Short answer: in general, no, absolutely not.

I will limit myself to examples from ChatGPT-3 (as opposed to the recently released ChatGPT-4) and how it answers questions in mathematics. The benefits of this are that for any question we ask, there will be an objectively correct answer, and since ChatGPT-3 is free to use, you can easily go see for yourself what I am talking about.

The drawback is that this isn't an entirely fair test, in the sense that we will not look at the latest version of the program, and ChatGPT wasn't really trained to handle math. Some of the issues I will bring up are fairly easy to fix. Some of them, I am quite certain, are not.

The Chat GPT is a useful tool, but you have to know how to use it, at an academic level it makes doctoral theses to many people, but for example for more banal and updated issues is obvious that it has many errors, because the Chat GPT, only has information until the year 2021, everything that is forwards, events, etc, it does not know.

That picks up all kinds of information and all kinds of sources, the information from sources, is like everything else, sometimes they are right and sometimes wrong, occasionally it gives errors, especially if the questions are not well formulated, but then it corrects the errors throughout the conversation.

The better known a particular topic is, the more sources of information there are on that topic, and so the GPT Chat, as it is programmed to catch everything, will most often get the answers to those topics wrong.

Often it can also get names, dates, etc wrong, especially if the question is not specific enough, but it always corrects the mistake throughout the conversation, which is fine.

The advice I give, is to be insistent and specific, let the GPT Chat process the information, when it doesn't process try again later.

So it's reliable if you know how to use it, depending on the topic and subject.

Read more: https://www.youtube.com/@panokondoyiannis7011