Almira Osmanovic Thunström, researcher at the Institute of Neuroscience and Physiology at the University of Gothenburg and organizational developer at the Department of Electronic Psychiatry at Sahlgrenska University Hospital, attempted an original experiment to say the least.
A particularly conclusive test
The scientist thus asked GPT-3, the artificial intelligence created by OpenAI, to write a 500-word academic article… regarding herself. To complicate things a bit, she also had to add scientific references and citations in the text. And the surprise was great for Almira Osmanovic Thunström when she found that the text generator had accomplished its mission perfectly.
As she explains: It looked like any other introduction to a fairly good scientific publication. “. It remains to be seen whether this article will be published in a scientific journal and whether the peers will see nothing but fire.
The researcher clarifies on this subject: In a few years, who knows what dilemmas this technology will inspire and that we will have to solve? All we know is that we opened a door. We just hope we haven’t opened a Pandora’s box. »
As pointed out by our colleagues from Giantfreakinrobot, this attempt also raises many questions, particularly in terms of copyright. Could the writing of scientific articles be delegated to an AI?
In the past, other experiments have been carried out using this text generator. Thus, the British newspaper The Guardian asked GPT-3 to write an editorial. However, this did not convince everyone, and in particular Daniel Leufer from Mozilla who described this test as ” absolute joke. « It would have actually been interesting to see the eight essays that the system actually produced, but editing them in this way only adds hype and misinforms people who won’t read the details “, he regretted.
