As artificial intelligence (AI) systems have become more widely available, scientific publishers are expressing concern with their use in both manuscript development and peer review.1–3 As publishers begin to establish guidelines for producing accurate and trustworthy manuscripts in the age of AI,4 issues of copyright and intellectual property must be considered: any data uploaded into a public AI are ingested and "remembered" for others to potentially use as their own. AI "hallucinations" of false information can also create significant verification work for data and references.5
When approached by Dr. Winkler-Schwartz5 with the idea of creating a controlled experiment to utilize the OpenAI GPT-4 algorithm without risking copyright issues or hallucinations, I was intrigued; such an exercise could potentially prove useful to our readers.
The Journal of Neurosurgery Publishing Group (JNSPG) in partnership with Dr. Winkler-Schwartz has set up an AI "conversational manuscript" interface that allows you, our readers, to interact with the information in a specific article6 published in our June issue of Neurosurgical Focus. The interface will respond only with information contained within the paper; unrelated questions such as "what is the best brownie recipe?" will receive the answer:
"I’m sorry, but as a neurosurgery-focused AI assistant, I don’t have information on brownie recipes. My primary function is to provide information and answer questions related to neurosurgery and related medical topics."
When you visit the manuscript link, you will see an interface embedded at the top of the article (Fig. 1). Please contribute data to this experiment: read the article, ask relevant questions within the interface, and give us feedback on whether this is a useful service for JNSPG to offer its readers. You will also find a link to a short anonymous survey; we hope you will take time to share your thoughts.
Screenshot of article with conversational manuscript interface (red circle). Figure is available in color online only.
This experiment will run through October 1, 2023, after which the functionality will be removed. Thank you in advance for helping JNSPG test drive this technology as we consider potential use cases and guidelines for our journals.
We look forward to your responses.
Disclosures
The author reports no conflict of interest.
References
- 1↑
The use of AI and AI-assisted writing technologies in scientific writing. Elsevier. Accessed July 14, 2023. https://www.elsevier.com/about/policies/publishing-ethics/the-use-of-ai-and-ai-assisted-writing-technologies-in-scientific-writing
- 2
Donker T. The dangers of using large language models for peer review. Lancet Infect Dis. 2023;23(7):781.
- 3↑
Conroy G. Scientists used ChatGPT to generate an entire paper from scratch - but is it any good?. Nature. Published online July 7, 2023. doi: 10.1038/d41586-023-02218-z
- 4↑
Editorial. Tools such as ChatGPT threaten transparent science; here are our ground rules for their use. Nature. 2023;613(7945):612.
- 5↑
Winkler-Schwartz A. When words leap off the page: conversational manuscripts as a novel interface for scientific publishing. J Neurosurg. Published online August 1, 2023. doi: 10.3171/2023.7.JNS231617
- 6↑
Shahrestani S, Chan AK, Bisson EF, et al. Developing nonlinear k-nearest neighbors classification algorithms to identify patients at high risk of increased length of hospital stay following spine surgery. Neurosurg Focus. 2023;54(6):E7.
