ChatGPT’s generative AI has proven particularly brilliant when it comes to healthcare. What to ask if, in the more or less near future, doctors will not be replaced by an artificial intelligence.
News: ChatGPT recently passed all three parts of the US medical licensing exam with flying colors in an experiment, it reports Axios.
- The researchers who conducted this experiment point out that first-year medical students typically spend hundreds of hours preparing for Part 1 of the exam. Only medical graduates pass Part 3.
The detail : Generative AI has proven particularly effective at answering various exam questions, without ever having been trained on a medical dataset.
- As part of their research, the researchers made sure that “none of the answers, explanations or related content was indexed on Google” and therefore the results provided by ChatGPT were indeed of its own initiative.
- However, the researchers say in their report that a number of “indeterminate” responses were excluded from the results, as ChatGPT was programmed to avoid providing what could be construed as medical advice.
- “These answers were so general that it was hard to tell whether they were right or wrong,” said Morgan Cheatham, co-author of the paper and a medical student at Brown University.
To notice : so if generative AI managed to pass the three exams, there’s still no doubt that it will replace a real-life doctor or medical student.
- However, it can be useful for explaining certain concepts to patients after consideration by a trained professional.
- It is for this purpose that the startup Ansible Health, focused on the treatment of chronic obstructive pulmonary disease (COPD), has used ChatGPT.
- The company’s CEO and former Google product manager, Jack Pro, wanted to see if the enthusiasm around this AI was rational or if it wasn’t just a fad. “When we started doing the validation, we were quite amazed with the results. Not just what was going well, but how it was explained,” he said.
However, artificial intelligence could prove useful in medicine in the more or less distant future.
Will AI be a doctor soon?
In fact, we can imagine that AIs will evolve in such a way as to go beyond the text phase, to be able to be endowed with speech and, later still, with a body and a face. At this stage, they would be able to modulate the tone of their voice, adapt their body language and facial expressions according to the situation, so that they could interact “humanly” with beings made of flesh and blood. .
In this case, AI could prove useful for ensuring the well-being of patients, but also for carrying out certain tasks of the general practitioner.
Yet, for the moment, we are still a long way off.
Reality : in fact, we are not close to seeing an artificial intelligence make a diagnosis on a patient autonomously, and this, even if it is able to swallow all the data of his medical record.
- AI models like ChatGPT still make claims today that turn out to be false, which could be dangerous, especially in a medical context.
- However, the AI has already revealed itself interesting help in the health sector and could grow in popularity as impressive, if not effective, models like ChatGPT emerge.
So ? While it’s still years, if not decades away from seeing an AI doctor, ChatGPT could now become an involuntary medical consultant, thus replacing “Dr. Google” to the general public looking for answers to their medical questions.