In a recent article in Kathimerini, Minister of State Akis Skertsos stressed the key role that Artificial Intelligence (AI) can play in the public sector reform process in Greece. Indeed, AI can act as an efficiency multiplier in many areas of public governance.

However, the rapid development of AI poses risks that have been highlighted even by pioneers and funders of this technology, such as Geoffrey Hinton, Elon Musk and Steve Wozniak. This is also the wavelength of the AI Safety Summit organised by the British Government on 1-2 November at Bletchley Park, the home of Alan Turing’s “Hut 8” team of decoders during World War II.

Nevertheless, it is not only the future, possibly existential, threat posed by the so-called Artificial Generative Intelligence (AGI) (intelligence which will no longer be limited in its capabilities or controlled by human intelligence) that requires special attention, careful study and analysis. Even existing “milder” forms of AI, which human intelligence is able to control (although it does not necessarily fully understand how they work, the so-called ‘black box’ problem), carry risks.

A concrete example of such risks is the case of the Dutch government’s SyRI (System Risk Indication) programme designed to identify the risk of fraud in the welfare system. According to the Dutch judiciary, due to the processing and meta-analysis of big data sets, SyRI could inadvertently create links influenced by biases. These biases could relate to factors such as a person’s lower socio-economic and/or immigration status, potentially leading to unfair and biased results which either by choice of the administration (as in the case of SyRI) or due to the complexity of the algorithms (black box problem) could not be judicially reviewed, thus undermining effective judicial protection.

There are of course areas of public administration where the application of AI can offer solutions in combination with risk mitigation measures. One such example is public procurement in the health sector. AI can offer solutions to a wide array of tasks, from the relatively simpler ones such as supplier management, invoice/payment processing, to more complex ones such as market horizon scanning for timely reaction in case of supply chain disruptions and compliance of the procurement ecosystem with the obligations arising from stricter carbon emission reduction targets.

However, it would be a big mistake to entrust solely AI with the task of simplifying the labyrinthine administrative procedures in Greece which, despite the commendable efforts of the former and current leadership of the Ministry of Digital Governance, continue to plague citizens in their dealings with the state. The simplification of administrative procedures is a separate process which must begin earlier to make the implementation of AI solutions more effective and to allow for the proper planning of risk mitigation measures. There are tools at EU level which can be used and in the development of which Greece should participate more actively.

The establishment of the AI Advisory Committee under Professor Daskalakis is a particularly encouraging step in the right direction.  In the age of AI, Greece cannot afford to be a mere observer.