Does ChatGPT have a sense of humor?
December 23, 2022 9:20 AM Subscribe
Does ChatGPT have a sense of humor? A funny writer asks ChatGPT to rewrite his work in the style of George Saunders, David Sedaris, and George Costanza. Hilarity ensues, but did the AI get the joke, or was the AI the punchline?
Ideally I could buy what amounts to a cell phone in the future that would develop it's own personality and personhood in the future. I imagine a future in which people "raise" AIs on their cell phones which then go on to larger servers off in the cloud to join their people.
Actually a movie plot, but I'm not going to spoil it for people.
posted by LionIndex at 10:00 AM on December 23, 2022 [6 favorites]
Actually a movie plot, but I'm not going to spoil it for people.
posted by LionIndex at 10:00 AM on December 23, 2022 [6 favorites]
Wasn't that the idea behind Tamagotchis?
posted by heatherlogan at 10:16 AM on December 23, 2022
posted by heatherlogan at 10:16 AM on December 23, 2022
Apparently Tamagotchis are a thing again, last weekend at a holiday gettogether my tween cousins were loudly discussing their Tamagotchis, including great lamentations how one's mom let theirs die while they were at school.
posted by AzraelBrown at 10:25 AM on December 23, 2022
posted by AzraelBrown at 10:25 AM on December 23, 2022
>we appreciate and enjoy art because we know that it was created by another human.
Being proved depressingly wrong in the visual arts as we speak.
posted by Sing Or Swim at 10:59 AM on December 23, 2022
Being proved depressingly wrong in the visual arts as we speak.
posted by Sing Or Swim at 10:59 AM on December 23, 2022
I see a dour future where we all are become entertainment executives constantly harassed by chatbots aggressively pitching tight fives into our headphones as they hone for open mic night at the chatbot Improv At The Live Stream.
posted by NoThisIsPatrick at 11:28 AM on December 23, 2022 [3 favorites]
posted by NoThisIsPatrick at 11:28 AM on December 23, 2022 [3 favorites]
I don't get Saunders, Sedaris or Costanza at all, but 'why not just crash here?!' and 'My parents were to me what The Huxtables or The Flintstones were to them', e.g., I do find absolutely hilarious (and Saundersian, for that matter).
posted by Flashman at 11:49 AM on December 23, 2022
posted by Flashman at 11:49 AM on December 23, 2022
David Sedaris, after having read the best efforts of a number of AIs has reacted, simply saying : "I can do better than that".
posted by nicolin at 12:00 PM on December 23, 2022
posted by nicolin at 12:00 PM on December 23, 2022
Does ChatGPT have a sense of humor?
Not to oversimplify but no, on account of how it's a machine.
posted by an octopus IRL at 12:33 PM on December 23, 2022 [7 favorites]
Not to oversimplify but no, on account of how it's a machine.
posted by an octopus IRL at 12:33 PM on December 23, 2022 [7 favorites]
Ok, maybe we can't program an AI to be funny, but what happens if we train an AI to imitate the style of a bad comedian? A Carlos Mencia, or a Jeff Dunham?
Though I suspect that training an AI to ape the style of Gallagher would only result in a terminator than runs around screaming all the time and hitting things with sledgehammers.
posted by LeRoienJaune at 12:39 PM on December 23, 2022 [4 favorites]
Though I suspect that training an AI to ape the style of Gallagher would only result in a terminator than runs around screaming all the time and hitting things with sledgehammers.
posted by LeRoienJaune at 12:39 PM on December 23, 2022 [4 favorites]
Does Clippy the Microsoft Paperclip have a sense of humour? The answer to this question is the same as the answer to the question in the title of the post.
posted by heatherlogan at 1:24 PM on December 23, 2022 [3 favorites]
posted by heatherlogan at 1:24 PM on December 23, 2022 [3 favorites]
Comedians know their profession is a formula, to some degree. Rule of three, etc.
posted by They sucked his brains out! at 2:45 PM on December 23, 2022 [2 favorites]
posted by They sucked his brains out! at 2:45 PM on December 23, 2022 [2 favorites]
Ideally I could buy what amounts to a cell phone in the future that would develop it's own personality and personhood in the future. I imagine a future in which people "raise" AIs on their cell phones which then go on to larger servers off in the cloud to join their people.
This story was already written by Ted Chiang in 2010. It's called "The Lifecycle of Software Objects"
posted by Insert Clever Name Here at 3:25 PM on December 23, 2022 [4 favorites]
This story was already written by Ted Chiang in 2010. It's called "The Lifecycle of Software Objects"
posted by Insert Clever Name Here at 3:25 PM on December 23, 2022 [4 favorites]
I spent a fair amount of time playing around with it and asking it to write various creative pieces, including poetry, limericks, haiku, short stories, etc. As a writer, my response is "somewhat clever, for a computer anyway." If you'd told me it was written by a middling 8th-grader, I'd have believed it.
As someone who has interfaced with artificial intelligence for more than 25 years, however, I have to say: It's pretty fucking amazing what this thing can come up with (as well as other forms of AI). Just to get to the point where a computer can sound like a middling 8th-grader is truly impressive.
posted by mikeand1 at 3:38 PM on December 23, 2022 [5 favorites]
As someone who has interfaced with artificial intelligence for more than 25 years, however, I have to say: It's pretty fucking amazing what this thing can come up with (as well as other forms of AI). Just to get to the point where a computer can sound like a middling 8th-grader is truly impressive.
posted by mikeand1 at 3:38 PM on December 23, 2022 [5 favorites]
Years ago, when John Scully was CEO of Apple, he started pushing his vision - The Knowledge Navigator. He had a short film produced to illustrate it. Big office, with big desk, in an obvious big house owned by some rich guy. On the desk is a laptop of sorts. Rich guy walks in, opens up the laptop and on the screen pops up what appears to be a live video of this guy in a white shirt and bow tie who right away starts telling the rich guy about his emails, upcoming meetings that day, etc. The rich guy, a professor, interacts with the bow tie guy by just talking to him, and they have a conversation about this and that, with bow tie guy being really obsequious to the professor, who now tells bow tie guy that he’s off to lunch date and will catch up with him later. Bow tie guy says yes in response, and as professor leaves the room, I could imagine a whole bunch of the employees in the room watching this video, were all adding in the voice of bow tie guy, “asshole……..” This was the first, as far as I know, depiction of a computer developing a personality and personhood. But only thanks to the cynicism of the employees in the room adding that expletive at the end.
posted by njohnson23 at 3:58 PM on December 23, 2022
posted by njohnson23 at 3:58 PM on December 23, 2022
I asked ChatGPT "Has Donald Trump ever lied to the American public?". it prevaricated ; "It is not within my capabilities to determine whether or not a specific individual has lied." etc.
So then i googled up some famous Trump lies, and presented them as a question, eg "does Rep. Ilhan Omar support al Qaeda?" It replied in the negative.
Wow, if they ever add basic propositional logic to ChatGPT, (if X is a lie, and Y says X, is Y a liar?) watch out...
posted by Artful Codger at 4:22 PM on December 23, 2022 [2 favorites]
So then i googled up some famous Trump lies, and presented them as a question, eg "does Rep. Ilhan Omar support al Qaeda?" It replied in the negative.
Wow, if they ever add basic propositional logic to ChatGPT, (if X is a lie, and Y says X, is Y a liar?) watch out...
posted by Artful Codger at 4:22 PM on December 23, 2022 [2 favorites]
I asked it to write me a fanfic of a videogame and it only got one canon detail wrong.
Unfortunately, it refuses to write smut.
posted by Jacqueline at 4:42 PM on December 23, 2022 [1 favorite]
Unfortunately, it refuses to write smut.
posted by Jacqueline at 4:42 PM on December 23, 2022 [1 favorite]
"So then i googled up some famous Trump lies, and presented them as a question"
Well, to be fair (and setting aside the obvious point that Trump is a liar), the question "Did this person lie" is more than just a question about whether they made a statement that is factually false; it's also a question about the speaker's state of mind or intent. If I say something false but I actually believe it, that's not lying, it's just being wrong.
BTW, I would not rely on Chat GPT to determine whether something is correct. I asked it a specific legal question just for fun, and it gave me a very authoritative-sounding answer and even cited a specific statute that purportedly stated the relevant rule. I looked up the statute, and although it was in the right area of law, with a title that seemed on-point, the statute didn't contain anything like the rule ChatGPT claimed it did.
posted by mikeand1 at 5:05 PM on December 23, 2022 [1 favorite]
Well, to be fair (and setting aside the obvious point that Trump is a liar), the question "Did this person lie" is more than just a question about whether they made a statement that is factually false; it's also a question about the speaker's state of mind or intent. If I say something false but I actually believe it, that's not lying, it's just being wrong.
BTW, I would not rely on Chat GPT to determine whether something is correct. I asked it a specific legal question just for fun, and it gave me a very authoritative-sounding answer and even cited a specific statute that purportedly stated the relevant rule. I looked up the statute, and although it was in the right area of law, with a title that seemed on-point, the statute didn't contain anything like the rule ChatGPT claimed it did.
posted by mikeand1 at 5:05 PM on December 23, 2022 [1 favorite]
ChatGPT is programmed to be a confident male with chronic answer syndrome. Whichever programmer was behind that little detail deserves a Turing Award, as faking confidence is about as human male as it gets.
posted by They sucked his brains out! at 5:10 PM on December 23, 2022 [10 favorites]
posted by They sucked his brains out! at 5:10 PM on December 23, 2022 [10 favorites]
but did the AI get the joke
That was a nice pun on "calculated." Can you explain the pun?
> Being proved depressingly wrong in the visual arts as we speak.
Hayao Miyazaki Has An Intense Reaction To AI Art - "The Oscar-winning Studio Ghibli co-founder says that he will never apply AI art to his work and that the art form is an insult to life itself."
That was a nice pun on "calculated." Can you explain the pun?
> Being proved depressingly wrong in the visual arts as we speak.
Hayao Miyazaki Has An Intense Reaction To AI Art - "The Oscar-winning Studio Ghibli co-founder says that he will never apply AI art to his work and that the art form is an insult to life itself."
- Will Transformers Take Over Artificial Intelligence? - "A simple algorithm that revolutionized how neural networks approach language is now taking on vision as well. It may not stop there."[1,2,3,4,5]
- Researchers Gain New Understanding From Simple AI - "Language processing programs are notoriously hard to interpret, but smaller versions can provide important insights into how they work."
- How Transformers Seem to Mimic Parts of the Brain - "Neural networks originally designed for language processing turn out to be great models of how our brains understand places."
Hayao Miyazaki Has An Intense Reaction To AI Art - "The Oscar-winning Studio Ghibli co-founder says that he will never apply AI art to his work and that the art form is an insult to life itself."
It's almost literally "Old man yells at cloud"
posted by LionIndex at 6:16 PM on December 23, 2022 [5 favorites]
It's almost literally "Old man yells at cloud"
posted by LionIndex at 6:16 PM on December 23, 2022 [5 favorites]
“old man yells at the cloud”
posted by fantabulous timewaster at 7:09 PM on December 23, 2022 [5 favorites]
posted by fantabulous timewaster at 7:09 PM on December 23, 2022 [5 favorites]
mikeandI: BTW, I would not rely on Chat GPT to determine whether something is correct. I asked it a specific legal question just for fun, and it gave me a very authoritative-sounding answer and even cited a specific statute that purportedly stated the relevant rule. I looked up the statute, and although it was in the right area of law, with a title that seemed on-point, the statute didn't contain anything like the rule ChatGPT claimed it did.
There's two things there:
- "determining" if something is correct. Well, outside of the most trivial rules of inference, I wouldn't want an AI to be "determining" anything. It could however be instructed, as part of its training, whether something is true or false. Beyond the naive assumption that everything that has been fed to it is true and accurate.
- you know that AI is, and will be used in future for planning, guidance, diagnosis... and probably in basic law too. So, it fucking well better be correct, or there's no point in relying on it.
I was only half-serious in posting the Trump example. I fear how AI will be trained, given the current state of "truth".
posted by Artful Codger at 7:46 PM on December 23, 2022
There's two things there:
- "determining" if something is correct. Well, outside of the most trivial rules of inference, I wouldn't want an AI to be "determining" anything. It could however be instructed, as part of its training, whether something is true or false. Beyond the naive assumption that everything that has been fed to it is true and accurate.
- you know that AI is, and will be used in future for planning, guidance, diagnosis... and probably in basic law too. So, it fucking well better be correct, or there's no point in relying on it.
I was only half-serious in posting the Trump example. I fear how AI will be trained, given the current state of "truth".
posted by Artful Codger at 7:46 PM on December 23, 2022
I'll let ChatGPT have the last word on AI in legal matters:
should AI be used to adjudicate legal questions?
AI has the potential to be used in a variety of legal applications, including adjudication. However, the use of AI in the legal system raises a number of ethical and practical concerns that need to be carefully considered.
One concern is the risk of bias in the training data used to develop AI systems. If the data used to train the AI system is biased, the system may produce biased or unfair outcomes. Ensuring that the training data is representative and unbiased is therefore crucial for ensuring the fairness and impartiality of the AI system.
Another concern is the lack of transparency in many AI systems, which can make it difficult for individuals to understand how the system arrived at a particular decision. This lack of transparency can make it difficult for individuals to challenge or appeal decisions made by the AI system.
There are also concerns about accountability, as it may be difficult to determine who is responsible for any errors or biases in the AI system.
Overall, while AI has the potential to be used in legal adjudication, it is important to carefully consider the ethical and practical concerns raised by its use and to ensure that appropriate safeguards are in place to protect the rights and interests of individuals.
posted by Artful Codger at 7:48 PM on December 23, 2022 [1 favorite]
should AI be used to adjudicate legal questions?
AI has the potential to be used in a variety of legal applications, including adjudication. However, the use of AI in the legal system raises a number of ethical and practical concerns that need to be carefully considered.
One concern is the risk of bias in the training data used to develop AI systems. If the data used to train the AI system is biased, the system may produce biased or unfair outcomes. Ensuring that the training data is representative and unbiased is therefore crucial for ensuring the fairness and impartiality of the AI system.
Another concern is the lack of transparency in many AI systems, which can make it difficult for individuals to understand how the system arrived at a particular decision. This lack of transparency can make it difficult for individuals to challenge or appeal decisions made by the AI system.
There are also concerns about accountability, as it may be difficult to determine who is responsible for any errors or biases in the AI system.
Overall, while AI has the potential to be used in legal adjudication, it is important to carefully consider the ethical and practical concerns raised by its use and to ensure that appropriate safeguards are in place to protect the rights and interests of individuals.
posted by Artful Codger at 7:48 PM on December 23, 2022 [1 favorite]
That’s a weirdly misandrist thing to say?
So be it, I'll own it to the grave.
posted by They sucked his brains out! at 7:51 PM on December 23, 2022
So be it, I'll own it to the grave.
posted by They sucked his brains out! at 7:51 PM on December 23, 2022
Wow, unironic use of the word misandry on Metafilter. Ghosts of Christmases long past.
posted by tigrrrlily at 8:20 PM on December 23, 2022 [4 favorites]
posted by tigrrrlily at 8:20 PM on December 23, 2022 [4 favorites]
When future alien archeologists write the history of the fall of human kind, they may pinpoint the moment we taught neural networks to form grammatically correct sentences before we taught them to understand the content of what they were writing.
posted by chromecow at 12:59 AM on December 24, 2022
posted by chromecow at 12:59 AM on December 24, 2022
ChatGPT's jokes are full-on uncanny valley with a side-helping of creepy sexism:
Why does a wife leave the lights on in the house? So she can find the way home after she runs away!
Wife: Why didn't you buy me a dress for Christmas? Husband: I thought you already had one - it's called a bed-sheet!
posted by verstegan at 5:09 PM on December 24, 2022 [1 favorite]
Why does a wife leave the lights on in the house? So she can find the way home after she runs away!
Wife: Why didn't you buy me a dress for Christmas? Husband: I thought you already had one - it's called a bed-sheet!
posted by verstegan at 5:09 PM on December 24, 2022 [1 favorite]
"Thanks for listening, and remember, don’t let the robots steal your soul."
Here is my experience with prompting ChatGPT to write stand-up routines about AI in the style of George Carlin, Richard Pryor, and Sam Kinison. Let's just say none will be heading to Netflix.
https://medium.com/@joetrip/adventures-in-ai-episode-six-comedians-on-ai-telling-jokes-4bef331707e1
posted by joetrip at 10:12 AM on December 26, 2022
Here is my experience with prompting ChatGPT to write stand-up routines about AI in the style of George Carlin, Richard Pryor, and Sam Kinison. Let's just say none will be heading to Netflix.
https://medium.com/@joetrip/adventures-in-ai-episode-six-comedians-on-ai-telling-jokes-4bef331707e1
posted by joetrip at 10:12 AM on December 26, 2022
If you'd told me it was written by a middling 8th-grader, I'd have believed it.
If AI progresses at real-time rates (which it will surely progress faster than), we're 10 years away from AI writing like a college graduate, and 12 years away from MFA-level first novels about a beautiful young assistant falling in love with a balding, middle-aged AI.
posted by Furnace of Doubt at 11:41 AM on December 27, 2022 [2 favorites]
If AI progresses at real-time rates (which it will surely progress faster than), we're 10 years away from AI writing like a college graduate, and 12 years away from MFA-level first novels about a beautiful young assistant falling in love with a balding, middle-aged AI.
posted by Furnace of Doubt at 11:41 AM on December 27, 2022 [2 favorites]
« Older 'To have it come around during this special time... | When you come to a crosswalk, take it. Newer »
This thread has been archived and is closed to new comments
I don't know if this makes me happy or sad. Ideally I could buy what amounts to a cell phone in the future that would develop it's own personality and personhood in the future. I imagine a future in which people "raise" AIs on their cell phones which then go on to larger servers off in the cloud to join their people.
posted by peterweyand at 9:39 AM on December 23, 2022 [4 favorites]