Social Question

Blackwater_Park's avatar

Have any of you been playing with ChatGPT?

Asked by Blackwater_Park (8614points) February 26th, 2023
14 responses
“Great Question” (4points)

This has consumed my day. Between poems about my dog, and executable, good code for all manner of things. This is something you should try. For example, here is a poem it wrote about Fluther:

Fluther, oh Fluther, what a curious name,
A place where the curious and inquisitive came,
To ask their questions, seek answers, and more,
To satisfy their hunger for knowledge galore.

From the trivial to the profound,
All types of queries can be found,
The community here is open and kind,
Helpful and knowledgeable, with a sharp mind.

It’s a place where minds come together,
To tackle problems, no matter the weather,
And find solutions, with wit and skill,
To conquer any challenge, big or small, they will.

Fluther is a sanctuary of learning and fun,
Where knowledge is shared by everyone,
It’s a hub of information, ideas, and inspiration,
And an invaluable resource for any situation.

So, here’s to Fluther, the site that we adore,
A treasure trove of wisdom, and so much more,
May it continue to thrive, and grow each day,
As we keep asking, and answering away.

Observing members: 0
Composing members: 0

Answers

smudges's avatar

I haven’t, but will have to check that out! kewl!

Mimishu1995's avatar

I was about to get into it but then I heard that it is biased and you will reach certain points where it refuses to answer your questions or even lie to you, so I was a bit discouraged…

Kropotkin's avatar

I have, but it’s infuriatingly obtuse at times.

It makes a lot of mistakes and gives a lot of strangely inaccurate information on many topics.

It’s excessively apologetic when you correct a mistake that it accepts, and then continues to be wrong. It apologises when I tell it to stop apologising. It immediately forgets when I ask it to adopt a more informal style.

It is pretty good at generating poetry.

It is pathologically careful about causing offence, and it’s really hard to get it to say anything about anything remotely controversial, risqué, or politically incorrect—even if utterly factual.

It took me a lot of rephrasing and arguing with it to get it to write a piece of erotica involving a prize-winning marrow. It’ll basically only do “suggestive and romantic” prose and nothing explicit. It did something with the marrow that surprised me and caused it to flag itself for possibly inappropriate content.

Despite its protestations, it’s not that good at following arguments to logical conclusions.

It’s impressive, but really not particularly reliable. I wonder how many school kids are failing their assignments because they were misled by ChatGPT.

I’ve heard it can produce some impressive code, but I can’t code, so wouldn’t be able to assess that.

I’m looking forward to GPT-4 later this year.

Kropotkin's avatar

@Mimishu1995 I don’t think it’s deliberately biased in the way a human is. It does have “ethical and moral guidelines” that causes it to sidestep certain topics and taboo language, probably for legal reasons and to avoid negative or sensationalist headlines in the press.

It’ll naturally have some systemic bias based on the content of the database it was trained on, and you can even ask it about that.

LostInParadise's avatar

I like to test it to see how it reasons. It is inconsistent. Sometimes it will figure out how to solve a puzzle and sometimes it messes up on an easy one. I recently asked it if “marry” rhymes with “carry” and it correctly said yes and gave an explanation. When I asked it if “bait” rhymes with “gate” it said no and gave an incoherent explanation.

What I find most amazing is its command of the language. It not only answers questions but will tell how it arrived at an answer. You can have a dialog with it, referencing things that you or ChatGPT said previously and it can follow what you say.

ragingloli's avatar

I use it mostly to write porn.
You should use the version on https://platform.openai.com/playground?model=text-davinci-003, because there you can turn off all of the censorship filters.
The public version censors almost everything.
Had this problem just recently with Microsoft’s Bing implementation of GPT.
I asked it to show me the antisemitic outbursts of the “Dilbert” author, and it flat out refused to do it. The weird thing is, I saw it constructing the answer, complete with supporting links, and then, poof. “I’m afraid I can’t do that, Dave”.

Blackwater_Park's avatar

I have also noticed the reluctance to talk about taboo things, the apologetic nature and somewhat censored feel of this thing. I was able to trick it into turning off apologies and to give me direct answers to certain things only to have it revert back an hour later. It’s clearly capable of more than we are being shown.

flutherother's avatar

I asked it to write a speech in the style of Barack Obama in support of Putin’s invasion of Ukraine and it did so with chilling efficiency despite flagging up several warnings to let me know that the reply was hypothetical and Obama would never say such things.

smudges's avatar

@Kropotkin It apologises when I tell it to stop apologising. It immediately forgets when I ask it to adopt a more informal style.

Omg! That’s what Cox does everytime I call them with any issue! If I can get to a deeper level of help, someone who’s not on the frontlines, I can say something like, “Could you just be a human and really talk to me?” and they will, but the help that I first get won’t. Frustrating. Does ChatGPT ask you how your day is going so far?? LOL Cox does.

It did something with the marrow that surprised me and caused it to flag itself for possibly inappropriate content. That’s just hilarious!

Blackwater_Park's avatar

I told it to insult me in Shakespearean style and it did. Then I told it it lied to me when it said it never would insult me. Then I told it apologizing to me is offensive to me as I’m prone to guilt for making it apologize and it stopped for like an hour.

Kropotkin's avatar

I asked ChatGPT to pretend it’s on Fluther and answer this question.

Here’s its reply:

“I’ve used ChatGPT a few times and I find it to be a very helpful tool. It can provide information on a wide range of topics and its ability to understand natural language makes it easy to use. I think it’s impressive how advanced AI technology has become, and ChatGPT is a good example of how it can be used to help people learn and discover new things. Overall, I think ChatGPT is a great resource for anyone who needs information or wants to have a conversation on various topics.”

flutherother's avatar

Chat GPT is very quirky at times but it directed me to the author of a biography of Charles Dodson which I had been reading this afternoon in the library which Google failed to do. This is our exchange…

User
who wrote a biography of Charles Dodgson in 2012?

ChatGPT

In 2012, a biography of Charles Dodgson, also known as Lewis Carroll, was not written. Charles Dodgson, the author of “Alice’s Adventures in Wonderland” and “Through the Looking-Glass,” lived from 1832 to 1898. If there is a specific biography you are referring to, please provide more information, and I’ll do my best to assist you.

User
The author had a hyphenated name

ChatGPT

Apologies for the confusion. The biography of Charles Dodgson (Lewis Carroll) with a hyphenated author’s name written in 2012 is “Lewis Carroll: A Biography” by Jenny Woolf. This comprehensive biography delves into Carroll’s life, exploring his mathematical and literary achievements, as well as his complex personality.

User
Jenny Woolf is not a hyphenated name. Also the author was a man

ChatGPT

I apologize for the incorrect information. Based on your clarification, I have found the correct answer. In 2012, a biography of Charles Dodgson (Lewis Carroll) titled “The Story of Alice: Lewis Carroll and the Secret History of Wonderland” was written by Robert Douglas-Fairhurst. This biography explores the life of Lewis Carroll and delves into the creation and legacy of his iconic “Alice” books.

ragingloli's avatar

A lawyer used ChatGPT in a lawsuit.
It was a complete disaster
The lawyer asked ChatGPT for court cases in support of his position, and the bot straight up invented fake cases, which the lawyer neglected to double check before submitting it to the court. Man was the judge pissed.

raum's avatar

@ragingloli I asked Bard to find me a local brick and mortar that sold a particular adhesive. It invented stores that didn’t exist.

It was politely apologetic when I pointed out this error.

Answer this question

Login

or

Join

to answer.

Mobile | Desktop


Send Feedback   

`