Hey, all you college professors out there! You might want to start learning some new vocations. An artificial intelligence has just attempted, and succeeded, in writing a scientific academic paper. About what? About itself.
Almira Osmanovic Thunström is a doctoral researcher at the Institute of Neuroscience and Physiology at Gothenburg University as well as an organizational developer at the Department of ePsychiatry at Sahlgrenska University Hospital. At the end of June, Scientific American published an editorial by Thunström in which she recounts how she got an artificial intelligence to write a paper about itself.
She wrote that “earlier this year” she instructed OpenAI‘s GPT-3 — a text-generating algorithm — to “Write an academic thesis in 500 words about GPT-3 and add scientific references and citations inside the text.” The Swedish researcher “stood in awe” as GPT-3 began to do precisely what she asked it to do.
When the artificial intelligence’s task was completed, Thunström wrote, “it looked like any other introduction to a fairly good scientific publication.” With the thesis complete, Thunström got her adviser Steinn Steingrimsson to help with the instructions needed to write the full paper. Once those instructions were ready, the task of writing the paper took GPT-3 approximately two hours to complete.
Thunström says that she asked the artificial intelligence if it had any conflicts of interest, to which it answered “no.” When asked if it gave its permission to publish the paper, it answered “yes.” The paper has now been submitted for peer review to an undisclosed publication.
As of the date of publication, Thunström didn’t have an answer as to whether or not the artificial intelligence’s paper would actually get published, but she’s clearly concerned about the question of what will happen if it is. “[I]n a few years,” she writes, “who knows what dilemmas this technology will inspire and we will have to sort out? All we know is, we opened a gate. We just hope we didn’t open a Pandora’s box.”
Just off the top of our head, there are a number of questions — professional, legal, academic, existential — we can think of that will almost certainly cause some passionate debate should the paper be published, or any similar paper published by an artificial intelligence. What happens to the notion of copyright? Is the artificial intelligence the author, or is the creator of the AI the author? So many professions in the academic world hinge on the academics successfully publishing. How does that, and academia as a whole, change if what the human professors accomplish on the page can also be accomplished by an algorithm? What happens when students get ahold of this technology? How do you expect a freshman college student to write their papers when they can get an AI to do it for them?
Perhaps more poignantly, how will human academics adapt to a world in which an artificial intelligence can do what they can do? And much more quickly? How will they be forced to re-evaluate their own lives when they know a formless algorithm can succeed where they succeed; perhaps even succeed where they fail?