ChatGPT versions 3.5 and 4 have been shown to be as good at determining the calorie and nutrient content of food products as professional nutritionists.
Although algorithms cannot replace doctors, scientists believe that the technology will be useful as an auxiliary tool for people who monitor the amount of energy and nutrients they consume.
The article was published in the journal JAMA Network Open;
How did the ChatGPT nutritionist skills test go?
Researchers at Taipei Medical University were interested in how ChatGPT, which is becoming increasingly popular among users looking for answers to medical questions, could handle the nutritional value of food.
They tested ChatGPT versions 3.5 and 4 by asking the algorithm to indicate the calorie and fat, carbohydrate, and protein content of 8 adult menus that contained 222 foods in total. It turned out that both chats indicate the calorie content of foods with a margin of error of up to 10 percent in 35-48 percent of cases. ChatGPT 4 was somewhat more accurate overall, but it had a higher error rate when estimating the protein content of foods.
The researchers conclude that the accuracy of the chats is not significantly different from the nutritionists’ evaluations of products. At the same time, they note that ChatGPT is not a search engine, so its answers may depend on many factors, such as the specifics of the query, the language of the query, and the chat environment, which users should pay attention to.
