Heh... Bing, Bard, BOOM! I've done some simple but non-basic tests on both and they've both failed miserably. Either the provide the wrong answer that wouldn't even run, produced the wrong answer because of "Silent Failures" in output, or produced code that was performance challenged.
As for making tables, it takes longer to verify that Bing, Bard, or ChatGPT wrote the correct code (and it's almost always missing the "pro" nuances that we all know, like the 2 part naming convention). That was also true of AI2SQL, which is supposedly written specifically for SQL. It didn't provide a correct answer either and, even worse, it wrote the answer in some other flavor of SQL than T-SQL even though I told it to use SQL Server.
And, no... I've not tried ChatGPT 4 with the new "code module". I'll let it mature for 6 months before I cough up the sawbucks to try it.