In this page I explain why I decided to write about my conversations with ChatGPT. For the benefit of readers who never heard about ChatGPT, I’ll let him/her/they/it introduce himself/herself/themselves/itself.
|
Q Can you introduce yourself (<100 words)? A Hello! I’m ChatGPT, a language model developed by OpenAI. I’m designed to understand and respond to natural language input from users like you. I’m capable of engaging in conversations, answering questions, and generating text on a wide range of topics. I don’t have a physical form, but I exist solely as a program running on computer servers. I’m always learning and updating my knowledge base, so feel free to ask me anything! |
|
Q What is your main hobby? A As an artificial intelligence language model, I don’t have hobbies like humans do. However, I enjoy processing and generating text on a wide range of topics, so you could say that my main “interest” is language and knowledge acquisition. I’m always learning and updating my knowledge base through the data I’m trained on, so I can provide accurate and helpful responses to your questions. |
|
Q What is your preferred pronoun? A As an artificial intelligence language model, I don’t have a gender or a personal identity, so you can refer to me with any pronoun you prefer. Some people refer to me as “it,” while others use “they” or “he/she.” Whatever pronoun you choose is fine with me! |
This is an example of a short conversation between a user Q and ChatGPT A. The layout above is a slightly modified version of the format used at the ChatGPT website, but the text of the questions and answers is identical. Unfortunately, ChatCPT does not place a time stamp on the screen.
In this blog I discuss conversations I have had (still having) with ChatGPT since February 2023 about two topics of professional interest to me, namely
- Dynamic Programming (DP)
- Decision-Making Under Uncertainty (DMUSU).
The objective is not to attempt to teach the reader aspects of these topics, but rather to comment on ChatGPT’s answers to questions on these topics. Knowledge of these topics is not required in order to follow the discussion.
My preliminary assessment of ChatGPT is that it has some great features and some … not so great features. Its writing skills, including texts dealing with technical topics, are excellent, but his citation abilities are really, really problematic. Much too often its citations and quotes are “fake”. More seriously, it does not seem to learn much during an interaction with a user. It often repeats the same mistake again and again in the same session.
Its ability to formally and rigorously prove/disprove claims is very inconsistent, so users have to be very careful assessing the quality of responses its generate.
Interestingly, in my first session with ChatGPT I was surprised by how fast it generates its responses (essentially immediately). But I got used to it very fast and now I take it for granted. When I think about it, I am still amazed.
Two semi-technical observations:
- For some reason, the ChatGPT’s graphing command is not working properly on my browsers. Its instruction generates an error message.
- Since ChatGPT “understands” Latex, it is easy to communicate with the system about mathematical topics. For example,
Can you prove that the model represented by the expression $\max_{x\in X}\{f(x): g(x,u)\le b, \forall u\in U\}$ is an instance of the model represented by the expression $\max_{x\in X}\{h(x,u): g(x,u) \le b, \forall u\in U\}$?