• Welcome to Avian Avenue! To view our forum with less advertisments please register with us.
    Memberships are free and it will just take a moment. Click here

Open AI Chat GPT

FiatLux

Jogging around the block
Celebirdy of the Month
Mayor of the Avenue
Avenue Spotlight Award
Joined
4/15/21
Messages
793
Real Name
Isabelle (she/her)
Wow! Has anybody tried open AI Chat GPT yet? I have just gotten so many of my questions about conures answered by this AI! I am kind of blown away by its capacity. @flyzipper? I’m sure we have members who are into AI but I just don’t know who they might be.
 

flyzipper

Rollerblading along the road
Celebirdy of the Month
Mayor of the Avenue
Avenue Spotlight Award
Joined
9/28/20
Messages
2,630
Location
Canada
Real Name
Steve
I've used it for a few things, but mostly I've been reading about its rapid progress and potential for disruption (and doom).

An example of that progress is the ChatGPT 3.5 engine scored a D on a certain college exam, but the ChatGPT 4 engine scored an A on the same test.
The other fascinating thing to observe has been the ethical and legal discussions that surround these technologies.

One anecdote from me is that I asked my cousin (a professional programmer) to write me a short script, and I used the same text description to have ChatGPT write me the same script -- I liked the ChatGPT result better.

This New Year image was generated by me with the help of an AI using text prompts (link).

I read one commentary that described two types of AI output: the first, is AI generated content that looks real at first, but upon close scrutiny we can see that it's not real, while the second emerging type is AI generated content that we believe is real even after we've been told it's not.

These are two examples:

1681278993376.png
(example of first type)

(example of second type)​

Another piece of commentary I read about modern music and musicians is that listeners have been conditioned to accept perfect beats from drum machines, and perfect pitch from auto-tune, so the listeners won't even notice when AI generated music is presented to them (like the example above).

Change is coming.
 

Xoetix

Biking along the boulevard
Mayor of the Avenue
Avenue Spotlight Award
Joined
6/12/22
Messages
6,954
Location
Southern VA, US
Real Name
Meghan (she/her)
I think it's fascinating, but a bit unsettling. I have a feeling the heavy reliance on it for so many things is going to disrupt a lot. Chatting with it is wild though.
 

anurim

Strolling the yard
Joined
1/16/20
Messages
148
I honestly find it pretty scary. It feels like AI in general has developed so so fast in the last few years.. like, let's be honest, the image with the pope?? It looks so real, especially for the untrained eye. When I saw it for the first time, I literally thought it was real, and considering I am a child of technology I find it even scarier.
As for ChatGPD, it's pretty cool. I have used it for some school work out of pure laziness, but some people do abuse it. Some schools have already banned it as far as I know. I don't take for granted the information it provides me with though, I don't feel like I can completely trust it yet.
They are still developing it, right?
 

BluJay

Sprinting down the street
Joined
12/22/22
Messages
356
Location
Eastern US
Real Name
Jay [he/him]
Oh yeah, they are still developing it. ChatGPT 4.0 is just the most recent release. Best believe that they have ChatGPT 5.0 in alpha or beta in the works if not fully developed and not released to the public at this point, it might be being used by government or military at this point. It is set to eventually replace some legal and financial jobs in the future. I read a article that they created an AI town with ChatCGT based characters similar to the Sims and they acted just like humans with their interactions and choices.

ChatGPT open gives better results than a Google search due to Google's ad optimization and other 'improvements' they have made over the past few years. Currently it's a tool, but I think it requires some oversight along with other current AI projects.

It gets things wrong on occasion because it is pulling from a collective data set, but there is some neat information you can discover on it. Take it with a grain of salt at this point.

I agree with the scary aspect of it. It's really the first AI (other than the art ones) open to the public. Who knows what is kept behind closed doors and security clearances at this point.

Jay
 

Tazlima

Jogging around the block
Avenue Veteran
Joined
3/7/19
Messages
632
Wow! Has anybody tried open AI Chat GPT yet? I have just gotten so many of my questions about conures answered by this AI! I am kind of blown away by its capacity. @flyzipper? I’m sure we have members who are into AI but I just don’t know who they might be.
I would STRONGLY caution against trusting Chat GPT to provide correct answers to questions. Honestly, I think the term AI to describe it is aspirational and not at all accurate. There's no thought or intelligence in what it does.

In its current form, Chat GPT is basically a really, really fancy version of your phone's predictive text function. It produces "answers" with no concern as to their accuracy, because it literally can't gauge accuracy. If you ask about conures, it will spit out what it can find about conures, but if you ask it about unicorns, you will get equally "confident" answers with perfect grammar and no concern for truth.

As a good example there are a lot of anecdotes about people asking it for a work with citations, and often it will just invent plausible-sounding titles and authors. It doesn't have a way to respond "sorry - there's no good source supporting the idea that the earth is made out of candy." It's just a phrase generator. You ask for words and it gives you words, even if you're asking for nonsense.

Even when it doesn't invent content out of whole cloth, it has all the biases and inaccuracies of whatever content was scraped to "teach" it. There are a lot of interesting tests people have done giving it identical prompts but changing one word and getting vastly different answers (the one I particularly remember was someone asked it to provide story ideas about a male/female detective, and the type of cases they might solve. The male detective got more action and adventure criminal-law type suggestions and the female detective got family law and social justice type suggestions).

Since you can't see where each "fact" came from, you have no way to judge whether the source is valid. Was that idea from Pamela Clark? An advertisement on Petco? PETA? Some loony who has a blog where they advocate pinioning as a good way to save the trouble of clipping wings on a regular basis? Who knows?! It's fact roulette!

If you want some wording for an important letter and don't know how best to phrase it, Chat GPT is a fantastic tool. But If you want to learn actual facts? A Google search is currently a thousand times more likely to get you a good answer, and you can also see where the answer came from and judge its accuracy. (Google has issues of its own, of course, but that's a whole other discussion).
 
Last edited:

Toy

Rollerblading along the road
Avenue Veteran
Mayor of the Avenue
Avenue Spotlight Award
Joined
4/14/10
Messages
1,731
Location
PA
With so much already fake stuff/info in today's world I fear this will just make things worse. It may be neat, but at what cost once the professional scammers get their hands on it?
 

FiatLux

Jogging around the block
Celebirdy of the Month
Mayor of the Avenue
Avenue Spotlight Award
Joined
4/15/21
Messages
793
Real Name
Isabelle (she/her)
I was actually quite impressed with the accuracy of the responses I got on chat GPT 4.0 I asked rather specific questions about Golden conures. I know from having to dig into published research articles and having to email an expert that the breadth and depth of the responses chat gpt provided were well sourced and accurate. Thst is not to say that I would rely on it blindly and not cross reference but still I am impressed. I actually do not think that the fact that it operates by scraping and sourcing from across the web universe of data points is much different than how the human brain synaptically processes information from a multitude of sources. Nor do I think that that way of coming up with answers is inherently less accurate than the way that humans collect information. I remain keenly aware of the dangers in AI.
 
Top