Over the past 25 years, the medical community has had to learn to use “Dr. Google,” thanks to the role the giant’s search tool has played in resolving health doubts. Now doctors, hospitals, researchers and startups are preparing for the arrival of a replacement that promises to be even more powerful as the “family doctor”: the ChatGPTthe chatbot from a powerful Artificial Intelligence (AI) from the American startup OpenAI.
The use of AI in healthcare is not exactly new areas like radiology that work with image recognition have already made important advances. For years, clinics and researchers have also been researching ways of introducing technology into the everyday lives of doctors, patients and institutions they should save lives and bring economic benefits. For example: the American NGO National Office for Economic Research estimates that the widespread adoption of AI systems could save the US between 5% and 10% in annual healthcare spending—somewhere in between 200 billion dollars It is $360 billion.
However, OpenAI’s “smart” chatbot sheds new light on the technology that is shaking the healthcare world and forcing the medical community to understand what to do with the tool. Judging by the initial interest, in the near future it will be very difficult not to visit a practice without a system.
Continued after the ad
On Monday 20th, the Faculty of Medicine of the University of São Paulo (FMUSP) sponsored a symposium on ChatGPT, where specialists from different areas of the institution presented the potential of the tool. A stream free on youtube it reached 1,500 concurrent viewers.
Among the possible applications, there seems to be a consensus that these systems will be fundamental in medical communication, which does not always go smoothly between professionals and patients. “ChatGPT can communicate about side effects and medications, answer pretest questions, and provide basic health advice. It can be a basic unit of interaction with the patient,” said Chao Wen, head of the faculty’s telemedicine department, at the event.
Although many people today refuse to communicate with simple chatbots (like those of a mobile operator), ChatGPT’s sophistication in text creation can do the opposite and create an approximation. “People are more likely to believe in the machine when it has anthropomorphic characteristics,” he explains Estao Alexandre Chiavegatto Filho, Director of the Big Data and Predictive Analysis in Health Laboratory (Labaps) from USP.
So trust grows when the AI resembles a person in any way, which ChatGPT does well in text building. This ability can enhance actions similar to searching for health information in the Google.
The direct relationship with the doctor can also benefit from the chatbot. With the ability to skillfully generate text, ChatGPT could fill out the medical records, freeing up consultation time, allowing the doctor to develop a closer relationship with patients. The American startup Nuance is dealing with this topic. In a relationship with Microsoftthe company has developed such a tool that GPT4latest version of ChatGPT’s “brain”.
Named after Dragon Ambient eXperience Express (DAX) it identifies the speeches of doctors and patients and automatically populates the document with symptoms and diagnoses to create a summary of the consultation it can record up to four hours of conversation. Those interested in the system should keep an eye out for Nuance, as the service hasn’t been made publicly available yet.
The automation goes even further: “ChatGPT can help fill out health plan approval forms, which also speeds up the speed of service,” says Chiavegatto.
Finally, the medical community can already think of ChatGPT as a remote patient monitoring tool that allows treatment management and followup. For example, the chatbot could serve as a reminder to take medication and record patient actions.
“We are at a point in healthcare where treatment is cost prohibitive. There is no other option but to rely on AI support if we want to provide people with the medical care they need,” says Mariano GarcíaValiño, CEO of Healthtech Axenya the startup has a solution that combines AI with one human team combined to monitor patients diabetes.
Left Quotation Arrow We are at a point in healthcare where treatment is cost prohibitive. There is no other option but to rely on AI support if we are to provide people with the level of medical care they demand from Arrow Aspas Direita
Mariano GarcíaValiño, CEO of Healthtech Axenya
human surveillance
In fact, nobody can imagine that intelligent systems will completely do without human help at least not in the short term. In the future, medical teams will need to monitor and filter what the machine is telling patients. “In the end, the doctor is still responsible,” says Giovanni Cerri, President of the Hospital das Clínicas Innovation Committee.
Concerns are pretty important. ChatGPT is a text generation tool probabilistic models, which means that the quality of the content is not the priority of the system. This means that he demonstrates certainty about every fact even if the probability that he is right is small. In addition, the database used in the system training has information up to 2021 and is not specific to the medical field.
“Perhaps in the future we can envision specific healthcare algorithms linked to institutional business intelligence. It would be a kind of health cloud,” Wen argued at the FMUSP event. Algorithms capable of searching medical knowledgefocused databases are on the radar of those developing the technology.
Last year Microsoft experimentally introduced the BioGPT, an AI using ChatGPTlike technology trained on data from the biomedical literature. In tests, BioGPT outperformed other general AIs in biomedical text generation tasks.
On Thursday 23rd, OpenAI also revealed ChatGPT’s ability to have “plugins” meaning it is now possible for companies to connect specific databases consulted by the chatbot, guaranteeing specific knowledge associated with the simple language of the tool is presented. . At first there is none plugins of health in the modality, but the new model proves that such a project is possible.
Future
Experts point out that a great revolution will take place when health is provided by predictive algorithms. So AI systems that read patientspecific data and diagnose diseases and Predicting the evolution of treatments. ChatGPT cannot do such a thing.
“These are very complex algorithms that require a lot of testing. “When it comes to health, the simplest answer is not the best. Simple answers are very rare,” says Chiavegatto. However, he points to the importance of the OpenAI tool to open the frontiers for technology.
Left Quote Arrow ChatGPT has put all AI in the spotlight. It goes down in history because it was the first time people actively searched for a Right Quote Arrow algorithm
Alexandre Chiavegatto, director of Labdaps
“ChatGPT has put all AI in the spotlight. It goes down in history because it was the first time people actively searched for an algorithm,” he says.
In an editorial, the scientific journal Nature pointed out the need for the entire medical community to devote attention to the topic. “The use of ChatGPT and other language models in healthcare requires careful consideration to ensure safeguards are in place for potentially harmful uses,” the text reads.
With the right protection, the new “family doctor” promises enormous potential. “Models like ChatGPT are an opportunity to expand human capabilities, not an opportunity to be a lazy shortcut,” says Chiavegatto.