ChatGPT from OpenAI leaves me speechless over and over again. I have been in the AI industry for many decades now and it has been a long time since I last had this feeling of utter fascination mixed with disbelief mixed with anxiety.
This is only a quick post in the context of R programming which I wanted to share with you, so read on!
So, I asked ChatGPT to create a sample dataset and write some R code to analyze it:
As you can see the code comes fully documented already!
The table looks nice but I wanted to have it in csv format:
When I ran the code, I encountered an error, so I asked ChatGPT to fix it:
After that the code ran without any problems:
# Load the data into R crm_data <- read.csv("data/crm_data.csv", header = TRUE) # View the first few rows of the data head(crm_data) ## Customer_ID Name Address City State Zip_Code Phone_Number ## 1 1 John 123 Main St. New York NY 10001 555-555-5555 ## 2 2 Jane 456 Park Ave. Los Angeles CA 90001 555-555-5556 ## 3 3 Sam 789 Elm St. Chicago IL 60601 555-555-5557 ## 4 4 Sarah 987 Main St. San Francisco CA 94102 555-555-5558 ## 5 5 Mike 321 Park Ave. New York NY 10001 555-555-5559 ## Email Purchased_Product Purchase_Date ## 1 john@gmail.com Apple Watch 01/01/2022 ## 2 jane@gmail.com iPhone 12 01/02/2022 ## 3 sam@gmail.com AirPods Pro 01/03/2022 ## 4 sarah@gmail.com MacBook Pro 01/04/2022 ## 5 mike@gmail.com iPad Pro 01/05/2022 # View the number of rows and columns in the data dim(crm_data) ## [1] 5 10 # View the summary statistics of the data summary(crm_data) ## Customer_ID Name Address City ## Min. :1 Length:5 Length:5 Length:5 ## 1st Qu.:2 Class :character Class :character Class :character ## Median :3 Mode :character Mode :character Mode :character ## Mean :3 ## 3rd Qu.:4 ## Max. :5 ## State Zip_Code Phone_Number Email ## Length:5 Min. :10001 Length:5 Length:5 ## Class :character 1st Qu.:10001 Class :character Class :character ## Mode :character Median :60601 Mode :character Mode :character ## Mean :52941 ## 3rd Qu.:90001 ## Max. :94102 ## Purchased_Product Purchase_Date ## Length:5 Length:5 ## Class :character Class :character ## Mode :character Mode :character ## ## ## # Create a bar plot showing the number of purchases by state barplot(table(crm_data$State))
# Create a pie chart showing the percentage of purchases by product pie(table(crm_data$Purchased_Product), main = "Percentage of Purchases by Product")
# Convert the Purchase_Date column to a date format crm_data$Purchase_Date <- as.Date(crm_data$Purchase_Date, format = "%m/%d/%Y") # Create a scatterplot showing the relationship between purchase date and zip code plot(crm_data$Purchase_Date, crm_data$Zip_Code, xlab = "Purchase Date", ylab = "Zip Code")
Ok, that’s it for today… this is just unbelievable, isn’t it? Please share your thoughts and experience with this tool in the comments below!
UPDATE December 14, 2022
I created a video where I test ChatGPT with real exam questions (in German):
UPDATE December 22, 2022
To understand the general principle that underlies the inner workings of Large Language Models (LLMs) like ChatGPT, read my new post here: Create Texts with a Markov Chain Text Generator… and what this has to do with ChatGPT!.
Helpful post!
Thank you, Paul!
That level of error message interpretation is off the charts 😉
Yes, it’s mind-blowing, isn’t it!
No, it is not.
It’s just a toy, not even a tool, yet.
Agree.
Ask the program to multiply two large numbers (e.g. 4594 and 4741).
It will fail every time.
Yeah, very human-like, isn’t it 😉
One can always look for something that the model can´t do, but in the end this is – for me – next level.
It is really close to be a really comprehensive and powerful chatbot which could be in some way
life-changing. I could imagine this to be the next Google. “If you want to know something – GPT it! ;)”
It doesn´t matter that it is not a capable to do math operations. People will find out when to use chatbots and when to use something different as you also do it today.
We maybe highly overestimate the impact of consciousness or general intelligence on powerful models.
Thank you, great post! Nice to see early adopters like you giving it a try. Fascinating. Fixing errors is almost incredible. Yes, a bit scary as well …