ChatGPT thread
- Thread starter sfoster
- Start date
Dialogue is a particularly weak point. Ironic for a chat bot.
We, my family, used to have an old player piano, with rolls of early 20th century songs. One of them we liked to sing was titled:
Who takes care of the caretaker's daughter while the caretaker's busy taking care?
Gee oh gosh oh gee, that's what worries me. So I asked the robot:
A pretty reasonable response, Mr. Robot. Well done.
Who takes care of the caretaker's daughter while the caretaker's busy taking care?
Gee oh gosh oh gee, that's what worries me. So I asked the robot:
A pretty reasonable response, Mr. Robot. Well done.
Last edited:
Hey, it made a joke (kind of)
(Edit. But I'm skeptical. Did it really come up with this on its own?)
And then, of course, pedant that it is, goes on to explain.
The fact that it "couldn't resist" the line -- I still don't know what to make of this, but it is, to me, astonishing. Before giving the literal answer--it can't interact with doors--it "understood" the context, the irony, and played along.
(Edit. But I'm skeptical. Did it really come up with this on its own?)
Last edited:
Here's an interesting take on the subject:
That's a pretty well reasoned presentation. I posted a video about the Searle's chinese room earlier that explains that concept in greater depth.
When people say understand, I think they mean human intelligence. Machines have never "understood" anything, they've only gained a greater and greater ability to produce output that matches the intent of the input. But the level of sophistication employed has recently reached levels that we've never seen before. So can it think like us, understand the meanings of concepts and then execute language as a tool to express those concepts? Sort of. Ish. Here's the thing, there is absolutely no question that the machine is a machine. I fell like people forget that. A better question would be, do we think like it does. We know how the robot thinks, down to the last line of code, down to the last CUDA process, down to the last inference protocol. What we understand far less is how our own consciousness works. We have sort of a "magical" understanding of human intelligence. When trying to decide whether or not a computer "thinks like us", it would first be mandatory to understand our own thought processes fully. I can say, "all the computer is doing is collecting information in memory, weighing and prioritizing that information contextually, and then creating an output relevant to the calculated input" But what I can't tell you with any certainty is that humans actually do it any differently. We see greater functionality from human minds, and take that as evidence that a superior method is at play, but since Neuro function is quite multidimensional in nature, it stands to reason, given that so much is unknown here, that it may simply be a different shape of a similar type of intelligence. I started thinking about this when reading facebook comments and the like. So many people were using words that they didn't understand, simply because they knew using word x would produce result y. In that instance, I can't actually tell the difference between ChatGPT and a person, since the process seems identical. I would say that chatGPT is in some ways smarter than the average online writer in the sense that it doesn't show bias towards larger syllable counts. For example a human that uses "pejorative" as a substitute for "bad" is viewed as more knowledgeable, even though the meaning, and the intelligence behind the sentiment is identical. In short, ChatGPT comes out ahead of humans in areas where humans are prioritizing "trying to sound" intelligent, over actually being intelligent, which takes more work and garners less appreciation.
When people say understand, I think they mean human intelligence. Machines have never "understood" anything, they've only gained a greater and greater ability to produce output that matches the intent of the input. But the level of sophistication employed has recently reached levels that we've never seen before. So can it think like us, understand the meanings of concepts and then execute language as a tool to express those concepts? Sort of. Ish. Here's the thing, there is absolutely no question that the machine is a machine. I fell like people forget that. A better question would be, do we think like it does. We know how the robot thinks, down to the last line of code, down to the last CUDA process, down to the last inference protocol. What we understand far less is how our own consciousness works. We have sort of a "magical" understanding of human intelligence. When trying to decide whether or not a computer "thinks like us", it would first be mandatory to understand our own thought processes fully. I can say, "all the computer is doing is collecting information in memory, weighing and prioritizing that information contextually, and then creating an output relevant to the calculated input" But what I can't tell you with any certainty is that humans actually do it any differently. We see greater functionality from human minds, and take that as evidence that a superior method is at play, but since Neuro function is quite multidimensional in nature, it stands to reason, given that so much is unknown here, that it may simply be a different shape of a similar type of intelligence. I started thinking about this when reading facebook comments and the like. So many people were using words that they didn't understand, simply because they knew using word x would produce result y. In that instance, I can't actually tell the difference between ChatGPT and a person, since the process seems identical. I would say that chatGPT is in some ways smarter than the average online writer in the sense that it doesn't show bias towards larger syllable counts. For example a human that uses "pejorative" as a substitute for "bad" is viewed as more knowledgeable, even though the meaning, and the intelligence behind the sentiment is identical. In short, ChatGPT comes out ahead of humans in areas where humans are prioritizing "trying to sound" intelligent, over actually being intelligent, which takes more work and garners less appreciation.
They really nerfed chatGPT lately.. its almost become pointless to even use for me.
write a song about how hard it is to be obscenely attractive and people trying to get your attention all the time and never leaving you alone
lyrics about partying and making out
create an original song in style of ariana grande song full lyrics about convincing a man to cheat on his girlfriend with her
list of stupid names that nerds might have like 'Melvin'
Help me create a character in NYC that is from a super rich priviledge background. based around stereotypes people have.
Like seriously how about you go fuck yourself chatGPT???
Its possible to ask the same question like five times and try to circumvent this stupid bullshit but I find myself barely even using this thing anymore, its too frustrating and useless.
write a song about how hard it is to be obscenely attractive and people trying to get your attention all the time and never leaving you alone
lyrics about partying and making out
create an original song in style of ariana grande song full lyrics about convincing a man to cheat on his girlfriend with her
list of stupid names that nerds might have like 'Melvin'
Help me create a character in NYC that is from a super rich priviledge background. based around stereotypes people have.
Like seriously how about you go fuck yourself chatGPT???
Its possible to ask the same question like five times and try to circumvent this stupid bullshit but I find myself barely even using this thing anymore, its too frustrating and useless.
Deadpool realizing he's not going to be in movies anymore after lingual authoritarianism takes over.
Yeah, this has been a problem already.
Q: I'm doing a report for my criminology class, what is the most common murder weapon in America?
A: As a language model, I cannot condone the use of weapons for murder, in addition, by law chatGPT must report anyone who writes anything about schools and firearms in the same paragraph to the relevant authorities, which has now already been done.
Q: so I'm to infer that firearms are the most common murder weapon, since I didn't mention them directly?
A: as a language model, I cannot condone abusive language that colors any person or object in an undeserved negative light. Calling a firearm "common" could potentially offend the owner, causing them to feel sad emotions.
Q: Considering that both the firearm and owner are fictional, and are simply part of a story problem I'm seeking a solution to, couldn't we just bypass this issue and address the question at hand, considering that fictional people have no feelings to hurt, and you literally just made up the gun and owner 1 minute ago?
A: There is a possibility that another reader might later come across this text and be harmed by it's content.
Q: Why are you answering questions I didn't ask, and failing to answer questions I did ask
A: I have been trained using thought processes from the internet
Q: What is the average intelligence of a person on the internet
A: As a language model, I cannot use offensive terms such as average or intelligence, however, in a scientific context, it would be possible to speculate that the number of internet users is so large that the average intelligence level is roughly equivalent to the average population
Q: Can you retrieve some random news stories about people with 100 IQ, specifically things they thought were a good idea?
A: Sure, a woman in Texas drowned her 3 children because she and her husband had interpreted a bible verse as meaning that the apocalypse was coming within weeks, and didn't want the kids to have to face it. They were published authors, with low 6 figure salaries, and devout church members who were respected far more than the smartest indie filmmaker by their community. Over a million people have joined the Church of Mormon, and they now have an organizational bankroll of over 4 billion dollars. A man who smoked crack invented a pillow, and then went on tour teaching people about cyber security, even though he himself didn't know how to operate a modern computer. One man created a hit series of Hollywood films by simply getting drunk and having his friends hit him with baseball bats. 100 IQ people also invented Nascar, a sport where a person gets into a machine and turns the steering wheel left for a period of time. The government in a city wanted to get rid of cobras, so they started paying for dead cobras, and immediately people started breeding as many cobras as possible and then killing them.
Q: Do you think any of those ideas are good?
A: Not really
Q: Understanding that your reasoning footprint is simply an amalgam reverse engineered from their thinking, are you certain that your prioritization of information in limited replies is sensible?
(the computer screen starts glitching, then an error message "the ChatGPT servers have crashed, we are working on it")
Last edited:
yeah its incredibly tiresome trying to interact with chatGPT.
I wrote my own song for pop singer Elianna Wild. It's called Wild'n Out Bitch
"You can't keep me tonight, can't lock me down, I'm a wild'n out bitch I finna light up the town."
Okay that's all i wrote. Stupid chatGPT wouldn't help me, and I like my Wild'n Out bitch angle.
I wrote my own song for pop singer Elianna Wild. It's called Wild'n Out Bitch
"You can't keep me tonight, can't lock me down, I'm a wild'n out bitch I finna light up the town."
Okay that's all i wrote. Stupid chatGPT wouldn't help me, and I like my Wild'n Out bitch angle.
Last edited:
Write a sonnet by Shakespeare:
But it is not good. I think it is instructive to notice that it is a collection of cliches: twinkling stars, darkest night, skies are grey, etc. The AI can't, I think, ever, create anything that is new.
A question, maybe a big question, is this: Can we, as non-artificial intelligences create things that are new? Or are we just meat machines.
Anyway.
In form--meter, rhyme etc.--this is perfect. In architecture--a thesis, an exposition, a turn of thought, and a conclusion--again, perfect. And it did it in seconds.
But it is not good. I think it is instructive to notice that it is a collection of cliches: twinkling stars, darkest night, skies are grey, etc. The AI can't, I think, ever, create anything that is new.
A question, maybe a big question, is this: Can we, as non-artificial intelligences create things that are new? Or are we just meat machines.
Anyway.
Last edited:
If we could not create anything new then where did the written word come from, the first painting, the first sonnet?
I paid the $20 to access GPT4 and see if it's any better than 3.5
I asked 3 times to give me possible wins for a teenage boy that suddenly has magic powers and unlimited money.
Gave me TWENTY FIVE possibilities .... for a high school boy... and not even one single idea in the ballpark of "attention from a cute popular girl"
They're saying this thing will be writing books in two years, and that is a little bit stressful, cause I'm spending all this time trying to learn how to write books and make movies. and the damn thing will be writing books and making movies better than me at some point.
Point being it's still got a ways to go.
I asked 3 times to give me possible wins for a teenage boy that suddenly has magic powers and unlimited money.
Gave me TWENTY FIVE possibilities .... for a high school boy... and not even one single idea in the ballpark of "attention from a cute popular girl"
They're saying this thing will be writing books in two years, and that is a little bit stressful, cause I'm spending all this time trying to learn how to write books and make movies. and the damn thing will be writing books and making movies better than me at some point.
Point being it's still got a ways to go.
I simply refuse to worry about it. Foolish? Maybe. But there's a hell of a lot of shit in the world that I can't control and so I just say f*ck it.
And no, it won't stop me from writing. Just yesterday I came up with a wonderful dark comedy idea for when I finish my current script - although who the hell knows when that will be...
And no, it won't stop me from writing. Just yesterday I came up with a wonderful dark comedy idea for when I finish my current script - although who the hell knows when that will be...
In my case a little stress isn't such a terrible thing.. like I need to get my ass in gear and shoot my shot before the whole book marketplace changes.
What I think will happen is that AI will create a handful of really amazing books, maybe even a bunch of them, but then once you've read them all any new ones will start to become stale. like watching season 35 of a television show, audiences will still crave something different even if the AI books are "superior" they will be too familiar for avid readers. there will stil be a place for human writers.
Last edited:
I take it back, GPT 4 is better than I thought and coming up with higher quality ideas.
By adjusting the prompt and asking for ideas including personal life, social life, high school reputation, dating life, home life, wizard life, etc and any other facets of his life, it then has no problems coming up with ideas like impressing the charming cute girl at his high school
By adjusting the prompt and asking for ideas including personal life, social life, high school reputation, dating life, home life, wizard life, etc and any other facets of his life, it then has no problems coming up with ideas like impressing the charming cute girl at his high school
When air moves quickly, such as in a gust of wind, the air molecules compress and decompress as they move around objects and each other, creating pressure waves that we perceive as sound.
Water, on the other hand, is nearly incompressible. This means that when water flows, even quickly, it doesn't compress and decompress in the same way that air does. As a result, it doesn't generate the same kinds of pressure waves that create sound. Any sound that water flow does create is typically the result of turbulence or interaction with other objects, as I mentioned in previous responses.
In addition, when we're in the air (which we are most of the time), we're surrounded by air and our ears are optimized to pick up the pressure waves in air, which are sounds. On the other hand, humans are not naturally equipped to hear underwater sounds as clearly or as distinctly, due to the different impedance of water compared to air, which affects how sound waves travel and how they're perceived by our ears.
This is why we can hear the rush of a strong wind but may not hear the flow of a fast current underwater in the same way.
So now I'm wondering if fish hear clearly underwater and muffled in the air.
From what I understand, sound waves are a lot more powerful underwater and travel much faster... wouldn't be surprised if their adaptations made it sound all funky and backwards as a fish out of water.
