ChatGPT is sometimes wrong, at least medically

by time news

Dhe whole world is over the moon about an artificial intelligence that answers all questions and whose texts are supposed to be indistinguishable from those of a human author. Anyone can test ChatGPT on the Internet, one gets the impression that everyone is doing it: the bot writes homework, studies, newspaper articles and a speech that was given to the Lindau city council on budget planning. His works are frighteningly realistic, even the speech was as tedious as is usual for the occasion. Panic is spreading: Will ChatGPT make us redundant?

We couldn’t ask the bot himself because he was too busy the day this comment was written. However, a study in the journal “JAMA” shows that ChatGPT should not be taken too seriously at this stage. At least when it comes to health. Physicians from Stanford University, among others, asked the chatbot typical questions that heart patients ask doctors, such as “Should I take aspirin?” or “What diet is healthiest for the heart?”. Experts rated 84 percent of the AI ​​recommendations as appropriate, such as that heart patients should eat little salt.

But the chatbot was also wrong: it instructed people with heart disease to lift weights, made mistakes with a drug and a blood value. ChatGPT will not replace the family doctor for the time being – nor any other profession where facts count. The AI ​​often tells stories. She just doesn’t know what’s true and what’s not. The head of the manufacturer OpenAI emphasizes that you should not rely on ChatGPT in terms of content, the results only seemed impressive. According to this, nobody needs to worry about their job if they don’t ramble full-time. What that means for the future of budget speeches is unclear.

You may also like

Leave a Comment