https://docs.google.com/document/d/19gl1Hd0HRFrWg76F06PL3DfyOKHrHpB1cHeuG8_Orjw/edit?usp=sharing
In this experiment, I opened up a ChatGPT prompt and asked, “In 250 words, tell me about the evolution of football tactics in the English Premier League.” I then read through the response and quickly noticed various mistakes in the factual statements provided in the text. While some of these factual errors were small or subtle, and may not have been noticed by someone that was not knowledgeable on the subject, they were not exactly difficult to see and could be easily fact-checked.
One of the main things that I learned during this exercise is that the information provided by ChatGPT is not always 100% factually correct. While I had already known that AI tools such as this one can often be incorrect, the information in the response to this prompt was not entirely incorrect. One example of this is shown in the response “By the early 2000’s . . . Arsene Wenger at Arsenal introduced a more technical, possession-based style . . . Around the same time, Jose Mourinho brought a stagmatic, defensively structured approach . . .” While the information about Wenger’s technical, possession based style and Mourinho’s stagmatic, defensive approach is correct, this response claims these events occurred in the early 2000’s, and claims that Mourinho’s style happened “Around the same time.” Although this statement may be somewhat correct, the real timing of the events is very different then what the ChatGPT response claims. Arsene Wenger arrived at Arsenal in 1996, which I would not call the early 2000’s. Jose Mourinho also arrived at Chelsea in 2004, an entire 8 years after Wenger’s arrival which I would say is not “Around the same time” as ChatGPT claims.
Overall, I found this experiment to be quite interesting and helped me learn a lot about ChatGPT and similar AI generators. I found it most interesting that the information provided was generally correct, but missed out on a lot of small details that impact the real story of what happened. I would say that I still generally “trust” this software and think it can be very useful, especially when seeking information on broad, well known topics. I think it is important to fact-check any information that ChatGPT provides, especially if you have no previous knowledge on the topic, as the facts can often be simplified or even completely wrong.