POSITIV Business & Style

Česko-anglický magazín mapující úspěchy českých podnikatelů, inovace, investiční příležitosti a trendy v lifestylu s distribucí po celém světě. / Czech-English Magazine Mapping the Successes of Czech Entrepreneurs, Innovations, Investment Opportunities, and Lifestyle Trends, with Global Distribution.

Pacient často přichází s vytištěným výstupem od chatbota.

www.posiv.cz ǀ 43

HEALTHCARE

questions people around the world ask language

models are related to health – either their own or that

of someone close to them. This is not a marginal issue.

And I can confirm it from my own neurology clinic:

patients come in with a printed document that was

created during a conversation with a chatbot. General

practitioners from the Moravian-Silesian Region tell me

the same thing. We have to be prepared for this.

Is it a problem, or an opportunity?

Both. The problem is that ordinary language models

are not medical devices, because they are not

trained primarily on medical data, they carry the risk

of hallucinations, and their outputs may not always be

reliable. I see the opportunity in the fact that a patient

who receives a complicated report from a doctor

can have it explained simply by a chatbot. That

is a legitimate use.

OpenAI, after all, recently launched ChatGPT Health,

a model focused on health-related topics. For now,

it only operates in the United States, because stricter

regulations apply in Europe. But the direction is clear:

both OpenAI and Anthropic are aware of how large

a share of health-related queries is directed at their

platforms, and they are responding by developing

more specialised tools.