Tuesday, September 17, 2024
No menu items!
HomeData Analytics and VisualizationChatGPT Isn't a Smart Analyst

# ChatGPT Isn’t a Smart Analyst

ChatGPT has been trending on social media platforms. It has crossed one million users in just a week time. Those who haven’t heard about ChatGPT, it’s a large language model trained by OpenAI. In simple words, it’s a chat bot which answers your questions and the responses it provides may sound human-like. It’s an impressive machine learning solution but you can’t rely on it over Google search for learning on any topic.

Why ChatGPT Isn’t Smart enough?

You can’t trust ChatGPT for preparation on any certification or exam. It’s a Big NO if you think you can use ChatGPT for answering questions in a telephonic interview round. Yes I know it’s a cheating if you even use Google for the same but wanted to give a WARNING as many people do this and many social media influencers posted on how to leverage ChatGPT for cracking interviews.

I asked ChatGPT a few questions related to statistics and programming. See the responses below.

Q1. How c statistics is calculated :

A. percent concordant + (1.5* percent tied)
B. percent concordant + (0.5 * percent tied)
C. percent discordant + (0.5 * percent tied)
D. percent discordant + (1.5* percent tied)

Correct Answer : B
ChatGPT : A

Surprisingly if you ask the similar question in a different way, answer would be totally different.

Now I ask ChatGPT “c statistic = percent concordant + (0.5 * percent tied)”. It returns that is correct. Does it sound like human? May be yes human-like nervousness in an interview.

Q2. Does multicollinearity influence the predictions, precision of the predictions?

Correct Answer : No, collinearity does not influence the predictions and the precision of those predictions.
ChatGPT : Yes, multicollinearity can influence the predictions made by a model and the precision of those predictions.

Q3. Does oversampling of rare events affect sensitivity and specificity in a rare event model?

Correct Answer : Oversampling does not affect sensitivity and specificity measures but false positive and negative rates are affected.
ChatGPT : Oversampling of rare events can potentially affect sensitivity and specificity in a rare event model.

Q4. Can PROC MEANS analyze ONLY the character variables in SAS?

Correct Answer : No
ChatGPT: Yes

Q5. You have two vectors in R as defined below –

x y
If you run this vector z
Correct Answer : It returns 8 15 12 with the warning message. The length of z is 3.
ChatGPT : Error message

Final Note

Yes ChatGPT answers some questions correctly as well which I have not included in the post. The whole intend of this post is to show the responses of the model is not reliable. It’s too dangerous to use it regularly as you become addictive to it, you would start believing it and then the problem starts due to its poor accuracy. As of now it can’t be an alternative to Google Search. ChatGPT may have many usecases in customer support and other domains but not going to replace human jobs for sure in the current form.

Read MoreListenData

RELATED ARTICLES