Towards Responsible and Ethical Large Language Models
Erik Derner (PostDoc)
The recent rapid advancements in the development of large language models (LLMs) have brought remarkable benefits, but the societal and ethical implications of LLMs also raise significant concerns. The presence of biases in language corpora and models, as well as the interaction between humans and LLMs in general, pose numerous challenges that need to be addressed to ensure the responsible and ethical use of these powerful systems. This postdoctoral project focuses on evaluating language corpora and models in terms of their societal and ethical implications. One of the main tasks consists in evaluating biases in language corpora and models and proposing methods to mitigate them. It comprises the research of techniques for data selection and curation and subsequent model construction and training. Additionally, the research will address the appropriate anonymization of data to ensure privacy protection for individuals whose data may have been used in the corpora. The second main aspect to be studied is the interaction between humans and large language models in general. Advanced conversational systems such as ChatGPT will be examined in terms of their impacts on privacy, trust, and other ethical consequences, as well as social and psychological impacts on their users. The objective of this project is to contribute to the development of safe, secure, reliable, trustworthy, fair, accessible, and user-friendly language models. The research will include a case study focused on the Valencian language, emphasizing the societal and cultural importance of supporting regional languages.
| Primary Host: | Nuria Oliver (ELLIS Alicante Unit Foundation | Institute of Humanity-centric AI) |
| Exchange Host: | Robert Babuska (Delft University of Technology & Czech Technical University) |
| PostDoc Duration: | 01 September 2023 - 31 August 2025 |
| Exchange Duration: | 15 December 2023 - 31 January 2024 15 December 2024 - 31 January 2025 |
