Written by Andrés García Higuera.
On 7 November 2023 the European Parliament’s Panel for the Future of Science and Technology (STOA) brought together experts on Artificial Intelligence and Large Language Models (LLMs), policy makers and representatives from civil society at a workshop entitled ‘Generative AI: opportunities, risks and challenges’.
STOA Vice‑Chair Ivars Ijabs (Renew; Latvia) opened the event by identifying the 2022 breakthrough in LLMs as a core technology and driver of change that heralded the release of Chat GPT. This technology brings immense opportunities in fields ranging from teaching to industry, but also poses challenges for data governance and has job market repercussions. In view of the economic impact on the EU of potentially missing-out on the technological revolution of recent decades, taking an active role in the AI revolution now appears crucial for the EU’s economic prosperity and competitiveness.
The event continued with a first panel concentrating on the technological aspects of AI. Professor Jan Hajič set the scene by commenting on different AI models, from single-task applications to machine learning‑based systems that focus on speech recognition, transformer architecture for choosing the next word in a sentence, and generative systems that produce a complete text. The panellists commented on current trends, the lifecycle of this technology and the scope of its application. Multilingualism, translation systems and digital language equality were also discussed in relation to digital training techniques and access to data and infrastructure. The assessment of risks and opportunities differ across categories of users, and public acceptability appears to be strongly linked to training. The code of conduct on disinformation, the Digital Services Act, risk assessment and mitigating measures were also mentioned, in relation to fake news and the enforcement of copyright rules within the framework of the AI act. Guardrails are necessary to moderate research and deployment so that innovation can move forward with the required guarantees.
Yordanka Ivanova (DG CNECT) represented the European Commission at the event. She commented on decisions taken while drafting the AI act, which is now under discussion. Future‑proof legislation requires clear definition of terms, systems, opportunities and risks, which is not always possible when dealing with emergent technologies like generative AI. The Commission’s coordinated plan on AI includes action to support AI excellence and development in the EU. Noting that the AI act has reached the final stage of trilogue negotiations, Yordanka Ivanova considers the three institutions (the Commission, Parliament and Council) could reconcile their different approaches and conclude negotiations by the end of the year. The proposed regulation specifies different levels of risk, and was drafted in a broad manner so that it will remain applicable to future technological developments. Since systems such as Chat GPT can be used for low and high-risk applications, the Council’s common position proposes to oblige providers to share the information they require to function, so that users are able to identify the level of risk. The Parliament’s draft focuses on the foundation models and the design and implementation phases, while searching for a common approach that would enable a global convergence with international players.
The second panel focused on the social and ethical effects of Generative AI. It was introduced by Ana García Robles, Secretary General of the Big Data Value Association, who highlighted multidisciplinarity and the need for ethical design, deployment and regulation of AI. The panel discussed the business aspect of AI, AI’s impact on European industrial development and the need to take advantage of the opportunities AI presents, including job creation, while addressing its challenges and risks. The panel called for AI governance and regulation as well as for investment in research and development. As generative AI is a technological breakthrough posing extraordinary regulatory challenges, legislators should protect people by requiring independent audits and enforcing regulations.
STOA Vice-Chair Ivars Ijabs closed the event by underlining the need for risk management while fostering development and innovation. He also pointed out that gaining public acceptance and trust in generative AI will require future‑proof regulation.
A web-stream recording of the event is available on our website.
Your opinion matters to us. To let us know what you think, get in touch via stoa@europarl.europa.eu and follow us on Twitter/X at @EP_ScienceTech.
Comments are closed for this post.