PEAX Equipment

Artificial Intelligence and Public Comment

I asked a programmer friend about it and he told me there is a feedback option where you can paste in text and ChatGPT will give a percentage chance the text was written by the program. It was apparently used recently to catch university students submitting ChatGPT written papers.
 
I asked a programmer friend about it and he told me there is a feedback option where you can paste in text and ChatGPT will give a percentage chance the text was written by the program. It was apparently used recently to catch university students submitting ChatGPT written papers.

I think this will work, until it won't. Basically, as these text-producing AI's get better and better, the difference between human-created and AI-created text will become more indiscernible. I think in many instances, on many topics, to many readers, it already is.

From what I have read, Parameters are a good metric to look at to understand the capability of the systems. ChatGPT-3 has 175 billion, ChatGPT-4 will have 100 trillion.

1673968670776.png
 
I have no doubt you are correct. The inherent imbalance between the cost of use and the cost of policing that use now is troubling. It will probably only grow as the program becomes more capable.
Are you aware of any current books that examine this issue in depth?
 
I have no doubt you are correct. The inherent imbalance between the cost of use and the cost of policing that use now is troubling. It will probably only grow as the program becomes more capable.
Are you aware of any current books that examine this issue in depth?

No. It's all pretty new. I myself am just someone who is interested and doesn't know much. A blog I read daily is Margina Revolution https://marginalrevolution.com/
They stay pretty on top of ChatGPT and it's potential reprucussions. For example, cyber criminals are already using it for numerous nefarious acts.

 
I think this will work, until it won't. Basically, as these text-producing AI's get better and better, the difference between human-created and AI-created text will become more indiscernible. I think in many instances, on many topics, to many readers, it already is.

From what I have read, Parameters are a good metric to look at to understand the capability of the systems. ChatGPT-3 has 175 billion, ChatGPT-4 will have 100 trillion.

View attachment 261137
This statement stuck out. This seems literally impossible. It has human supervisors and is learning from human-created content, but aims to be unbiased? Someone need to include some data about today’s humans.

Better human-supervised training may help reduce the likelihood of GPT/ChatGPT generating toxic/biased content, and may also help reduce instances of misinformation.
 
This statement stuck out. This seems literally impossible. It has human supervisors and is learning from human-created content, but aims to be unbiased? Someone need to include some data about today’s humans.

Better human-supervised training may help reduce the likelihood of GPT/ChatGPT generating toxic/biased content, and may also help reduce instances of misinformation.
I heard Musk was gonna buy it.
 
No. It's all pretty new. I myself am just someone who is interested and doesn't know much. A blog I read daily is Margina Revolution https://marginalrevolution.com/
They stay pretty on top of ChatGPT and it's potential reprucussions. For example, cyber criminals are already using it for numerous nefarious acts.

After messing around with it a little, I swear I’ve gotten phishing-type emails that sound very similar to the ChatGPT text outputs I was getting.
 
Since I doubt I'll ever try it, could someone please ask it how to: 1. Avoid an argument with my wife. 2. Win inevitable argument. 3. Handle aftermath of winning inevitable argument. 😂 J/K. I rarely argue with my wife as she is way smarter than me.

As a person that routinely solicits public comment for various projects, I can see this being used both by both sides. I've never thought about questioning the validity of a letter or person sending it. Also, a letter gets the same amount of consideration as a phone call or face to face comment.
 
I asked a programmer friend about it and he told me there is a feedback option where you can paste in text and ChatGPT will give a percentage chance the text was written by the program. It was apparently used recently to catch university students submitting ChatGPT written papers.
Even if the AI response isn't directly copied into a writing assignment, how hard izzit to rewrite the AI response that was the totality of research one did for the assignment?
 
Big Tech Exec1: We can have AI write our new software product for $10 Million, or we can hire 50 US developers, or 500 Indian developers.

Big Tech Exec2: Screw it, human software always sucks. Let's just pay the money and get the software complete today.
 
"ChatGPT, the popular chatbot from OpenAI, is estimated to have reached 100 million monthly active users in January, just two months after launch, making it the fastest-growing consumer application in history, according to a UBS study on Wednesday."


 
I would wager that sooner than later it will be capable of generating that which is “novel” in a way that will be indiscernible from the way humans create novelty.

What’s happening on the backend may or may not really matter. What will matter is whether or not anyone can tell a difference.

Some would say Creativity itself is just a complex algorithm...

Resurrecting this because last night I read this paper which came out 3 days ago. It's short, but interesting.


"We found no qualitative difference between AI and human-generated creativity, although there are differences in how ideas are generated. Interestingly, 9.4 percent of humans were more creative than the most creative GAI, GPT-4. "

In the paper, they do kind of question what it means to be truly creative, as these AIs require a prompt - an ask - before they will create. When you have created things, or when I have, is it "prompted"? It doesn't feel like it, but I don't know if that matters. I think we are in the middle of a highly disruptive event - as disruptive as the internet was - but it's tough to yet tell how.

In the context of the OP - Public Comments - it makes the need for some sort of verification, even more dire. Right now, the vast majority of humans can't concoct arguments, and the cases for those arguments, better than these AIs.
 
Leupold BX-4 Rangefinding Binoculars

Forum statistics

Threads
110,816
Messages
1,935,439
Members
34,888
Latest member
Jack the bear
Back
Top