Browsed by
Category: AI

Study shows: GPT-3 can—technically—be the author of scientific papers. An interview with Almira Thunström.

Study shows: GPT-3 can—technically—be the author of scientific papers. An interview with Almira Thunström.

A system could technically be considered a co-author—or even the main author!—of a scientific paper! This is the mind-boggling insight presented by human AI researchers Almira Osmanovic Thunström and Steinn Steingrimsson from Gothenburg – with the assistance of their co-author GPT-3.

Article: Does GPT-3 qualify as a co-author of a scientific paper publishable in peer-review journals according to the ICMJE criteria? – A Case Study.

“We tested the system against the ICMJE criteria, and we found that a system could technically be considered a co-author—or even the main author—of a paper,” says Almira Osmanovic Thunström. “The system would have to be the one to come up with the topic, write paragraphs, analyze and critically evaluate its production and be responsible—correct and discover mistakes—in its paper.”

Almira and her human co-author Stein Steinn Steingrimsson—both from the University of Gothenburg—noted that the system had no problem to fulfill three out of the four main criteria for co-authorship: “drafting the work or revising it critically for important intellectual content; final approval of the version to be published; and agreement to be accountable for all aspects of the work.” Only when it came to referencing, the system could not fulfill the criteria.

Almira Osmanovic Thunström is a project manager and researcher in child and adolescent psychiatry. Currently, she is working mostly with chatbots and virtual reality. Her background, however, is somewhat mixed.
“I am a biologist on paper, an assistant nurse in my heart, and an engineer by trade,” she says. “I get to do quite a lot of fun and exciting things, but at the moment, I am mostly interested in developing digital tools for treating anxiety disorders.”

In the summer of 2022, Almira published an article in Scientific American that spurred some attention: We Asked GPT-3 to Write an Academic Paper about Itself—Then We Tried to Get It Published. After exploring automated writing for fun in 2021, she realized that with a little bit of luck and the right settings, she could ask the system to help her with academic work. Fascinated, she started to experiment, and in the process, it became important for her to document her thoughts. Her notes eventually became an op-ed for what is probably the world’s most famous popular science magazine.
“As someone who has been creating chatbots for the past two decades, I’ve been around for many of the twists and turns regarding natural language processing and large language models. I usually approach all these things with a childish sense of exploration and with each new model, software, or hardware I encounter, I want to do something silly with it.”

After realizing that paper-writing text robots are an interesting and philosophical dilemma, Almira noted that in a world where scientists write like robots, we are bound to create robots that write like humans.
“We wanted to raise a debate on transparency and authorship,” she says. “The question of authorship and accountability is a big pink elephant in the room we academics would rather avoid talking about. After all, that senior researcher on your paper that hasn’t even read it, is paying your bills. Any paper where your name is first will keep paying your bills in terms of grant money, which is often dependent on having a steady production rate of papers. With the help of a system like GPT-3, one could produce papers very fast; after all, if your text can be expanded and completed by a system, you can do everything in half the time.”

Hi, Almira. This is truly fascinating! Please tell us how you proceeded when you wrote the study.

“We asked the system to write about itself. Mainly because it has so little data on itself (Terminator plot headache warning: the system is trained on 175TB of text data from a variety of databases up until 2020/2021, however, very little if any data existed on GPT-3 because it didn’t exist at that time, only GPT-2 did). We wanted to test its capability. It was a case of childish curiosity and an interesting thought experiment. During the peer review of the submitted paper, we realized that the main issue reviewers had (remember this is pre-ChatGPT and ChatGPT publications!) was if the system could follow the ICMJE criteria for authorship on a paper. So we did what the reviewers asked of us: checked the system against the criteria.”

What was your reaction when you went through the results?
‘“The world should see this!’ was my first reaction. And ‘Oh no!’ was the second. How on Earth do I present this data? What will this mean for myself and others as we advance!?!”

 What should scientists over the world learn from this study?
“Have an open mind, and dare to experiment. Let us write less like robots. Let us not blindly embrace authorship on papers; both humans and systems need to pass the ICMJE criteria. ChatGPT for example did not pass the ICMJE criteria on many different levels. Our paper has a very extensive discussion part that I think a lot of scientists will appreciate.”

How do you think the advent of ChatGPT and similar technology will affect science in the long run?
“I’m not so sure they will, or at least I think it will take time. My hope was that we’d rethink how we teach, write, design studies, and collaborate. I’m seeing institutions play out the worst scenarios in their head about the consequences of these systems, which I think is valid to some point, but not as productive as acknowledging a challenge and finding an innovative way forward. The most excited I’ve seen the universities get is when they discovered Crossplag.”

“In our SciAm article, I said we hope we hadn’t opened a Pandora’s Box. I think in the eyes of academic institutions, we very much did. Unlike Pandora’s box, hope was however not left inside. I see individual teachers and academics who are looking at how we can innovate how we teach in lieu of the takeover of automated writing—that gives me hope!”


Please note: all texts on this blog are produced by me, Olle Bergman, or invited human writers. When robot-written sections are added to serve as examples or to demonstrate a point, this is clearly indicated.

Feeling like a 21st century man

Feeling like a 21st century man

Sometimes the 21st century feels overwhelming. For example when you read the The Memo from Life Architect and ponder over where mind-bending initiatives like Muse and Leta will take us.

Then I feel comfort in Ray Davies’s bluesy howl of alienation in 20th Century Man.

Play it loud and enjoy the autotune-free vocals, the energizing noises from the acoustic instruments—especially the scratching of the (metal?) bottleneck on the fretboard–and the drummer’s (Mick Avory) non-perfect but groovy beat.


”There is no doubt NLG could have a profound impact“ – The AI ONE ON ONE interview

”There is no doubt NLG could have a profound impact“ – The AI ONE ON ONE interview

Anyone interested in how AI will affect the world should have a look at the LinkedIn-based blog/newsletter AI ONE ON ONE, run by the two academic researchers Aleksandra Przegalinska and Tamilla Triantoro. The purpose of their blog is to discuss advantages and limitations of current AI systems and explore the possibilities of using AI for good. RWR got in touch for an interview, focusing on NLG in general and the advent of ChatGP3 in particular.

Behind AI ONE ON ONE are two colleagues residing at different sides of the Atlantic Ocean. Aleksandra Przegalinska is a professor and Vice-President of Kozminski University in Poland, and a Senior Research Associate at the Harvard Labour and Worklife Program. Tamilla Triantoro is a professor at Quinnipiac University in the United States, and a leader of the Masters in Business Analytics program.
“The blog has been a good run so far,” says Aleksandra Przegalinska. “In summer of the last year, we started experimenting with different features of GPT-3, and then, in December, ChatGPT took the world by storm, making these conversations even more relevant.”



Hi, Aleksandra and Tamilla! What motivated you to start this blog?
“We both have a solid background in human-computer interaction research. One of the interesting things about working with new cutting-edge technologies is to observe how humans interact with them, and what are the possible improvements that can be implemented. Right now is the first time when natural language generation has been democratized. As more people get acquainted with this technology, it becomes more relevant to explore the possibilities. Through our blog we explore the creative applications, and in our research, we focus on the applications for business, and the effect of generative AI on routine and creative processes at the workplace.”

Let’s talk NLG in general and GPT-3 in particular! You recently noted that there is “incredible hype surrounding [ChatGPT] these days”. What do you think spurred this hype?

”The hype around the GPT family of models, particularly ChatGPT, is largely due to the impressive capabilities it has demonstrated in natural language processing tasks. The key strength of ChatGPT is its ability to generate highly coherent and human-like text in the form of a conversation. Unlike the previous version of GPT, ChatGPT requires minimal fine-tuning, and we observed this during our experiments. Newer technology, GPT-3.5 that powers ChatGPT, allows for a more seamless interaction and that is truly captivating.”

“The hype is also related to the excitement regarding the transformers’ remarkable ability to perform creative tasks – such as storytelling, generating interview questions, article titles and writing programming code. The transformers such as ChatGPT perform tasks, that were once considered impossible for machines to accomplish. This allows for more innovation and creativity while making it all accessible not only to industry giants but also to small businesses and individuals.”

Have you used GPT-3 for other things than studying how it works and experimenting with its features? 

”We both have interest in the future of work. AI influences the way the work will be performed in the future, and an interesting question is in what circumstances humans will collaborate with AI. Currently we study the collaboration of humans and AI via the lens of individual acceptance. We created an application for business use that integrates GPT-3 technology. The application assists knowledge workers in performing tasks of various difficulty and levels of creativity. We measure the levels of collaboration and attitudes towards AI and devise a program of improvement of the human-AI interaction and collaboration at the workplace.”

In what situations do you think GPT-3 can be useful for professional groups such as researchers, investigating journalists, and business people?

“GPT-3 certainly has the potential to be a useful tool for a wide range of professional groups due to its ability to generate human-like text and understand natural language input. Some specific examples includes, for instance, researchers: here GPT-3 can be used to assist with literature reviews and data analysis. It could help generate summaries of scientific papers and articles, assist with data analysis and visualization, and even support writing research proposals and papers. Another example could be investigating journalists: GPT-3 could be used to assist with fact-checking and research by quickly providing additional information on people, organizations, or events mentioned in articles. It could also be used to generate summaries of complex documents such as legal filings, and help analyze and organize large sets of data.”

“Obviously, the biggest impact will be seen for businesspeople: here GPT-3 can be used for automation of repetitive and low-value tasks such as writing reports, composing emails, and creating presentations. Additionally, it could also assist with market research and trend analysis, as well as generate strategic plans, and help identify new business opportunities.”

Can you see any ethical problems arise – short-term and long-term?

“Well, it is important to keep in mind that GPT-3 is a machine learning model and the results it generates still need to be verified, but it can save time for professional groups by automating time-consuming, repetitive and low-value tasks, supporting with research and analysis, and generating new insights and ideas. We need to understand that we humans are the real fact-checkers here.”

“Other issues include the following:

  • bias – because language models like GPT-3 are trained on a large dataset of text, which can perpetuate and amplify societal biases that are present in the data;
  • privacy – because the use of large language models for natural language processing can raise privacy concerns, particularly when the models are used to process sensitive or personal information;
  • explainability – because language models like GPT-3 are complex systems that are difficult to understand and interpret, which can be a problem in situations where it is important to know how a model arrived at a certain decision or conclusion.”

Do you think that upcoming generations of NLG technology  (GPT-5…6 …7 …) may attain better results than humans in certain situations? 

“It is possible that future generations of natural language generation technology may attain better results than humans in certain situations. The capabilities of language models like GPT-3 have already surpassed human-level performance in tasks such as language translation and text summarization. As the technology continues to improve and more data is used to train these models, it is likely that they will perform even better on these and other tasks.”

“For example, GPT-5 or GPT-6 models might be able to generate news articles or summaries, they might be able to compose a novel or a poem, it might be able to compose persuasive or convincing fake news or deepfake videos, even able to impersonate a person on internet, these are areas where human are currently capable but natural language generation technology may outmatch them in future.

“However, it’s worth noting once again that such models have significant limitations. They can only perform as well as the data and the algorithms that they have been trained on, so if the training data is biased or inaccurate, the model will be too. Additionally, Language models currently lack common sense, self-awareness or creativity. So, While they may be able to perform certain tasks better than humans, they may not be able to understand context.”

How will humanity’s relation with the written word be affected by NLG technology? Is it a bigger disruption than Gutenberg’s printing technology?

“Natural Language Generation (NLG) technology has the potential to significantly change the way people interact with written text. With NLG, machines can automatically generate written content in a way that mimics human writing. This means that people may rely more on computer-generated text for tasks such as writing reports, news articles, and other types of written content.”

“It’s hard to say whether NLG will be a bigger disruption than Gutenberg’s printing press, as it depends on how widely the technology is adopted and how it’s used. Currently it’s a bit of a social experiment on mainstreaming AI, we will see where it takes us. Having that said, the printing press allowed for the mass production of books, making written information more widely available and helping to spread knowledge and ideas. In a similar way, NLG has the potential to increase the efficiency and speed of producing written content, but it’s still in early stages and it’s hard to predict how it will evolve over time.”

“There is no doubt though that NLG could also have a profound impact on how people write and consume written information, as it may change the way we think and communicate, allowing us to express ourselves in new ways, and could have implications on areas such as employment, education, and language.”


Please note: all texts on this blog are produced by me, Olle Bergman, or invited human writers. When robot-written sections are added to serve as examples or to demonstrate a point, this is clearly indicated.