Since the end of last year (Dec 2022), there has been lots of buzz about a new artificial intelligence (AI)-chatbot called ChatGPT, which uses natural language processing (NLP) to generate conversations. Its release has since gone viral, and many people are using this new generative AI to do all sorts of things – like coding for beautiful logo images, creating three-dimensional (3D) images, creating music and composing songs and their lyrics, and even developing lecture notes and writing books!
So what is the hype about ChatGPT? As a chatbot, it is able to have human-like conversations with users. It is different from search engines as it is not able to search the internet for information, but rather, it learns from the tons of training data that has been “fed” to it to generate a response. Interestingly, it was also described to be “woke” because it was able to differentiate users’ requests and refuse to answer some queries on grounds of being “inappropriate” and “harmful”. Best of all, it is currently free for public use, since it is in its research phase right now! Developed by OpenAI – the company behind the famous DALLE•2 AI art generator and its automatic Whisper speech recognition system – this chatbot piqued the interest of the world and was even described by Elon Musk as a “scary good” and “dangerously strong AI”!
If you want to know more about the capabilities of this amazing chatbot, take a crash course by Adrian Twarog on his YouTube video:
What would ChatGPT write about it’s own role in healthcare?
Following Reddit user Jackb1980’s experiment on asking the chatbot to write a textbook for him, I was also interested in knowing what the chatbot would think of itself and it’s role in healthcare. Through a short conversation, the following excerpt is what ChatGPT replied!
Check out the section below to read its “thoughts” on its role in healthcare… (Credit to ChatGPT!)
| 1. Introduction Brief overview of ChatGPT and its capabilities ChatGPT, short for “Generative Pre-trained Transformer,” is a large language model developed by OpenAI. It is trained on a massive amount of text data from the internet and is able to generate human-like text in a variety of styles and formats. How ChatGPT is being used in healthcare The focus of this blog post is to explore how ChatGPT is being used in the healthcare industry. Healthcare is a field that heavily relies on communication between patients, doctors and other healthcare professionals, and ChatGPT’s ability to understand and generate natural language makes it a valuable tool in this field. The post will explore various applications of ChatGPT in healthcare, the benefits and challenges of using it, as well as a comparison to other AI language models and NLP techniques in healthcare. The post will also look into the future possibilities of ChatGPT in healthcare. 2. Applications of ChatGPT in Healthcare Triage chatbots One of the most popular applications of ChatGPT in healthcare is the development of triage chatbots. These chatbots use natural language processing (NLP) and machine learning (ML) to understand the symptoms and concerns of patients and provide initial guidance on how to proceed with their healthcare. They can also help to reduce the workload of doctors and other healthcare professionals by triaging patients based on their symptoms and directing them to the appropriate level of care. Medical transcription ChatGPT can also be used to transcribe medical dictations, which can save time and improve accuracy. This can be particularly useful in fields such as radiology, where doctors need to document the findings of imaging tests quickly and accurately. Virtual assistants for patients ChatGPT can be used to create virtual assistants for patients. These assistants can help patients to schedule appointments, provide information about their health, and even remind them to take their medication. This can be particularly useful for patients who have difficulty communicating with healthcare professionals, such as those with hearing or speech impairments. Other examples ChatGPT can also be used in other healthcare applications such as summarizing clinical notes, automating the generation of clinical documents, and natural language understanding of electronic health records. 3. Benefits of using ChatGPT in Healthcare Improved efficiency ChatGPT can help to automate repetitive tasks such as data entry, transcription, and triage, which can save time and increase efficiency for healthcare professionals. This can lead to faster and more accurate diagnoses, as well as more time for healthcare professionals to spend with patients. Cost savings Automating tasks with ChatGPT can also lead to cost savings for healthcare organizations. For example, using a chatbot for triage can reduce the need for expensive human operators. Additionally, using ChatGPT for tasks such as medical transcription can save money by reducing the need for human transcribers. Better patient outcomes ChatGPT can also improve patient outcomes by providing more accurate and prompt diagnoses, as well as by providing patients with more convenient and personalized healthcare. For example, a virtual assistant that uses ChatGPT can help patients to manage their health more effectively, which can lead to better outcomes. 4. Challenges of using ChatGPT in Healthcare Need for large amounts of training data One of the main challenges of using ChatGPT in healthcare is the need for large amounts of training data. Because healthcare is a complex and dynamic field, it is difficult to train a model on the vast amount of data that would be required to accurately understand and generate medical text. This can make it difficult to achieve the level of accuracy and reliability that is necessary for healthcare applications. Ethical considerations Another challenge of using ChatGPT in healthcare is the ethical considerations that come with using AI in this field. For example, there are concerns about the privacy and security of patients’ personal and medical information, as well as the potential for bias in the model’s predictions. It is important to ensure that the model is developed and used in an ethical manner, and that it is transparent and explainable. 5. Comparison to other AI language models and NLP techniques in healthcare ChatGPT is not the only language model that is being used in healthcare. Other popular models include BERT and GPT-2. These models also use pre-training and fine-tuning to understand and generate text. However, ChatGPT has some advantages over these models, such as its ability to generate more human-like text and its ability to handle longer input. In addition to language models, there are other NLP techniques that are being used in healthcare, such as rule-based systems, and machine learning-based systems. Rule-based systems use a set of pre-defined rules to understand and generate text, while machine learning-based systems use a training data to learn how to understand and generate text. Rule-based systems are generally simpler and more transparent than machine learning-based systems but may lack the flexibility and performance of the latter. 6. Future possibilities for ChatGPT in healthcare Advanced chatbots One potential future development for ChatGPT in healthcare is the creation of more advanced chatbots. These chatbots could use more sophisticated natural language understanding and generation techniques, such as dialogue management and context tracking, to provide more accurate and helpful responses to patients. They could also be integrated with other healthcare technologies, such as electronic health records, to provide more personalized and efficient care. Integration with other healthcare technologies Another potential future development is the integration of ChatGPT with other healthcare technologies. For example, ChatGPT could be used to analyze and understand the vast amounts of data generated by wearable devices and remote monitoring devices, to provide more accurate and personalized care. It could also be used to assist with the interpretation of medical images and radiology, to support diagnosis and treatment planning. 7. Conclusion Summary of key points In this blog post, we have discussed how ChatGPT, a large language model developed by OpenAI, is being used in the healthcare industry. We have explored the various applications of ChatGPT in healthcare, including triage chatbots, medical transcription, and virtual assistants for patients. We have also looked at the benefits and challenges of using ChatGPT in healthcare, such as improved efficiency, cost savings, and better patient outcomes, as well as the need for large amounts of training data, and ethical considerations. Final thoughts on the potential of ChatGPT in healthcare ChatGPT has the potential to revolutionize the way healthcare is delivered by providing more accurate, efficient, and personalized care. By automating repetitive tasks and providing more convenient and accessible healthcare, ChatGPT can help to improve patient outcomes and reduce costs. However, it is important to ensure that the model is developed and used in an ethical manner, and that it is transparent and explainable. As technology advances, we expect to see more applications of ChatGPT in healthcare and more developments in this field. |
My experiment on ChatGPT on COVID-19 vaccine information
Since ChatGPT is still “learning the ways of the world”, as a healthcare professional, I echo the concerns in healthcare regarding the ethical use and quality of information provided by this AI chatbot, in terms of its comprehensiveness, accuracy and reliability. So I decided to try asking some queries on COVID-19 vaccines, using the “gold standard” answers that our team compiled in our publication on the Quality of COVID-19 Vaccine Videos on Video-sharing Platforms. While the answers provided by ChatGPT are quite comprehensive and accurate (see Comparison Figure below), with some of its answers incorporating more information than the “gold standard” answer, it may not be clear to the layman as to whether its answers are inferred or evidenced-based. While there was already some research published on the GPT-3 architecture that this AI chatbot is based on, I would definitely expect more research being published on the capabilities and applications of ChatGPT this new year. Nonetheless, the answers that it provides are still quite impressive considering that its algorithm is still learning, and it may ultimately improve the quality of other chatbots, mobile health/mHealth apps and digital voice assistants like Siri, Alexa, Bixby and Google Assistant on health-related information, such as for surgical preparations; medical conditions like eating disorders and mental health conditions; as well as public health concerns on pandemics like COVID-19 and COVID-19 vaccines.
An impressive, yet scary AI chatbot, ain’t it?
The cool thing about ChatGPT is that, unlike other chatbots or digital voice assistants, it is able to generate answers based on the conversation it has with the user – something like a WhatsApp conversation where it remembers what was posed to it before in the conversation. In fact, there are already concerns regarding its potential use by students for cheating, the possible spread of misinformation through its answers, and its potential to replace some jobs. Given the fact that it is still a young “baby” chatbot, its capabilities are quite impressive compared to its predecessors, I must say! However, being this young and experimental, there are also many limitations tagged to it.
Limitations of ChatGPT
Having tried ChatGPT myself, I found several limitations to it. My questions or statements had to be worded in a specific way, with as little spelling and grammatical errors as possible, or it would not be able to understand my question. Due to the high traffic flow on its website, there were times whereby I had to refresh my webpage as the conversation got “stuck” with error messages appearing. Furthermore, the accuracy or lack thereof, and biasness of the chatbot were also noted by other users. For example, if one were to force an opinion to a query in the conversation, the chatbot might also accept that “answer” as the correct answer. Other limitations include its dependence on the quality and quantity of data being used to train its algorithm, its limited understanding and reasoning abilities, potential bias in its answers, as well as technical errors and mistakes.
Summary thoughts
Despite the shortcomings of ChatGPT, the field of AI is expected to exponentially accelerate in the coming years post-COVID. In fact, Gartner has predicted that even though technologies like AI and machine learning, and digital twins are still at the technology trigger phase, they will become mature enough to be fit for mainstream adoption in 5 to 10 years time. As such, even though it sounds scary, it is only a matter of time where AI will be part of healthcare practices and education, hence as a digital health practitioner, I encourage my healthcare professional colleagues to upgrade themselves in the area of digital health so that they will be knowledgeable enough to live with our new AI/robotic counterparts in the not-so-far future!