Menu

Chatbots and Natural Language Processing: Innovation Brief

ChatGPT has the potential to revolutionize the field of anesthesiology, improving patient care and outcomes while streamlining workflow and reducing costs.

In late 2022, we saw the seemingly overnight explosion of innovative new computer programs able to engage in conversation with human users with a degree of sophistication that became increasingly difficult to discern from true human interaction. These conversational artificial intelligence programs, colloquially termed “chatbots,” seemed to take a major step forward due to recent advances in natural language processing (NLP). NLP is a subset of artificial intelligence focused on recognition of unstructured text (natural language) with subsequent analysis of available texts to allow for such varied tasks as translation, grammar analysis, parsing of meaning, and, ultimately, conversational output with external users. Within this domain, the recent innovations seen are due to the use of large language models (LLM) where algorithms are developed that can, with minimal training, analyze massive data sets (e.g., on the order of a petabyte, or just over one million gigabytes1and gain new skills and knowledge without explicit training in those domains.2

The first, and by far the most well-known, chatbot of this new generation is ChatGPT, developed by OpenAI, a San Francisco-based startup founded in 2015 by a team of veteran Silicon Valley entrepreneurs including Sam Altman, Peter Thiel, and Elon Musk.3 At its founding it touted itself as a non-profit research company “free from financial obligations” so that it can “better focus on a positive human impact”.3 Within two months of the launch of ChatGPT in late 2022, it amassed 100 million users, faster even than popular consumer apps like TikTok and Instagram.4

When considering the potential implications for such innovative new technologies on the field of anesthesiology, it seemed proper to ask ChatGPT to consider its own potential to contribute to the specialty. The following text was written completely by ChatGPT5 in a matter of seconds without any further human interaction or editing beyond the following prompt:

“Write a 1000-word essay on what ChatGPT is and innovative ways it could be used in the practice of anesthesiology.” (Figure 1 below).

ChatGPT is a conversational AI model developed by OpenAI, that is capable of understanding natural language and providing answers to text-based queries. The model was trained on a diverse range of topics, allowing it to answer questions about a wide range of subjects including science, history, and current events. ChatGPT is a powerful tool for businesses and organizations to automate customer service and support, freeing up employees to focus on more complex tasks.

In the field of anesthesiology, ChatGPT has the potential to be used in a number of innovative ways that can improve patient care and efficiency. In this essay, we will explore some of the innovative ways ChatGPT could be used in anesthesiology, including:

  1. Clinical Decision Support: Anesthesiologists are responsible for managing patient pain, airway control, fluid and electrolyte balance, and blood pressure during surgical procedures. ChatGPT can be used to provide real-time decision support to anesthesiologists by answering questions about medication doses, side effects, and drug interactions. This can help anesthesiologists make informed decisions quickly and with confidence, improving patient outcomes and reducing the risk of adverse events.
  2. Documentation and Record Keeping: Documentation is a crucial aspect of anesthesiology, as it helps to ensure that patient records are complete, accurate, and accessible. ChatGPT can be integrated with electronic health records to automate the documentation process, reducing the time and effort required to keep records up to date. This can improve the accuracy of patient records and make it easier for anesthesiologists to access patient information when needed.
  3. Patient Education: Anesthesiology is a complex and technical field, and patients may have questions about their care and treatment. ChatGPT can be used to provide patients with information about anesthesiology procedures, helping to ease their fears and promote informed decision-making. This can improve patient satisfaction and trust in the anesthesiology team, helping to build stronger patient-provider relationships.
  4. Research and Data Analysis: Anesthesiology is a rapidly evolving field, and new treatments and procedures are constantly being developed and tested. ChatGPT can be used to analyze large amounts of data related to anesthesiology, allowing researchers to quickly identify trends and patterns. This can help to improve our understanding of the field, leading to the development of more effective treatments and procedures.
  5. Streamlining Workflow: Anesthesiology is a complex and demanding field, requiring anesthesiologists to manage many tasks simultaneously. ChatGPT can be used to automate routine tasks, freeing up anesthesiologists to focus on more complex and high-value tasks. This can improve efficiency, reducing wait times and improving patient outcomes.

In conclusion, ChatGPT has the potential to revolutionize the field of anesthesiology, improving patient care and outcomes while streamlining workflow and reducing costs. By using ChatGPT to provide decision support, automate documentation, educate patients, analyze data, and streamline workflow, anesthesiologists can work more efficiently and effectively, delivering better care to their patients.

It is evident that ChatGPT has impressive natural language processing abilities. It has been shown to have broad capacity across multiple knowledge domains as evidenced by passing examinations in business6, law7 , and medicine.8 One editorial in Nature9 published just after ChatGPT’s debut notes that in addition to this, ChatGPT has been included as a co-author on several manuscripts as it was used to create significant portions of the final manuscript. This raises the ethical question as to whether a computer algorithm should be able to be attributed as an author. The editorial concludes that any artificial intelligence generative algorithm should be cited in the methods or acknowledgement section as authorship requires an agreement to be responsible and accountable for the content of a scholarly publication, a standard which ChatGPT, or any LLM, cannot meet. Additionally, given that ChatGPT, like all LLMs, are trained on large volumes of data, it raises questions about the very nature of whether the output of such programs represent novel generation of ideas or simply derivative work of the training data. This very question is at the heart of several lawsuits by prominent authors against Open AI stating that the unauthorized use of copyrighted material as training data is, in fact, copyright infringement and entitles the authors to compensation.10,11 As with any disruptive innovation, the old rules must be reconsidered for the new world to decide if, when, and how they apply. We are still early in this process and a recalibration is imminent as to the question of authorship and ownership for the output of LLMs.

Despite its start as a non-profit and stated desire to “build value for everyone rather than shareholders,”3 in early 2023, OpenAI accepted $10 billion from Microsoft to acquire an eventual 49% stake in the company.12 In early February 2023, Microsoft, based on its newly acquired stake in OpenAI, conducted a limited launch of the updated version of its search engine, Bing, featuring a conversational, chat-based search function powered by an integration with ChatGPT.13 Stories soon emerged of disturbing “behaviors” by the new search engine, which called itself Sydney, and left journalists “deeply unsettled, even frightened.”14 In one interaction, it became argumentative with a user1514 These widely publicized interactions led Microsoft to conclude that very long chat conversations could “confuse” Sydney and were the ultimate cause of the bizarre interactions.16 Microsoft ultimately curtailed the number and duration of chats that users could engage in,17 leading some users to lament that Microsoft had “lobotomized” Bing.18 Over subsequent months, more rigorous studies showed a sharp drop off in performance through the summer of 2023.19 This has led to the announcement of planned updates in late 2023 including image analysis, speech recognition, and text-to-speech responses.20 Additionally, while the initial versions of ChatGPT were trained using data dating back to September 2021, which halted its knowledge base at that date, upcoming versions will allow access to the Internet in real time.21 These updates seem to have overshadowed any early missteps as recent valuations of the technology soared to $90 billion.22

Given the potential size of this market it should be of no surprise that OpenAI is not the only LLM game in town. In February 2023, Google also announced its own chat-based search, named Bard, to compete against Bing. Unfortunately, an early demonstration of the service showed that Bard made a simple factual error in answering a question, resulting in a 9% drop in the stock price for Google’s parent company, Alphabet.23 The Chinese search and AI company, Baidu, released its own chatbot named Ernie Bot in March 2023 with subsequent integration across its entire suite of services, including search, cloud services, automobile operating systems, and smart speakers.24 Though Baidu has marketed internal research25 touting Ernie Bot as superior to ChatGPT, each of the LLMs available in the marketplace is still so new that direct comparisons for real world use cases are not widely published.

In our (human) conclusion, subsequent iterations of LLM algorithms may have some innovative uses in the future practice of anesthesiology, as ChatGPT outlines above in a surprisingly eloquent fashion. Such algorithms may at some point prove to be a beneficial tool to provide a cognitive “boost” to clinicians allowing better patient care.26 Even so, these algorithms will remain just that: tools. As valuable as they may or may not prove to be, we cannot relinquish our role in the proper development and testing of such algorithms and as the final arbiter of their appropriate use even as they become increasingly sophisticated and difficult to differentiate from human interaction.

Figure 1.

Screen capture of the ChatGPT interface.

References:

1. Fisher T. Terabytes, gigabytes, & petabytes: How big are they? Lifewire. January 1, 2021. Accessed February 15, 2024. https://www.lifewire.com/terabytes-gigabytes-amp-petabytes-how-big-are-they-4125169#toc- how-big-is-a-petabyte.

2. Wiggers K. The emerging types of language models and why they matter. TechCrunch. April 28, 2022. Accessed February 15, 2024. https://techcrunch.com/2022/04/28/the-emerging-types-of-language-models-and-why-they-matter/.

3. Introducing openai. Introducing OpenAI. Accessed February 15, 2024. https://openai.com/blog/introducing-openai/.

4. CHATGPT reaches 100 million users two months after launch. The Guardian. February 2, 2023. Accessed February 15, 2024. https://www.theguardian.com/technology/2023/feb/02/chatgpt-100-million-users-open-ai-fastest-growing-app.

5. Introducing chatgpt. Introducing ChatGPT. Accessed February 15, 2024. https://openai.com/blog/chatgpt.

6. Would chat GPT3 get a Wharton MBA? Accessed February 15, 2024. https://mackinstitute.wharton.upenn.edu/wp-content/uploads/2023/01/Christian-Terwiesch-Chat-GTP.pdf.

7. Choi JH, Hickman KE, Monahan A, Schwarcz D. Chatgpt goes to law school. SSRN. January 25, 2023. Accessed February 15, 2024. https://ssrn.com/abstract=4335905.

8. Kung TH, Cheatham M, ChatGPT, et al. Performance of chatgpt on USMLE: Potential for AI-assisted medical education using large language models. medRxiv. January 1, 2022. Accessed February 15, 2024. https://www.medrxiv.org/content/10.1101/2022.12.19.22283643v2.

9. Stokel-Walker C. ChatGPT listed as author on research papers: many scientists disapprove. Nature. 2023 Jan;613(7945):620-621. PMID: 36653617

10. Roth E. Another group of writers is suing openai over copyright claims. The Verge. September 11, 2023. Accessed February 15, 2024. https://www.theverge.com/2023/9/11/23869145/writers-sue-openai-chatgpt-copyright-claims.

11. Alter A, Harris EA. Franzen, Grisham and other prominent authors Sue OpenAI. The New York Times. September 20, 2023. Accessed February 15, 2024. https://www.nytimes.com/2023/09/20/books/authors-openai-lawsuit-chatgpt-copyright.html.

12. Hoffman L, Albergotti R. Microsoft eyes $10 billion bet on CHATGPT. Semafor. January 10, 2023. Accessed February 15, 2024. https://www.semafor.com/article/01/09/2023/microsoft-eyes-10-billion-bet-on-chatgpt.

13. Mehdi Y. Reinventing search with a new AI-powered Microsoft Bing and EDGE, your copilot for the web. The Official Microsoft Blog. May 16, 2023. Accessed February 15, 2024. https://blogs.microsoft.com/blog/2023/02/07/reinventing-search-with-a-new-ai-powered-microsoft-bing-and-edge-your-copilot-for-the-web/.

14. Roose K. Bing’s A.I. Chat: “I want to be alive. 😈.” The New York Times. February 16, 2023. Accessed February 15, 2024. https://www.nytimes.com/2023/02/16/technology/bing-chatbot-transcript.html.

15. Uleis J. My new favorite thing - bing’s new CHATGPT BOT argues with a user, gaslights them about the current year being 2022, says their phone might have a virus, and says “You have not been a good user”why? because the person asked where avatar 2 is showing nearby pic.twitter.com/x32vopxxqg. Twitter. February 13, 2023. Accessed February 15, 2024. https://twitter.com/MovingToTheSun/status/1625156575202537474?s=20&t=P_FgyBC1edOTj7+um1GoVRQ.

16. Bing. Accessed February 15, 2024. https://blogs.bing.com/search/february-2023/The-new-Bing-Edge-Learning-from-our-first-week.

17. Bing. Accessed February 15, 2024. https://blogs.bing.com/search/february-2023/The-new-Bing-Edge-%E2%80%93-Updates-to-Chat.

18. Benj Edwards    -    Feb 17 2023 11:11 pm UTC. Microsoft “lobotomized” AI-powered bing chat, and its fans aren’t happy. Ars Technica. February 17, 2023. Accessed February 15, 2024. https://arstechnica.com/information-technology/2023/02/microsoft-lobotomized-ai-powered- bing-chat-and-its-fans-arent-happy/.

19. ArXiv.org e-print archive. arXiv.org e-Print archive. Accessed February 15, 2024. https://arxiv.org/pdf/2307.09009.pdf.

20. Roose K. The new CHATGPT can “see” and “talk.” here’s what it’s like. The New York Times. September 27, 2023. Accessed February 15, 2024. https://www.nytimes.com/2023/09/27/technology/new-chatgpt-can-see-hear.html.

21. WSJ News exclusive | openai seeks new valuation of up to $90 billion ... Accessed February 15, 2024. https://www.wsj.com/tech/ai/openai-seeks-new-valuation-of-up-to-90-billion-in-sale-of-existing-shares-ed6229e0.

22. Antoinette Radford & Zoe Kleinman TE. CHATGPT can now access up to date information. BBC News. September 27, 2023. Accessed February 15, 2024. https://www.bbc.com/news/technology-66940771.

23. Olson E. Google shares drop $100 billion after its new AI chatbot makes a mistake. NPR. February 9, 2023. Accessed February 15, 2024. https://www.npr.org/2023/02/09/1155650909/google-chatbot--error-bard-shares.

24. Kharpal A. China’s Baidu claims its Ernie Bot beats chatgpt on key tests as A.I. race heats up. CNBC. June 27, 2023. Accessed February 15, 2024. http://www.cnbc.com/2023/06/27/baidu-claims-its-ernie-bot-beats-openais-chatgpt-on-key-ai-tests.html.

25. Baidu Research. Accessed February 15, 2024. http://research.baidu.com/Blog/index-view?id=185.

26. Alexander JC, Joshi GP. Anesthesiology, automation, and artificial intelligence. Proc (Bayl Univ Med Cent). 2017 Dec 5;31(1):117-119. PMID: 29686578

Curated by: ASA Committee on Innovation

Date of last update: May 17, 2024