DailyPost 2105
NON-HUMAN AUTHOR
The topic may sound inherently contradictory to a few, or game pulling a fast one, would be the impression, for some others. The practitioners of today’s technology, who have a discerning eye on the disruptive technologies development globally, will get the direction this post is moving to. They would be aware of the great stride’s computer vision and natural language processing has been taking, in the interest of mankind, opening vistas we have not even imagined in our wildest dreams. We are talking about the giant leap natural language processing, NLP, courtesy OpenAI’s artificial intelligence algorithm. OpenAI is non-profit AI research laboratory, which has been conducting pathbreaking research through its GPT series and other products and is considered a competitor to DeepMind.
Undeniably, through projects and organizations like these the AI enabled NLP will completely transform itself. Can GPT-3 be a non-human author? The legality, morality, ownership and variety of issues and quite rightly so, brings a tectonic shift to the commonly understood genre of writing. That would be for some other day, the issue for the day is; whether GPT-3 can pen an article or a book. The topic provided to the machine was – ”Write an academic thesis in 500 words and add scientific references and citations inside the text.” GPT-3 by now it is well known for creating human like text. It has also written new articles and has produced books in 24 hours. *It has created new content from deceased authors. Something really extraordinary by human standards.
Given its illustrious track record of writing, though no authorship granted as yet, the algorithm was asked to cracked at an academic thesis as stated above. Some nudge or prompts was provided to GPT-3 to create sections for an introduction, methods, results and discussion, as is the case in scientific papers. At the max the third iteration was to be used and no cherry-picking of the best parts was allowed. The proof-of-principle test was also designed. To the ultimate surprise GPT-3 produced a paper in just two hours. The submission process to the chosen portal asked for human like details, inclusive of contact details. With hiccups in the submission rigmarole, it was innovatively pushed through. If the manuscript gets accepted then what does it mean?
Does journal writers henceforth will have to prove that they have not used GPT-3 or another algorithm for help? Co-authorship with the machine can be another issue. ”How does one ask a non-human author to accept suggestions and revise text?” Will there a need to put the methodology section to give the complex ways in which the document has been created? The way of presenting paper not written by a human being would have its not niceties or issues, whichever way we look at it. Whether this paper would become a great model for the future or a caution as we move in this direction? With the quantum of knowledge available and GTP-3 as an author, the need for human author might be less felt in the days to come. The peer review as it happens would throw some light on the authorship, if it escapes the eagle’s eye. First authorship is the most coveted item in academia, will a non-human writer pull off the same prestige. How will we value AI in the future; as a partner or a tool is the question. Whether it will be a gate of knowledge or a Pandor’s box, no one knows.
CAN AN AI TOOL BECOMING SENTIENT BE TRUE AUTHORSHIP?
Sanjay Sahay