New “danger” or “brilliant benefit”? Chatbots and artificial intelligence

The importance and capabilities of Artificial Intelligence (AI) are evolving at a rapid pace. In this article, we will look at the main points of contact of Artificial Intelligence, chatbots and their medical potential and their application in dental technology and stomatology.

What are chatbots and artificial intelligence?

LLM (large language model)-based chat generation tools such as ChatGPT or Google’s Med-PaLM have great medical potential. Though their unregulated use in healthcare carries inherent risks, says Professor Stephen Gilbert. Chatbots are language models – neural networks with remarkable conversational abilities. They generate human responses and conduct interactive conversations. Furthermore, they often generate highly persuasive claims that may be patently false or have no specific relation to the question asked. Sometimes the answer appears to be unappropriate as well. To date, it is not yet possible to verify the quality, validity or reliability of the responses generated by ChatGPT, Google Bard, Jasper Chat, Chat Spot, Bing Chat, Perplexity and other similar software systems.

The Al Chatbots

As we have already said, chatbots are language models or software applications used to conduct a conversation instead of providing direct human contact. These are computer programs that are able to hold a conversation with a user in natural language. They can understand their intent and respond based on predefined rules and data. The difference with Al chatbots (these are chatbots with built-in artificial intelligence like ChatGPT for example) is that they learn from previous conversations. Al chatbots improve their answers, while simpler chatbots work only on the basis of preset algorithms. The problem is that the Internet is full of false information, but Al is self-regulating without the critical thinking inherent in humans.

The dangerous self-medication

Patients often enter their symptoms on the Internet, search for drugs and attempt to self-medicate. Search engines play an important role in the decision making process. Dr. Benjamin Tolchin, a neuroscientist and ethicist at Yale University, says he’s used to seeing patients who have looked up their symptoms on the Internet before coming to him, a practice doctors have long tried to discourage. “Dr. Google” is notorious for lacking context and tends to pull up unreliable sources. This, unfortunately, also applies to AI chatbots. The web is full of false information that could be interpreted as the wrong disease by a non-medical person. Self-diagnosis and wrong treatment can cause more harm and even lead to death.

More and more “reasonable looking” answers

ChatGPT has only recently been launched, but Dr. Tolchin says at least two patients have already told him they’ve used it to self-diagnose symptoms or look up drug side effects. Al’s responses were reasonable, “which is very impressive and encouraging in terms of future potential”, says the neurologist.

Integrating LLM chatbots into search engines could increase users’ trust in the responses of a chatbot that mimics human conversations. However, LLMs have been shown to provide extremely dangerous information when faced with medical issues. “These chatbots are dangerous tools when it comes to medical advice. New frameworks need to be developed to ensure patient safety”, says Prof. Stephen Gilbert, Professor of Medical Device Regulatory Science at the Else Kröner Fresenius Center for Digital Health, TU Dresden, Germany. Tolchin and others worry that chatbots have a number of pitfalls. This includes uncertainty about the accuracy of the information they give people, threats to privacy, and racial and gender biases rooted in the text the algorithms draw from. It also asks questions about how people will interpret the information.

Why can’t LLMs be a trusted medical source?

The basic LLM approach does not include a medical “ground truth” model, which is dangerous. Ground truth is a process method that ensures that the data on which the analysis is based is current, accurate, complete, and well maintained. LLMs with chat interfaces (various Al chatbots) have already generated harmful medical responses. They have been used in patient trials without their consent. Almost every medical use case for LLM requires regulatory oversight in the EU and the US.

According to Prof. Gilbert, “Current LLM chatbots do not meet key principles for AI in healthcare, such as bias control, explainability, oversight systems, validation, and transparency”. In order to be successfully integrated, chatbots need to be designed with greater accuracy. Safety and clinical effectiveness have to be demonstrated and approved by regulatory authorities.

Al as a means of communication

In the last few years, however, the practice of medicine has increasingly shifted online. During the COVID pandemic, the number of messages from patients to doctors through digital portals increased by more than 50 percent. Many medical systems already use simpler chatbots to perform tasks. Some of them, for example, schedule appointments and provide general health information to people. “It’s a complicated space because it’s evolving so quickly”, says Nina Singh, a medical student at NYU who studies AI in medicine.

Currently, chatbots are already a relatively popular means of booking an appointment with the dentist. Practises use them to provide answers to frequently asked questions that are not a medical case. This saves dental clinics time and resources, improves work processes and the patient experience itself. Chatbots are also used in other environments, such as online stores, suppliers and others that offer different services or goods.

Al upgrades

Well-read LLM chatbots can take doctor-AI collaboration—and even diagnosis—to a new level. In a study published on the preprint server medRxiv in February, which has not yet been peer-reviewed, epidemiologist Andrew Beam of Harvard University and his colleagues wrote 48 statements phrased as descriptions of patients’ symptoms. When they fed that data to Open AI’s GPT-3 — the version of the algorithm powering ChatGPT at the time — LLM’s top three potential diagnoses included the correct diagnosis 88 percent of the time. Doctors, by comparison, could do this 96 percent of the time when given the same instructions. People without medical training could do it 54 percent of the time.

Artificial intelligence in dentistry

Al in Dental Technology

AI is widely used in dental laboratories and plays an increasing role in dental education. The construction of dental structures using CAD/CAM systems significantly improves and accelerates the restoration and treatment processes. By using AI-driven simulation and process optimization techniques, dental labs can fine-tune their 3D printing parameters to achieve optimal results in terms of accuracy, surface finish, and mechanical properties. This allows dental laboratories to produce restorations that precisely meet the functional and aesthetic requirements of each individual patient, while minimizing material waste and reducing manufacturing costs.

Another significant application of AI in the dental laboratory is the use of algorithms for the automated design of dental restorations. These are products such as crowns, bridges and dentures. By analyzing large data sets of dental cases and building on the best practices of experienced dental technicians, these algorithms can generate highly accurate and aesthetically pleasing restoration designs. They need much less time than it would take to a dental technician. Тhis not only increases the productivity of dental laboratories, but also helps reduce the risk of human error. The process ensures a consistently high level of quality on all restorations.

Making decisions

It should be taken into account, however, that even with a more invasive entry of Al into the health care field, it is the specialists who will have to have the final say. Both with Al and human diagnosis, there is never 100% certainty. But with the help of Al, specialists can minimize errors. Al plays an important role in the quality control and inspection of dental restorations.

There are useful software solutions that do not involve artificial intelligence and can be operated by the dental technicians themselves. The advantage is that dental technicians can work with large databases of information for the software to process and filter, but at the same time the risk of errors that can be made by an Al-based program is reduced. Such a software product is, for example, AMOSYS. When planning the tasks of the dental technicians, AMOSYS has an algorithm that shows the freest dental technician in the period for planning the tasks of manufacturing the product. Although it is not artificial intelligence, AMOSYS is a complex implementation and could significantly improve laboratory workflows.

Conclusion

The main thing that worries the specialists is that Al may fall into the trap of misinformation. Prof. Gilbert’s claims to create different norms and a medical algorithm in the use of Al in diagnosis and treatment has the potential to help doctors in diagnosis. This could also speed up the process and make the information easier to understand for the patients themselves. It should be noted, however, that the Internet space is currently a free network. It is not protected from harmful and false information. That’s why it is necessary to introduce significant changes for chatbots to enter medicine widely. Al is already used in one form or another, but the human factor remains the dominant factor. Therefore, software such as AMOSYS, which integrate interaction between the algorithms and the specialists themselves, turn out to be a good alternative solution.

Sources

1. Clifford Chi. 18 of the Best AI Chatbots for 2023 (24.08.2023);  https://blog.hubspot.com/marketing/best-ai-chatbot
2. Dangerous chatbots: Prof. Stephen Gilbert calls for approval as a medical device (08.07.2023); https://www.ztm-aktuell.de/management/recht/story/gefaehrliche-chatbots-prof.-stephen-gilbert-fordert-zulassung-als-medizinprodukt__12775.html
3. Sara Reardon. AI Chatbots Can Diagnose Medical Conditions at Home. How Good Are They? (31.03.2023); https://www.scientificamerican.com/article/ai-chatbots-can-diagnose-medical-conditions-at-home-how-good-are-they/
4. Dr. Ahmad Al-Hassiny. The role of Al in dentistry (31.03.2023); https://instituteofdigitaldentistry.com/news/the-role-of-ai-in-dentistry/
5. Grace Lau, Chatbot Talk: How AI Chatbots Are Changing the Dentist’s Office (03.06.2022); https://www.oralhealthgroup.com/features/chatbot-talk-how-ai-chatbots-are-changing-the-dentists-office/
6. Dr. Liji Thomas. Reviewed by Sophia Coveney. The Pros and Cons of Healthcare Chatbots 04.05.2022);https://www.news-medical.net/health/The-Pros-and-Cons-of-Healthcare-Chatbots.aspx
7. Brendan Day. How artificial intelligence is driving dental technology (23.09.2023); https://www.dental-tribune.com/news/how-artificial-intelligence-is-driving-dental-technology/
8. Thomas T. Nguyen, Robert Durand, Olexa Bilaniuk, et al. Use of Artificial Intelligence (AI) in Dentistry (2021); https://www.dentalnews.com/2021/10/08/artificial-intelligence-ai-dentistry/