforming areas like advertising, media, finance, insurance, manufacture, health and medicine, weather forecast, catastrophe prevention, justice administration… It is a universal purpose technological revolution with huge impact on work, culture, civil life and social structures. Regulation is needed on input (quality of data, bias avoidance, data ownership, purpose of automated processes) and output (reasonability of results) rather than on software itself. Public intervention is needed to protect competition against monopolistic practices, which is difficult because tech markets work in ways which differ from classical ones. Above all, there is a need to educate a capacity for discernment in digital environments at all levels and age groups, and to promote universal access to digital services.

Domingo Sugranyes

Director of a seminar on Ethics and Technology at Pablo VI Foundation, past Executive Vice-Chairman of MAPFRE international insurance group.


“…it is vital to avoid the exclusive focus on economic efficiency…”

AI comprises forms of intelligence demonstrated by machines in three different areas: 1). Advanced automation; 2). Computer-based central-nervous-system research; and 3). Bridging 1 and 2 through the use of neurophysiological models in designing machines to perform practical tasks, mostly robots (see the well-known classification of James Lighthill). Already important parts of social and economic organization have been affected by AI, and this process can be expected to continue. In addressing institutional design it is vital to avoid the exclusive focus on economic efficiency as narrowly understood, in particular replacing mostly but not exclusively semi-skilled and unskilled labour with machines. Innovations can already be conceived that increase the tasks ordinary workers are able to perform, for example, through new technologies allowing workers to perform tasks previously performed by more skilled people or enabling the provision of more specialized services by existing workforces.

Andrew Cornford

Counsellor, Observatoire de la Finance; past staff member of UNCTAD, with special responsibility for financial regulation and international trade in financial services.


 “ . . . AI may become the single most influential human innovation in history…”

Artificial intelligence is raising important questions for society, economy, and governance. The world is on the cusp of revolutionizing many sectors through AI, but the way AI systems are developed need to be better understood due to the major implications these technologies will have for society as a whole. In order to balance innovation with basic values, recommendations include improving data access, increasing government investment in AI, promoting AI workforce development, creating a civic advisory committee, engaging with state to ensure they enact effective policies. These processes need to be better understood because they will have substantial impact on general masses soon, and for the foreseeable future. AI may become the single most influential human innovation in history.

Archana Sinha

Head, Department of Women’s Studies, Indian Social Institute, New Delhi, India.

Commentaires

Virgile Perret / Paul H. Dembinski

Observatoire de la finance Project Manager / Director

Les articles suivants pourraient aussi vous intéresser