Unless you have been hibernating in a cave for the last year, you are surely aware of all the excitement surrounding the recent advances in artificial intelligence (AI)!
The deployment of the most advanced large language models (LLMs) and the chatbots based on their fine-tuned versions (ChatGPT being the most famous) has generated a turbulent mixture of unseen levels of enthusiasm and concern, reaching the mainstream media and even the governments of many countries. At the extreme of this concern, a group of well-known AI people (including from OpenAI, maybe ironically, the very company that created ChatGPT) has claimed that AI poses an existential risk for humanity, on par with pandemics and nuclear wars, which isn't being given adequate attention.
This focus on AI’s existential threat has raised criticism: on the one hand, many experts see it as unrealistic in the near to medium future; on the other hand, it may overshadow other significant risks, downplaying urgent global issues such as climate change. Instead of focusing on uncertain future risks, attention should be paid to pressing issues related to the current use of AI: exacerbating inequalities and biases, undermining freedom and/or privacy, producing disinformation that could destabilize democracies. These issues demand political and regulatory solutions, but they also present technical and scientific challenges in the pursuit of responsible AI (RAI). Conducting research towards RAI is essential to harness the huge potential of AI to serve humanity, while ensuring its responsible and fair use across various scientific, technological, and business domains. A balance must be sought between exploiting AI's potential and mitigating its possible adverse effects through responsible practices.
The newly created Center for RAI (CRAI), funded by the Portuguese Recovery and Resilience Plan, is at the forefront of R&D towards RAI and aims to position Portugal as a major player in this field. IT is one of the research partners in this consortium, with our participation spearheaded by André Martins, a world-leading expert (and two-time ERC grantee) in natural language processing (NLP), the area of AI within which LLMs and chatbots, including ChatGPT, were developed.
IT is thus well positioned to play a relevant role in the crucial quest for responsible AI.
(Coordinator of the thematic line Information and Data Sciences at IT and Professor at Instituto Superior Técnico)