Ever since computers began to be used, people have speculated that one day a computer would be created that could outthink a human. The release of the ChatGPT app a few months ago seemed to indicate we are on the verge of that becoming reality. It will answer any question you ask it, in clear coherent English, or French, or any one of 93 other languages.

Researchers at the two children’s hospitals in Montreal recently decided to test the accuracy of ChatGPT’s answers. They asked ChatGPT 20 medical questions taken from recently published studies. ChatGPT responded to every question with seemingly authoritative answers. According to a report in la Presse, a Montreal newspaper, those answers were correct 60% of the time. The researchers concluded that it could be hazardous to your health to trust ChatGPT for medical information.

Labeling what ChatGPT does as Artificial Intelligence may give the impression that the app is capable of thinking. It is not. It scours the internet and vacuums up information of all kinds and when asked a question, it selects and repackages that information in clear, grammatically correct sentences. That is remarkable, but it is not evidence of thought, and if the Montreal study applies in other fields, it may not be very trustworthy. Somehow, it makes me think of the Wizard of Oz, just a masterful illusion.

One of the problems the Montreal researchers pointed out was that ChatGPT had an answer for every question; it never admitted “I don’t know.” I suppose in time AI apps will have some capacity to discern between reliable and unreliable information. Would I ever fully trust an app’s discernment? I don’t believe that would be wise; it would seem more like an attempt to manipulate public opinion than an attempt to provide unbiased information.