• Please review our updated Terms and Rules here

ChatGPT gotcha - it gave me a straight wrong answer!

krebizfan said:
I received an email offering a product that would use ChatGPT to write the T-SQL code underpinning my databases. I welcome all my competitors to try it out first.
Click to expand...
Cue in techo-voice: "I apologize for any confusion in my previous response, and that the query I provided has wiped your database. Here is the right query for you to run it again. You do have backups, don't you?"
 
Last edited:
After watching multiple ChatGPT programming sessions, it seems to me that ChatGPT never gives someone what they want (it is incorrect) until they have talked it and talked it and talked it into what they want again and again. I'm not talking about 3 requests, but they are always like 20+ requests. That one about the radio shack trainer for example. It got the instruction set wrong right off the bat!
 
Yes, arguing with ChatGPT is the current national pastime until Baseball season proper starts up.
 
I honestly dont care about it and am frankly tired of it being forced on everyone as a fix all for everything. Much like how Facebook was forced on everyone...

The whole conversation is like someone talking to you about cell phones. Who cares its a phone...
 
One thing I notice about ChatGPT is it takes the dates and locations of historical events, throws them into a blender and presents them as fact. Or mixing up locations that happen to have the same or similar names. Here's my best impression: On November 15, 1989, US President George H.W. Bush and Georgy Malenkov signed the 1990 Chemical Weapons Accord in Huntsville, Alabama. On the same day, an F4 tornado touched down in Berlin, Nebraska.

ChatGPT may one day be useful for something, but as of now it's pretty much an experimental play toy.
 
ChatGPT is just fancy pattern matching. Useful for generating art. Not useful for anything requiring "correctness", like programming, engineering, or research.

Numerous lawyers have ended up in hot water over the past year after using ChatGPT to generate legal arguments which cited non-existent cases.
 
What has a bunch of vendors salivating is the prospect of ditching their entire tech support staff and replacing those people with a LLM AI engine.
 
For most vendors, the customers wouldn't notice, as the response you get, if you get one at all, is always from the same song sheet. Just the sort of task that suits the use of Automated Interaction.
 
Chuck(G) said:
What has a bunch of vendors salivating is the prospect of ditching their entire tech support staff and replacing those people with a LLM AI engine.
Click to expand...
Yep, and even more mediocrity than we have now will become the norm.
 
It'll be all over when the IRS and other federal agencies start replacing their phone agents with AI. Dealing with them is a headache enough NOW, but imagine having to be on the phone with the IRS for 2 hours because someone used your SSN to file a 2032 tax return, and the only responses you can get are along the lines of "As an AI language model, I am incapable of [insert action here], so you will have to use your implantable microchip to complete this action."
 
Even with humans on the far end, the IRS "help line" comes with the caution that "our answers to your question may be utter garbage and you'll still be on the hook if we're wrong". Yup. Nice job that an AI could do.
 
It's very-very easy to get absolutely wrong answers to straightforward questions, which completely defeats the purpose. I see no point in trying to get a quick answer to something if you need to go and verify it afterwards.
 
One of my chief complaints with how ChatGPT is used comes from a family member who is in Engineering school. He use it to generate references (that he has to verify) for papers that he is writing. I am glad he is actually writing his own papers (supposedly) but I learned so much during my time in school from doing my own research for sources, being exposed to ideas I had never heard of, and even completely changing the topics of papers after learning from unexpected search results. I'm not one of those that thinks you have to have every piece of knowledge in you head at all times, definitely go search for that formula that you seldom use but we are now losing critical thinking skills. ChatGPT is often wrong and humans are going to be unable to detect the errors.
 
Chuck(G) said:
What has a bunch of vendors salivating is the prospect of ditching their entire tech support staff and replacing those people with a LLM AI engine.
Click to expand...
At least then there would be NO EXCUSE for awful grating Indian accents. Give it a pleasant female voice, perhaps even make it sound ditzy and people will forgive it for being garbage.

Of course, I don't talk to machines unless I am YELLING at them.
 
You must log in or register to reply here.
Back
Top