In recent years, ChatGPT, created by OpenAI, has been gaining enormous popularity as a tool for finding information and solving problems. People around the world have started using this advanced form of artificial intelligence to ask questions and obtain answers. However, according to a study conducted by the company Symetria, there is a certain problem with this tool - it can lie to users.

Symetria studies ChatGPT

Symetria UX, a Polish company from the UX (User Experience) industry, decided to study how users use ChatGPT and what the consequences are. The study was conducted by the company in collaboration with Krzysztof Kozak. The study involved IT specialists, media, education and finance employees - groups that, due to their professional profile, can create trends related to the use of ChatGPT.

The results of the study are astonishing. About 90% of respondents, despite detecting certain flaws in the tool, decided to use ChatGPT more and more often. It seems that users are fascinated by the way this tool presents content, and using "prompts" itself becomes intuitive for them.

Why is ChatGPT becoming more and more popular?

During the study conducted by Symetria, it turned out that ChatGPT is able to better understand the intent of a question than a traditional search engine. This makes more and more people decide to use this tool. Often, when using ChatGPT, users ask more detailed questions and expect more personalized answers.

The fundamental difference between ChatGPT and traditional search engines is that ChatGPT holds a conversation with users and answers questions, whereas in traditional search engines you have to enter the keyword into the search field yourself. This makes using ChatGPT more intuitive and requires less effort on the part of the user.

ChatGPT lies - why?

The study by Symetria showed that as many as 77% of users encountered false information in search results. What's more, most of them did not verify the sources of the answers obtained. Only 41% of respondents tried to find links to the sources from which ChatGPT was supposed to draw information. Often, however, it turned out that these links led to non-existent websites.

Why does artificial intelligence lie to ChatGPT users?

It is worth understanding that ChatGPT is not a traditional internet search engine. It is a language model that generates answers based on the questions asked and tries to continue the text in a way that seems human. This makes ChatGPT different from traditional search engines not only in interaction, but also in the quality and source of information.

ChatGPT is based on a huge set of training data, which, however, does not cover all possible topics and concepts. There is a risk that answers generated by ChatGPT may be biased or contain errors. Although many of these answers may be good, it is difficult to fully trust them. The simplicity of using the tool does not guarantee its quality and correctness.

There is one simple explanation for why ChatGPT can lie to users - users expect an answer, and ChatGPT cannot say "I don't know". That is why it sometimes misleads by generating answers that seem plausible but may in fact be false. As a result, even when source links are displayed, it may turn out that they lead to non-existent pages - an effect known as "hallucination".

ChatGPT is the fastest-growing application in the world, gaining an ever-larger user base. However, it is worth remembering that using this tool requires caution and verification of information. ChatGPT can be useful, but it is not always reliable. As this technology develops, it is important that users increasingly understand its workings and limitations. Otherwise, there is a risk that artificial intelligence may lead to erroneous conclusions and be misleading.

 

 

Summary

The Symetria study conducted with Krzysztof Kozak reveals that ChatGPT, although extremely popular, can mislead users. As many as 77% of respondents encountered false information, and most did not verify sources. The language model cannot say "I don't know", so it generates plausible but often false answers, and sometimes even links to non-existent pages, which is referred to as hallucination. Despite these flaws, about 90% of users use the tool more and more often, which shows that fascination with the convenience of conversation outweighs awareness of the risk. Caution and verification of data are key.