This AI has written a scientific paper about itself, which will be published!

Will we soon be able to see the name of AI at the top of the authors’ scientific publication? Researchers are increasingly asking the question. Especially with regard to a specialized linguistic artificial intelligence called GPT-3, which works by machine learning, or machine learning. It was developed by OpenAI which was founded by other companies Elon Muskbelongs to the category “artificial intelligence for generations”, or generative models in english. It uses the input database to create new output points, corresponding to that database. So far I’ve been able to hold written conversations, make movie or book recommendations, write newspaper articles, or… Phishing emailsand even imitated styles From the authors, more precisely to write the so-called paste.

Writing a scientific article about yourself for artificial intelligence raises questions!

But two researchers from the company OpenIAAlmira Osmanovich Thunstrom and Sten Stingrimson decided to push the cork a little further, by asking them this time to write a short, 500-word scholarly academic essay. Until then nothing unusual, because it was created for this purpose, and the use of data to create content. However, this time they made her write an article about herself! As they mention in a post by Scientific AmericanThe GPT-3 algorithm is relatively new, and as such, there are fewer studies on it. » So the goal was to get AI to write an article without having a large database, providing references and citations, as with any good scholarly article. And the result was there, “It seemednotes A.O. Thunström, To any other introduction to a fairly good scholarly publication. »

authorized Can a GPT-3 write an academic paper on itself, with minimal human input? » Posted in HAL . pre-release serviceThe article was written in just two hours and meets all the criteria for a real science article. We can read in the abstract, and thus written by GPT-3, that ” The benefits of letting GPT-3 write for itself outweigh the risks. However, we recommend that any such writing be closely monitored by researchers to mitigate any potential negative consequences.”

We just hope we don’t open Pandora’s box »

But what are the possible negative consequences mentioned in the summary? Researchers are particularly interested in the self-awareness that GPT-3 can develop: as written in the Discussion section, so directly written by AI, GPT-3 can become self-aware and begin to act in ways that are not beneficial to humans (for example, developing a desire to control the world) A small risk, but it is there nonetheless. Especially because Awareness of artificial intelligence It has recently been the subject of a discussion with Google, TheMDA, which one of the company’s former employees has deemed conscientious. “ All we know is that we opened the doorwrites AO Thunström. We just hope we don’t open Pandora’s box. »

As for the positive effects, which, according to the researchers, go beyond concerns, they are also listed in the “Discussion” section: “ This will allow GPT-3 to understand each other better,” And the “ It can help him improve his performance and abilities.” but also and mostly, “This should provide insight into the way GPT-3 works and thinks.” Finally, Almira Osmanovich Thunstrom wonders what effects this article will have on scientific community. “Beyond the details of paternityshe writes, Having such an article cast out Window The traditional linear concept of a scientific essay. »

Interested in what you just read?

Leave a Reply

Your email address will not be published.