583974_en-US

liliconcom

Terrifying: ChatGPT writes halfway decent JSL code

Watch out SEs, the robots are coming for us.

I asked ChatGPT this morning: "Can you write a JMP script in JSL for me that will create a distribution analysis of all continuous columns in the dataset?" Here's what it returned:

// Get the current data table
dt = Current Data Table();

// Get a list of all continuous columns in the data table
continuousCols = dt << Get Column Names( Continuous );

// Loop through each continuous column
for( i = 1, i <= N Items( continuousCols ), i++,
  // Get the current column name
  colName = continuousCols[i];

  // Create a distribution analysis for the current column
  dt << Distribution( 
    Continuous Distribution( Column( colName ) ),
    Fit Distribution( Normal )
  );
);

Check it out, it runs!! Then I replied: "that was pretty good, but I want the results in a single distribution report, not multiple reports." It gave me this:

// Get the current data table
dt = Current Data Table();

// Create a new distribution report
distReport = Distribution( Continuous );

// Get a list of all continuous columns in the data table
continuousCols = dt << Get Column Names( Continuous );

// Loop through each continuous column
for( i = 1, i <= N Items( continuousCols ), i++,
  // Get the current column name
  colName = continuousCols[i];

  // Add a distribution analysis for the current column to the report
  distReport << Add Column(
    Continuous Distribution( Column( colName ) ),
    Fit Distribution( Normal )
  );
);

// Display the distribution report
distReport << Show;

This one doesn't run for a couple of reasons: The Distribution() call is wrong, and the <

So I guess we don't have to worry about becoming redundant. Yet. But still, it freaked me out a little.