ChatGPT: Chances and Challenges for Dentistry
Florin Eggmann, Dr. med. dent.; and Markus B. Blatz, DMD, PhD
The artificial intelligence (AI) chatbot ChatGPT has generated both huge interest and deep concern since its launch in November 2022.1 ChatGPT, a large language model (LLM) with a conversational interface, has been trained on vast amounts of human-generated text and has the ability to respond to questions and complete various text-related tasks. The use of ChatGPT and similar LLMs in dentistry is unlikely to significantly impact the daily routine of most dental healthcare personnel but could streamline administrative workflows and potentially serve as an additional tool for clinical decision support in the future. However, this is contingent on the availability of comprehensive, up-to-date, and unbiased data. The use of LLMs also causes privacy and cybersecurity concerns. It is therefore crucial to implement robust data protection measures and strong defenses against malicious use of LLMs. Although ChatGPT provides succinct answers to most queries, its lack of reliability, transparency, and up-to-date knowledge compared with conventional search engines is a major drawback, particularly for health-related queries.
This article provides an overview of the broad and complex implications that LLMs carry for dentistry. It emphasizes the need for rigorous scientific evaluation and strict legal regulation to realize the potential benefits of LLMs for patient care, education at dental schools, and dental research and to safeguard against the serious risks inherent in such AI technologies.
Answering Life's Questions
Within just 3 months of its launch last fall, ChatGPT had amassed 100 million users, making it one of the fastest-growing internet applications in history.2 Developed by OpenAI, a US company backed by Microsoft, ChatGPT can give answers to the biggest and smallest questions in life, draft passable college essays, and produce job application letters, short stories, sonnets, song lyrics, stir-fry recipes, and more. It is also adept at finding and fixing bugs in computer code.3
The ChatGPT chatbot can fulfill such text-related tasks because it has been trained on massive amounts of human-generated text, with detailed guidance from AI experts and instructive interactions with volunteers during its development.4 When replying to a question or prompt, ChatGPT generates statistically likely sequences of words, using statistical patterns learned from vast amounts of textual data. In some sense, this is like predictive text on a cell phone but on a much larger scale and with advanced sophistication.
Consider this example: If you ask ChatGPT, "How long should I etch lithium-disilicate restorations with hydrofluoric acid?" it replies, "20 to 30 seconds." It provides this answer by analyzing the statistical distribution of linguistic tokens in a huge corpus of English text and determining that "20 to 30 seconds" is the most probable response. ChatGPT has no actual knowledge of dentistry, practitioners will be glad to learn.
LLMs such as ChatGPT possess extraordinary capabilities and hold considerable potential for use in dentistry. The following paragraphs offer a cursory survey of the implications of LLMs for dentistry, including their current limitations and drawbacks.
Everyday Working Life of Dental Healthcare Personnel
Given that the delivery of dental care heavily relies on in-person communication, clinical and radiographic assessments, and hands-on procedures, LLMs are unlikely to bring significant changes to the daily routine of most dentists, dental assistants, and hygienists. LLMs could, however, change the way dental telemedicine services operate and alter the work of administrative personnel, such as office employees in dental clinics and insurance claims processors.
Routine Administrative Tasks
LLMs hold promise for streamlining administrative work. In a recent TikTok video, Dr. Clifford Stermer, a rheumatologist based in the United States, demonstrated how the use of ChatGPT can improve the efficiency of writing preauthorization requests for insurance companies.5 This highlights the potential of LLMs in assisting healthcare personnel to effectively perform routine administrative duties, such as maintaining patient records and submitting insurance reimbursement claims. Some LLMs are also adept at translating languages, making them useful tools for facilitating communication across multiple languages.
Clinical Decision Support
LLMs have the potential to become an additional tool for clinical decision support. For instance, LLMs that also take into account patients' electronic health record data may soon be used to enhance evidence-based selection of radiological imaging exams.5 However, the accuracy of LLMs depends on the quality and type of data used in their training, and their usefulness as a clinical support tool is contingent on access to comprehensive, up-to-date, and unbiased data.6 ChatGPT, with its limited knowledge base and lack of internet access, is liable to occasionally produce inaccurate, biased, misleading, or harmful content. It is therefore unsuitable for clinical support today but offers a glimpse into the future of the potential role of LLMs in clinical practice.
Data Confidentiality and Cybersecurity
The use of LLMs in dentistry raises privacy and security concerns as they may gather personal and medical information. To mitigate these risks, it is crucial to put in place robust data protection measures. The terms of use for ChatGPT allow OpenAI to collect usage data. To ensure patient data privacy, one must therefore enter no patient information into ChatGPT. LLMs also involve dangers for cybersecurity because threat actors may exploit the capabilities of LLMs to create malware and craft phishing messages.7 The entities involved in building LLMs, including technology companies, research organizations, and governmental agencies, must therefore take proactive measures to prevent malicious use. Additionally, to protect patient data and healthcare providers' technology systems, it is critical to implement and maintain strong defenses against cyberattacks.8,9
Online Information on Dental Health
With its user-friendly dialogue-based interface, ChatGPT provides clear and succinct answers to most queries without the need to navigate to other websites. LLMs such as ChatGPT are therefore predicted to be increasingly used for online queries instead of conventional search engines, such as Baidu, Bing, Google, and Yahoo. However, the accuracy of information provided by ChatGPT is a major concern, particularly for health-related queries. Shortcomings of ChatGPT are that it can provide factually incorrect or biased answers with confidence, its knowledge base ends in 2021, and it cannot access the internet. Moreover, it is all but impossible to verify the sources of its replies. This lack of reliability, transparency, and up-to-date knowledge is a major drawback compared with conventional search engines. Considering that many people seek medical information online, it is vital to safeguard against inaccurate, outdated, and biased responses from LLMs to health-related queries.10
Education at Dental Schools
LLMs present both opportunities and challenges for higher education. It is necessary to adapt some higher education curricula, especially in academic disciplines that strongly rely on written assignments.11-13 While dental schools commonly evaluate students through oral exams, multiple-choice tests, practical assessments, and supervised patient treatments, and are relying less on written assignments, it is important to incorporate education on AI technologies into the curriculum to develop students' understanding of AI applications relevant to dentistry.14
Scientific Writing
Authors can garner assistance from LLMs to improve the fluency, grammar, and spelling of their writing, which potentially can promote a level playing field for non-native English speakers looking to publish in English-language journals.15 However, there are serious concerns that the use of LLMs may result in an increase in flawed, faulty, and fabricated research.16-18 A recent study found that scientific abstracts generated by ChatGPT evaded plagiarism checks and often deceived human reviewers.19 This highlights the central importance of scientific integrity, methodological rigor, and thorough peer review in all academic fields, including dental research.20
Conclusions
LLMs such as ChatGPT have both potential benefits and potential risks. In dentistry, LLMs hold promise for improving administrative efficiency and facilitating multilingual communication. In the near future, LLMs may become an additional tool for clinical decision support. However, their usefulness as an aid in dentistry is contingent on the quality and comprehensiveness of data used in training and during user interactions. The use of LLMs in healthcare fields such as dentistry raises privacy and cybersecurity concerns. Effective measures must be taken to mitigate these risks and prevent malicious use. Although ChatGPT offers a user-friendly interface for answering queries, it is currently unsuitable to reliably answer health-related questions and prompts due to its lack of accuracy, transparency, and up-to-date knowledge.
The implications of LLMs for dentistry are far-reaching and complex. Therefore, their use must be approached with caution to ensure the veracity of information, patient data privacy, and the security of technology systems. Both rigorous scientific evaluation and careful legal regulation are necessary to harness the potential benefits of LLMs for patient care, education, and dental research and to guard against the risks inherent in such AI technologies.
Acknowledgment
The authors recognize, with gratitude, the support of the Gottfried and Julia Bangerter-Rhyner Foundation and the Freiwillige Akademische Gesellschaft Basel (FAG), funders of scholarships for Florin Eggmann. The authors disclose the use of ChatGPT (January 30 version [free research preview]) during the revision of the draft of this report to enhance its fluency. An online repository holds the version of the draft that was revised using ChatGPT.21 ChatGPT was not used for conceptualization, literature review, data interpretation, or drawing conclusions.
Disclosure
The authors report no competing financial and/or nonfinancial interests in relation to this article.
About the Authors
Florin Eggmann, Dr. med. dent.
Research Fellow, Department of Preventive and Restorative Sciences, Robert Schattner Center, University of Pennsylvania School of Dental Medicine, Philadelphia, Pennsylvania; Lecturer, Department of Periodontology, Endodontology, and Cariology, University Center for Dental Medicine Basel UZB, University of Basel, Basel, Switzerland
Markus B. Blatz, DMD, PhD
Professor of Restorative Dentistry, Chair, Department of Preventive and Restorative Sciences, Robert Schattner Center, and Assistant Dean, Digital Innovation and Professional Development, University of Pennsylvania School of Dental Medicine, Philadelphia, Pennsylvania
References
1. Stokel-Walker C, van Noorden R. What ChatGPT and generative AI mean for science. Nature. 2023;614(7947):214-216.
2. Milmo D. ChatGPT reaches 100 million users two months after launch. The Guardian website. February 2, 2023. https://www.theguardian.com/technology/2023/feb/02/chatgpt-100-million-users-open-ai-fastest-growing-app. Accessed February 24, 2023.
3. Castelvecchi D. Are ChatGPT and AlphaCode going to replace programmers? Nature. 2022 Dec 8. doi: 10.1038/d41586-022-04383-z.
4. Shanahan M. Talking about large language models. Cornell University website. Revised February 16, 2023. https://arxiv.org/pdf/2212.03551. Accessed February 24, 2023.
5. Shen Y, Heacock L, Elias J, et al. ChatGPT and other large language models are double-edged swords. Radiology. 2023;230163. doi: 10.1148/radiol.230163.
6. Patcas R, Bornstein MM, Schätzle MA, Timofte R. Artificial intelligence in medico-dental diagnostics of the face: a narrative review of opportunities and challenges. Clin Oral Investig. 2022;26(12):6871-6879.
7. Kulesh S. Why ChatGPT can be dangerous for every internet user. The Times of India website. December 21, 2022. https://timesofindia.indiatimes.com/gadgets-news/why-chatgpt-can-be-dangerous-to-every-internet-user/articleshow/96393104.cms. Accessed February 24, 2023.
8. Iacono L, Wojcieszek K, Glass G. Q2 2022 threat landscape: ransomware returns, healthcare hit. Kroll website. August 2022. https://www.kroll.com/-/media/kroll-images/pdfs/q2-2022-threat-landscape-report.pdf. Accessed February 24, 2023.
9. Singleton C, DeBeck C, Chung J, et al. X-force threat intelligence index 2022. IBM website. February 2022. https://www.ibm.com/downloads/cas/ADLMYLAZ. Accessed February 24, 2023.
10. Masukume G, Kipersztok L, Das D, et al. Medical journals and Wikipedia: a global health matter. Lancet Glob Health. 2016;4(11):e791.
11. Graham F. Daily briefing: will ChatGPT kill the essay assignment? Nature. 2022 Dec 12. doi: 10.1038/d41586-022-04437-2.
12. Stokel-Walker C. AI bot ChatGPT writes smart essays - should professors worry? Nature. 2022 Dec 9. doi: 10.1038/d41586-022-04397-7.
13. Taylor P. On ChatGPT. London Review of Books. January 5, 2023;45(1). https://www.lrb.co.uk/the-paper/v45/n01/paul-taylor/on-chatgpt. Accessed February 24, 2023.
14. Schwendicke F, Chaurasia A, Wiegand T, et al. Artificial intelligence for oral and dental healthcare: core education curriculum. J Dent. 2023;128:104363. doi: 10.1016/j.jdent.2022.104363.
15. Kitamura FC. ChatGPT is shaping the future of medical writing
but still requires human judgment. Radiology. 2023;230171. doi: 10.1148/
radiol.230171.
16. Else H. Abstracts written by ChatGPT fool scientists. Nature. 2023;613
(7944):423. doi: 10.1038/d41586-023-00056-7.
17. Gordijn B, Ten Have H. ChatGPT: evolution or revolution? Med Health Care Philos. 2023 Jan 19. doi: 10.1007/s11019-023-10136-0.
18. Thorp HH. ChatGPT is fun, but not an author. Science. 2023;379
(6630):313. doi: 10.1126/science.adg7879.
19. Gao CA, Howard FM, Markov NS, et al. Comparing scientific abstracts generated by ChatGPT to original abstracts using an artificial intelligence output detector, plagiarism detector, and blinded human reviewers [preprint]. bioRxiv. December 27, 2022. https://www.biorxiv.org/content/10.1101/2022.12.23.521610v1. Accessed February 24, 2023.
20. Tools such as ChatGPT threaten transparent science; here are our ground rules for their use. Nature. 2023;613(7945):612. doi: 10.1038/d41586-023-00191-1.
21. Eggmann F, Blatz MB. ChatGPT: chances and challenges for dentistry. 2023. Zenodo. https://doi.org/10.5281/zenodo.7617123.
