Katharina Buczek
“We expect to see a significant change in the way medical AI will operate. Next, we will have devices that– rather than doing just a single task– can do maybe a thousand tasks, some of which were not even anticipated during model development,” Moor detailed.
So, in the study, the authors outlined how GMAI could work in a variety of applications– from taking notes to helping doctors make bedside decisions.
In radiology, for instance, models could theoretically draft radiology reports that visually distinguish abnormalities. At the same time, GMAI could take a patient’s medical history into account.
Still, there are certain capabilities and requirements needed in order to ensure GMAI is trustworthy technology.
Primarily, the model must consume both personal medical data and historical medical knowledge. Thereafter, the model should only refer to these data when used by authorized persons.
The model also must be able to maintain a conversation with a patient or doctor in order to collect new data or recommend different courses of treatment.
Among the authors, though, the largest concern is verification. In other words, ensuring that GMAI models are relaying accurate medical information.
Right now, AI chatbots like ChatGPT are gaining both widespread praise and criticism on the internet. Sometimes, this language model is spot-on. Other times, it spits out inaccurate information.
In addition to verification, concerns about privacy also represent a reasonable reservation with the technology.
“This is a huge problem because with models like ChatGPT and GPT-4, the online community has already identified ways to jailbreak the current safeguards in place,” Moor explained.
Finally, distinguishing between social biases and data is also a challenge for GMAI models. But, according to Moor, this problem must be tackled by the owners and developers of models– making sure biases are both identified and addressed before deploying GMAIs in hospitals.
