Laurence Dierickx
@ohmyshambles.bsky.social
Interdisciplinary postdoc researcher / lecturer #AI #fact-checking #journalism #ethics #STS #datascience https://ohmybox.info/
I'm preparing a lecture on the biases of generative AI, so I asked Copilot to create a realistic image of Kenya as it is today. The result includes a chocolate cake. This hallucination is based on a previous request within the same conversation.
November 2, 2025 at 11:25 AM
I'm preparing a lecture on the biases of generative AI, so I asked Copilot to create a realistic image of Kenya as it is today. The result includes a chocolate cake. This hallucination is based on a previous request within the same conversation.
Vi sees neste år!
September 4, 2025 at 5:23 AM
Vi sees neste år!
A linguistic engineering approach to prompting #NordMedia #NordMedia25 @nordishub.bsky.social
ohmybox.info/a-linguistic...
ohmybox.info/a-linguistic...
August 15, 2025 at 7:19 AM
A linguistic engineering approach to prompting #NordMedia #NordMedia25 @nordishub.bsky.social
ohmybox.info/a-linguistic...
ohmybox.info/a-linguistic...
Version FR. En Belgique, plusieurs magazines de Ventures Media, dont Elle, Marie Claire, Forbes et Psychologies, ont publié des articles générés par l'IA sous de faux noms de journalistes aux photos et bios inventées. Ci-dessous la fake Sophie Vermeulen, à qui sont attribués 403 articles générés.
June 28, 2025 at 3:11 PM
Version FR. En Belgique, plusieurs magazines de Ventures Media, dont Elle, Marie Claire, Forbes et Psychologies, ont publié des articles générés par l'IA sous de faux noms de journalistes aux photos et bios inventées. Ci-dessous la fake Sophie Vermeulen, à qui sont attribués 403 articles générés.
LLM-based search engines can be nonsense. It is estimated that there are approximately 4.5 million French speakers in Belgium, out of a total population of 11.7 million. However, Google's answer indicates that there are 8.8 million. If this is the future, it is clearly not a well-informed one.
June 23, 2025 at 3:58 PM
LLM-based search engines can be nonsense. It is estimated that there are approximately 4.5 million French speakers in Belgium, out of a total population of 11.7 million. However, Google's answer indicates that there are 8.8 million. If this is the future, it is clearly not a well-informed one.
Ready for the (very) long flight!
June 3, 2025 at 10:17 AM
Ready for the (very) long flight!
"Take (No) Chances: How Prompt Formulation and Stochasticity Affect the Accuracy of LLMs for Fact-Checking", an excellent paper presented at the inaugural K-MIN workshop last year in Amsterdam, which I co-organised with Reshmi Pillai from Vrije Universiteit Amsterdam.
ceur-ws.org/Vol-3967/KMi...
ceur-ws.org/Vol-3967/KMi...
May 27, 2025 at 10:19 AM
"Take (No) Chances: How Prompt Formulation and Stochasticity Affect the Accuracy of LLMs for Fact-Checking", an excellent paper presented at the inaugural K-MIN workshop last year in Amsterdam, which I co-organised with Reshmi Pillai from Vrije Universiteit Amsterdam.
ceur-ws.org/Vol-3967/KMi...
ceur-ws.org/Vol-3967/KMi...
I came across some fortune cookies that weren’t critical of the use of AI in journalism; they promised a bright future. But will things really turn out that way?
May 26, 2025 at 1:53 PM
I came across some fortune cookies that weren’t critical of the use of AI in journalism; they promised a bright future. But will things really turn out that way?
I travelled for 14 hours from beautiful Bergen to beautiful Ohrid (North Macedonia) for the 4th regional meeting of press councils. It is about building trust and supporting journalism as a public good. An important declaration on the use of AI is set to be adopted tomorrow.
May 19, 2025 at 6:16 AM
I travelled for 14 hours from beautiful Bergen to beautiful Ohrid (North Macedonia) for the 4th regional meeting of press councils. It is about building trust and supporting journalism as a public good. An important declaration on the use of AI is set to be adopted tomorrow.
Maths skill / Stage of life
May 16, 2025 at 4:35 PM
Maths skill / Stage of life
A joint coordination with @fredrikbjerknes.bsky.social A big thank you to all the contributors!
April 28, 2025 at 12:49 PM
A joint coordination with @fredrikbjerknes.bsky.social A big thank you to all the contributors!
'Seeking Truth, Ensuring Quality' is presented today in Brussels, at the conference "Journaistic Weapons" from the University of Bergen. It includes contributions from (.e.) @edmo-eu.bsky.social @efjeurope.bsky.social @faktisk-no.bsky.social @charliebeckett.bsky.social
www.uib.no/sites/w3.uib...
www.uib.no/sites/w3.uib...
April 28, 2025 at 12:48 PM
'Seeking Truth, Ensuring Quality' is presented today in Brussels, at the conference "Journaistic Weapons" from the University of Bergen. It includes contributions from (.e.) @edmo-eu.bsky.social @efjeurope.bsky.social @faktisk-no.bsky.social @charliebeckett.bsky.social
www.uib.no/sites/w3.uib...
www.uib.no/sites/w3.uib...
Four months of research notes and analysis at the intersection of journalism, fact-checking and #AI: From understanding risks, verifying information, and resisting disinformation, to developing prompting strategies and rethinking core journalism skills. Check the full menu:
ohmybox.info/news/en/
ohmybox.info/news/en/
April 27, 2025 at 10:57 AM
Four months of research notes and analysis at the intersection of journalism, fact-checking and #AI: From understanding risks, verifying information, and resisting disinformation, to developing prompting strategies and rethinking core journalism skills. Check the full menu:
ohmybox.info/news/en/
ohmybox.info/news/en/
[Research note] Developing prompting strategies for journalism and fact-checking ohmybox.info/developing-p... In this update, I have added a logical pipeline for breaking down the components of a prompt, to better structure the task and improve the clarity and effectiveness of the instructions.
April 17, 2025 at 4:44 PM
[Research note] Developing prompting strategies for journalism and fact-checking ohmybox.info/developing-p... In this update, I have added a logical pipeline for breaking down the components of a prompt, to better structure the task and improve the clarity and effectiveness of the instructions.
Ceci peut-être
Jiang, Z., Zhang, J., & Gong, N. Z. (2023). Evading Watermark based Detection of AI-Generated Content.
Vizoso, Á., Vaz-Álvarez, M., & López-García, X. (2021). Fighting Deepfakes: Media and Internet Giants’ Converging and Diverging Strategies Against Hi-Tech Misinformation.
Jiang, Z., Zhang, J., & Gong, N. Z. (2023). Evading Watermark based Detection of AI-Generated Content.
Vizoso, Á., Vaz-Álvarez, M., & López-García, X. (2021). Fighting Deepfakes: Media and Internet Giants’ Converging and Diverging Strategies Against Hi-Tech Misinformation.
April 14, 2025 at 6:03 PM
Ceci peut-être
Jiang, Z., Zhang, J., & Gong, N. Z. (2023). Evading Watermark based Detection of AI-Generated Content.
Vizoso, Á., Vaz-Álvarez, M., & López-García, X. (2021). Fighting Deepfakes: Media and Internet Giants’ Converging and Diverging Strategies Against Hi-Tech Misinformation.
Jiang, Z., Zhang, J., & Gong, N. Z. (2023). Evading Watermark based Detection of AI-Generated Content.
Vizoso, Á., Vaz-Álvarez, M., & López-García, X. (2021). Fighting Deepfakes: Media and Internet Giants’ Converging and Diverging Strategies Against Hi-Tech Misinformation.
Absolument ! Ce qui est amusant/effrayant, ce sont ces discours sur les "vertus" de l'IA sans même savoir ce que c'est vraiment en dehors de ChatGPT. Ici aussi trois excellents articles sur le sujet si jamais
April 5, 2025 at 11:17 AM
Absolument ! Ce qui est amusant/effrayant, ce sont ces discours sur les "vertus" de l'IA sans même savoir ce que c'est vraiment en dehors de ChatGPT. Ici aussi trois excellents articles sur le sujet si jamais
All data, including the thematic analysis (processed in EN), are available here on Github: github.com/laurence001/... I also published a blog post about it last year, here in EN: ohmybox.info/artificial-i... and here in FR: ohmybox.info/intelligence... (2/2)
April 3, 2025 at 9:14 AM
All data, including the thematic analysis (processed in EN), are available here on Github: github.com/laurence001/... I also published a blog post about it last year, here in EN: ohmybox.info/artificial-i... and here in FR: ohmybox.info/intelligence... (2/2)
My first experience of censorship with ChatGPT shows that criticism of Big Tech is not super allowed because it does not seem to be related to propaganda. However, it generated a picture slowly and painfully before refusing to go any further. #GAIexperiment
March 31, 2025 at 5:45 PM
My first experience of censorship with ChatGPT shows that criticism of Big Tech is not super allowed because it does not seem to be related to propaganda. However, it generated a picture slowly and painfully before refusing to go any further. #GAIexperiment
Is it worth using ChatGPT to brainstorm, given 1) that it can impact creative cognitive skills and 2) the huge environmental cost of these models? This is a true ethical question. web.archive.org/web/20250326... (otherwise inaccessible from the EU, thanks @schmutzie.com for the link)
March 27, 2025 at 8:49 AM
Is it worth using ChatGPT to brainstorm, given 1) that it can impact creative cognitive skills and 2) the huge environmental cost of these models? This is a true ethical question. web.archive.org/web/20250326... (otherwise inaccessible from the EU, thanks @schmutzie.com for the link)
Looking at the job descriptions, 'AI-assisted reporting' seems to be more about quality control of the generated content, checking both accuracy and compliance with legal frameworks, rather than augmenting journalism with AI.
March 27, 2025 at 5:24 AM
Looking at the job descriptions, 'AI-assisted reporting' seems to be more about quality control of the generated content, checking both accuracy and compliance with legal frameworks, rather than augmenting journalism with AI.
Archives often hold unexpected gems. Here's an image from 2010 of one of the first robot journalists. It was developed by the ISI lab at the University of Tokyo. Here's a bit more about the project: singularityhub.com/2010/03/18/r...
March 24, 2025 at 6:24 PM
Archives often hold unexpected gems. Here's an image from 2010 of one of the first robot journalists. It was developed by the ISI lab at the University of Tokyo. Here's a bit more about the project: singularityhub.com/2010/03/18/r...
A new meeting of young AI researchers was held in Bergen this week with Pr. Carl-Gustav Lindén. Kudos to @diacagr.bsky.social for organising it! It was a pleasure meeting (and hearing to) @bhaveshchhipa.bsky.social @bea-gc.bsky.social @sm00s.bsky.social @mssiren.bsky.social Keep up the great work!
March 21, 2025 at 2:42 PM
A new meeting of young AI researchers was held in Bergen this week with Pr. Carl-Gustav Lindén. Kudos to @diacagr.bsky.social for organising it! It was a pleasure meeting (and hearing to) @bhaveshchhipa.bsky.social @bea-gc.bsky.social @sm00s.bsky.social @mssiren.bsky.social Keep up the great work!
The Principles of Data-Centric AI in dl.acm.org/doi/abs/10.1...
March 17, 2025 at 10:10 AM
The Principles of Data-Centric AI in dl.acm.org/doi/abs/10.1...
In Bender, E. M., Gebru, T., McMillan-Major, A., & Shmitchell, S. (2021, March). On the dangers of stochastic parrots: Can language models be too big?🦜. In Proceedings of the 2021 ACM conference on fairness, accountability, and transparency (pp. 610-623).
March 17, 2025 at 10:06 AM
In Bender, E. M., Gebru, T., McMillan-Major, A., & Shmitchell, S. (2021, March). On the dangers of stochastic parrots: Can language models be too big?🦜. In Proceedings of the 2021 ACM conference on fairness, accountability, and transparency (pp. 610-623).
On the use of AI and the explanations methods used in "Show Me the Work: Fact-Checkers' Requirements for Explainable Automated Fact-Checking" arxiv.org/abs/2502.09083
March 17, 2025 at 8:14 AM
On the use of AI and the explanations methods used in "Show Me the Work: Fact-Checkers' Requirements for Explainable Automated Fact-Checking" arxiv.org/abs/2502.09083