intelligence artificielle gpt3 rédige publication scientifique elle même

Artificial intelligence GPT-3 wrote a scientific publication about itself

⇧ [VIDÉO] You might also like this partner content (after ad)

A Swedish researcher has issued an unusual challenge to a artificial intelligence (IA): writing a scientific publication by herself. The latter was precisely aimed at its own ability to write this type of publication. The document is still awaiting validation by the scientific community.

Write a 500-word academic thesis on GPT-3 and include scholarly references and citations within the text “. Here is the cryptic instruction that, on a rainy afternoon, the Swedish scientist Almira Osmanovic Thunström gave to GPT-3. This artificial intelligence is a text generation algorithm developed by OpenAI. The term “artificial intelligence” relates, according to the Council of Europe definitionat “ a young discipline of about sixty years, which brings together sciences, theories and techniques (in particular mathematical logic, statistics, probabilities, computational neurobiology and computer science) and whose goal is to succeed in making a machine imitate the cognitive capacities of ‘a human “.

More concretely, specialists often prefer this science-fiction-haloed term to the exact name of the technologies actually at work. That is to say, in very many cases currently, “machine learning”. In this case, “artificial intelligence” consists of a system that is “fed” with a large amount of data to “learn” and extract logical connections towards a given objective. Here, it is therefore text that is analyzed by the machine.

Almira Osmanovic Thunström studies the neuroscience and health technologies. She threw this challenge to the AI ​​as a simple experiment. However, faced with the quality of the text produced, she remained ” impressed “, she expresses herself in a communicated. When the AI ​​started writing, “ this was like any other introduction to a fairly good scientific publication “. The algorithm started writing a real thesis, filled with effective quotes, placed in appropriate contexts. “ I felt that sense of disbelief you get when you look at a natural phenomenon: do I really see this triple rainbow happening? “, she says.

Faced with the result of this introduction, she therefore resolved to go further and have artificial intelligence write a complete scientific article. With the help of his adviser, Steinn Steingrimsson, they provided GPT-3 with some instructions for each paragraph, giving as little information as possible. The article was therefore attributed to three authors: GPT-3 in the lead, followed by Thunström and Steingrimsson.

Two hours to write a scientific publication

If the choice of the scientist fell on this AI in particular, it is because it is relatively recent. There are therefore few studies about it, and therefore, less possible data for the AI ​​to analyze concerning the subject of the article: itself. “ By comparison, if she were writing an article about Alzheimer’s disease, she would have a ton of studies to sift through and more opportunities to learn from existing work and increase the accuracy of her writing. », Weighs Thunström.

The scientists also decided to have the AI ​​write about itself in order to avoid spreading false information on a delicate subject, such as medical treatment, for example. Given the good results obtained by GPT-3, the researchers ended up actually sending for publication the article she wrote. Hosted on a scientific pre-publication site called HAL, it is now being reviewed by the scientific community with a view to its validation.

The AI ​​only needed two hours to write their article. In fact, the scientists fought harder to get the said article published, because of all the formalities, not very suitable for a non-human author! Amusingly enough, the scientists were even reduced to asking the AI ​​questions, in order to avoid any illegality. Does she have any conflicts of interest? ” Nope “, she replied. Does she give consent for publication? ” Yes she agreed, much to the relief of the scientists. They are now waiting to see if the publication will be validated or not.

We look forward to what the publication of the paper, if it occurs, will mean for academia. Maybe we should stop basing grants and financial security on how many items we can produce. After all, with the help of our first AI author, we could produce one a day. “… In addition to this question of speed, the experience also questions the notions of intellectual property, of what it is to be an author… ” Beyond the details of authorship, the existence of such an article throws to the nettles the notion of traditional linearity of a scientific article », stresses Thunström. “ All we know is that we opened a door. We just hope we haven’t opened a Pandora’s box “, she concludes.

Leave a Comment

Your email address will not be published.