AI at Infomedia: 2023 in Review
2023 proved to be a prolific year for Infomedia. Our researchers were among those at the forefront of scientific advancements, introducing innovative tools, and making substantial contributions to informed discussions on AI.
![ai](https://www.uib.no/sites/w3.uib.no/files/styles/content_main/public/media/adobestock_166966052_0.jpeg?itok=C_RztK5W×tamp=1704978050)
Main content
In 2023, generative artificial intelligence (AI) emerged as a key storyline. The technology has conquered our imagination, even clinching the title of "word of the year" in various corners of the globe, including Norway.
The preceding year witnessed a growing discussions on the nuanced nature and capabilities of artificial intelligence. Simultaneously, actors worldwide acknowledged the substantial risks posed by AI.
“In 2023, we could see calls for transparency and regulation gaining urgency. Towards the end of year, European Union achieved a milestone by finalizing world’s first AI regulation aiming at ensuring better conditions for the development and use of artificial intelligence”, explains Infomedias’ head, Marija Slavkovik.
In parallel, the Norwegian government announced boost research efforts in the field, committing a minimum of NOK one billion over the next five years.
Driving Innovation with Responsibility
Amidst these developments, Infomedia remained committed to researching on responsible technology.
2023 proved to be prolific at the department - marked by an impressive array of scientific publications. Key research areas encompassed large-scale language models, the ethical challenges facing machines, and the application of AI for content production and verification.
Below highlights of some scientific advancements:
- In the first-ever study of this type, NLP researcher Samia Touileb contributed to examining the harmfulness and toxicity of Scandinavian pre-trained Language Models. The study demonstrated that some of the popular Norwegian, Danish, and Swedish language models generate harmful and gender-biased stereotypes. Touileb has been vocal about the potential adverse outcomes of using such models in real-world situations.
- In collaboration with Faktisk.no., PhD candidate Sohail Ahmed Khan developed two new AI prototypes : a deep learning-based military vehicle classification web prototype which can predict the make/model of a military vehicle present in an image, and a language detection, transcription and translation web service based on OpenAI’s Whisper model.
- A content-filtering tool on social media identifies a picture of a child without a shirt by the water's edge. Should it be treated as child pornography and removed, or is it a harmless vacation memory shared by a grandmother after a trip to Gran Canaria? The ethical challenge highlightes the increasing considerations machines face as tasks are automated. Marija Slavkovik emphasizes that when machines become our moral deputies, we bear a responsibility to equip them with the tools to handle ethical challenges. In an article published in the Journal of Artificial Intelligence Research, the Infomedia’s head together with colleagues from China and Luxembourg demonstrated how such tool, an artificial ‘moral counselor’, can be programmed.
Making headlines
Last year, our researchers have garnered significant visibility in the public domain, frequently featured in the media.
Among others:
- Lars Nyre and Bjørar Tessem contributed to discussions surrounding the billion-kroner allocation for artificial intelligence research in Khrono.
- Marija Slavkovik offered insights on the regulation of AI at the European level on tek.no.
- Per Christian Magnus and Asbjørn Leirvåg explored the evolving perception of journalists as historically insensitive and inhumane, considering the emergence of AI-driven journalism in Medier24.
Curious to know more? Headlines featuring our researchers and their expertise can be found at AI at Infomedia.