ChatGPT

I was reading the Malwarebytes newsletter story called Trusting AI not to lie, a few days ago, I’m a big fan of their software and I’m subscribed. It speaks of a lawyer who asked ChatGPT for a list of previous cases that would support their main argument and allow them to continue the lawsuit. “But when the court reviewed the lawyer’s citations, it found something curious: Several were entirely fabricated. “ 

This is the main reason why AI Whisperers are being paid so much money. The AI didn’t lie, I don’t know the details of the story, but I do know that the AI did not lie. I’ve dealt with similar issues over the years when fixing software issues for people. The computer did it, they’ll say, or I didn’t do anything, it just happened. 

Computers don’t just do things, they do exactly as they’re told and this is also the case with AI. It will do exactly as it’s told. If asked for a list of cases to support an argument, it will provide them, that’s what you asked for. If you don’t state that they must be real world examples of actual Court cases, why would it give you those?  

ChatGPT was wrong, it says confidently, I sincerely doubt that. 

Anyone who has ever coded knows the difference a comma in the wrong place can make, well, the same thing is true when speaking to AI. Yesterday I was on Midjourney, trying to find a particular look for an image, the prompt I used first was this: 

Beautiful, pensive, wistful woman, pale skin, wearing a bejewelled headdress, detailed close-up, in white, intricate, sophisticated white, luxury, clothing with subtle details, pastel colours, porcelain doll, extraordinary makeup, pink and white clouds of smoke, studio lighting and volumetric lighting Photo taken by Karolina Skorek, taken with a Canon EOS R with a Canon RF 35mm F1.8 MACRO IS STM lens at 1/200 sec, f/2.8 and ISO200, Award-winning photography style, 4k –v 5.1 

I removed one sentence and got this: 

It gave me exactly what I asked for on both occasions. The woman in the picture does not exist in this world, she is a fictional character made up by the AI. If I wanted an image of a woman who actually exists, I can ask for one, but it won’t just do that, you must be clear about what it is that you want. 

The article closes by comparing ChatGPT to a cheap translation AI and states this: “That is an example of exactly this technology coming in and being treated as the arbiter of truth in the sense that there is a cost to how much truth we want.” 

They’re confusing apples and oranges with that nonsense. One AI is not the other AI. An AI based on ChatGPT is not ChatGPT. This is the same as me saying that all malware tools are rubbish because I have used one that didn’t meet my expectations. Over the years I’ve heard many complaints about computers which were a worthless waste of time, every time it turned out that the computers had been used improperly and worked perfectly well after some maintenance. 

I understand the attitude from people who write and work with code every day and whose living depends on it, but it is a fact that AI writes code 4 times better and far faster than any human can and in a few short years, coders will be obsolete. 

For a long time now, the phrase learn how to code was used as a joke when people were made redundant by progress and forced to change careers. That phrase won’t be around much longer.  

Perhaps it’s time we all learned how to Whisper. 

Cheers,

Old Man