Writers are having an existential crisis! Why?, You ask, well this recent article by Stanford talks about artificial intelligence, CoAuthor which helps people in writing, and how it poses a threat to traditional writers.
The report described how an AI is fully capable of grasping the topic at hand and producing relevant content. The idea of using AI for writing is floating around for quite a while now. Although, it makes things easier for some people but is it idealistic?
What is the aim behind CoAuthor?
Firstly the article starts off by quoting the idealism behind an AI “There’s a lot of nagging doubt in my mind about this. Is that okay? I mean, shouldn’t humans write their own content? And does this mean the writing is on the wall for an entire profession?“
Will there be no more writers? We all have to ask ourselves what our roles in this brave new world will be,” immediately shocking the readers right after telling them that this whole bit was written by Co-Author.
CoAuthor is an interface, a dataset, and an experiment all in one. It comes from a doctoral student in computer science at Stanford University, Mina Lee, and her advisor Percy Lang who is an associate professor of computer science at Stanford.
Lee’s aim was to ease up the writing process using these language models. She believes that these models have a lot of potential and can help a lot of people. She even gave an example of several awards winning books and essays Co-authored by such models.
According to the article, enhancing human productivity and creativity is the goal here, rather than replacing them. “We think of a language model as a ‘collaborator’ in the writing process that can enhance human productivity and creativity, helping to write more expressively and faster,” Lee says.
Google’s predictive search is an easy example of a writing-help model. Its algorithms help by suggesting the next word while typing. Although it’s a hit and miss sometimes.
How does it work?
CoAuthor is predominantly based on GPT-3, a large language model from open AI, trained on multiple datasets available on the internet. Although, the idea of seeing something original comes out of a model that works on ‘already existing texts’ seems like a stretch. But hey, It works! And works so good that traditional writers have acknowledged its feat.
Lee and her team went far and beyond to push writers out of their comfort zone. They were on the lookout for writers deviating from their routine and writing stuff they wouldn’t have written otherwise.
How is CoAuthor deemed effective?
Lee and her team built CoAuthor as an interface that records writing sessions at a keystroke level, selecting large datasets as writers worked with GPT-3 and analyzing the collaboration between human writers and AI.
The researchers asked 60 people to write 1440 stories all assisted by GPT-3. Accepting the AI suggestions was completely up to the individuals. CoAuthor can hence collect all sorts of data for researchers to analyze like; how often the writer accepts suggestions, which suggestions get accepted, how they were edited, and how they influenced the subsequent writing.
Since it is an analytical tool CoAuthor can determine how “helpful” were the accepted suggestions. It can even interpret rejected suggestions as a proxy for the writer’s taste to improve its suggestions for future language models.
Do writers like it?
In conclusion, the writers took a survey about their relative satisfaction with the resulting work. They said the ideas and suggestions were quite new and useful. At times writers disregarded its suggestions as they took the writer in a different direction than intended.
“I especially found the names helpful,” wrote one of CoAuthor’s participants after the survey. “I was actually trying to think of a stereotypical rich jock name and the AI provided me with one. Perfect!”
Lee found out that however useful, it may have a negative influence on the writer’s sense of ownership in the resulting text. CoAuthor’s creators also found that the use of large language models increased writer productivity as measured in the number of words produced and the amount of time spent writing with considerably less grammatical errors.
Are Large Language models a cause of concern? Is it a threat to the profession of traditional writers? Comment down below.