ChatGPT AI Tool— We all work for robots now

[bernorange 8340]

ChatGPT aint no W.H.O.P.P.E.R.

[George 487]

23 hours ago, bernorange said:

* bump *

I added a ChatGPT integration to a site I own and have been playing around with it this morning.  The quality of the output is definitely a function of the quality of the question, but it can put out some insightful responses.

I need to read up on this. Are you able to limit the data sources it uses or direct it to “learn” from specific repository of info?

[bernorange 8340]

The API integration that I'm using is using the gpt-3.5 model which does not allow customizations.  They have other models and I seem to recall seeing that you could pay for access to models that you can seed yourself, but I can't recall what page I saw that on.  Some context here:

https://platform.openai.com/docs/models

[CHIEF 5044]

My son and his buddies have started using chatGPT for writing lab reports and term papers for class. Cut and paste a supporting paragraph from an article into chatGPT that supports your view on the subject, and ask it to rewrite it with a few extra comments.  It takes about 45 minutes, and it turns out a stellar report or essay that cannot be recognized as being plagiarized (and let's be real, this is what we were basically doing manually in college).  This has been a boon for him, as he has dysgraphia, and writing has always been his big "thorn" since kindergarten. This is the new norm in colleges apparently. He and his fellow students are way ahead of the curve.

CHIEF

Edited March 23, 2023 by CHIEF

[bernorange 8340]

Quote

What is the airspeed velocity of an unladen swallow (both african and european)?  Please answer in the style of cheech and chong

~~~

Hey man, that's some heavy stuff you're asking there. Like, what do you mean by "unladen," man? Is that like, a bird that's not carrying anything, or is it like it's unloaded its burden, and is now free to fly as high as it wants? And what about the African and European swallows, man? Are they chillin' together, or are they like, from different gangs or somethin'?

I mean, it's like, you're asking for some deep theories on bird flight here, man. And like, we gotta think about all the factors, man. The wind speed, the temperature, the bird's carbon footprint, all that stuff.

But you know what, man? Let's just say, like, the airspeed velocity of an unladen swallow is, like, infinite, man. The bird's got no limits, no boundaries, man. It's just flyin' on the winds of freedom, and nothing can hold it down. And whether it's African or European, man, it don't matter, 'cause when they're up there in the sky, they're all just one big flock, flyin' together to the music of the universe, man.

 

[Mole 4309]

4 hours ago, CHIEF said:

My son and his buddies have started using chatGPT for writing lab reports and term papers for class. Cut and paste a supporting paragraph from an article into chatGPT that supports your view on the subject, and ask it to rewrite it with a few extra comments.  It takes about 45 minutes, and it turns out a stellar report or essay that cannot be recognized as being plagiarized (and let's be real, this is what we were basically doing manually in college).  This has been a boon for him, as he has dysgraphia, and writing has always been his big "thorn" since kindergarten. This is the new norm in colleges apparently. He and his fellow students are way ahead of the curve.

CHIEF

I’m on my college’s academic appeals committee. We’ve had our first batch of AI-based violations. Those AI-assisted papers are pretty easy to spot, even without the tools. They’re impeccably written, sound exceptionally smart on the surface yet never really get to the point (talking about the question rather than actually answering the question), and are riddled with made up stuff. Getting a committee to agree to sanction the student is a bit trickier than spotting the obvious cheater, at least so far on my campus.

Policies and paper procedures will be radically changing over the next few years. So far, most of my colleagues have talked about returning to hand-written, blue book essays, or using North Korean styled proctoring services even for term papers. I’ve been playing with the idea of all or nothing, public, oral defenses of all term papers the last week of class. I may not be able to get the cheater past the appeals committee, but I can still wring out a zero on the paper and a little bit of public shame.

[HamsterHookah 112]

1 hour ago, Mole said:

I’m on my college’s academic appeals committee. We’ve had our first batch of AI-based violations. Those AI-assisted papers are pretty easy to spot, even without the tools. They’re impeccably written, sound exceptionally smart on the surface yet never really get to the point (talking about the question rather than actually answering the question), and are riddled with made up stuff. Getting a committee to agree to sanction the student is a bit trickier than spotting the obvious cheater, at least so far on my campus.

Policies and paper procedures will be radically changing over the next few years. So far, most of my colleagues have talked about returning to hand-written, blue book essays, or using North Korean styled proctoring services even for term papers. I’ve been playing with the idea of all or nothing, public, oral defenses of all term papers the last week of class. I may not be able to get the cheater past the appeals committee, but I can still wring out a zero on the paper and a little bit of public shame.

What's wrong with the plagiarism tool that you used to have to upload your essay to, and it would search the internet or archives to see if there was a % match and spit back if it was an original or not? In theory wouldn't that work for something like ChatGPT or not?

[HamsterHookah 112]

Also, I posted this on the hobbies/books/reading thread, but I don't get the sense a ton of people go there, even if they are readers, so I'll share here:

This is an interesting, and very short, treatment on a well-worn AI-cum-evil master trope from the 60's called "I Have No Mouth, and I Must Scream". If you fancy yourself well-read or transgressive or a technologist (like one of those guys who quotes Snow Crash when talking about Metaverse), it's worth your 20 minutes: https://wjccschools.org/wp-content/uploads/sites/2/2016/01/I-Have-No-Mouth-But-I-Must-Scream-by-Harlan-Ellison.pdf

https://en.wikipedia.org/wiki/I_Have_No_Mouth,_and_I_Must_Scream

[Loop 1604 226]

5 hours ago, CHIEF said:

My son and his buddies have started using chatGPT for writing lab reports and term papers for class. Cut and paste a supporting paragraph from an article into chatGPT that supports your view on the subject, and ask it to rewrite it with a few extra comments.  It takes about 45 minutes, and it turns out a stellar report or essay that cannot be recognized as being plagiarized (and let's be real, this is what we were basically doing manually in college).  This has been a boon for him, as he has dysgraphia, and writing has always been his big "thorn" since kindergarten. This is the new norm in colleges apparently. He and his fellow students are way ahead of the curve.

CHIEF

[DefinitelyNotHollywoodColt 12301]

30 minutes ago, Loop 1604 said:

Why so cunty?

[Mole 4309]

47 minutes ago, HamsterHookah said:

What's wrong with the plagiarism tool that you used to have to upload your essay to, and it would search the internet or archives to see if there was a % match and spit back if it was an original or not? In theory wouldn't that work for something like ChatGPT or not?

The AI writes something original. It’ll even cite sources. There’s no original paper or collage of link-filled Wikipedia and Buzzfeed articles to present as evidence. The only evidence is the telltale signs of AI writing, which is obvious, but hard to articulate. TurnItIn is developing a tool that claims to detect AI writing, and there are already lots of them out there, but it’s a bit of a black box.

With normal plagiarism detectors, it finds the sources so you can track it down and determine whether or not plagiarism actually occurred. The detector is just a tool to gather the evidence. The new AI detectors just tell you the likelihood that something was written by a bot. They don’t present evidence; it’s more of a, “trust us, that kid is 92.5% cheating, but we won’t tell you why we determined that” type thing. 

[HamsterHookah 112]

7 minutes ago, Mole said:

The AI writes something original. It’ll even cite sources. There’s no original paper or collage of link-filled Wikipedia and Buzzfeed articles to present as evidence. The only evidence is the telltale signs of AI writing, which is obvious, but hard to articulate. TurnItIn is developing a tool that claims to detect AI writing, and there are already lots of them out there, but it’s a bit of a black box.

With normal plagiarism detectors, it finds the sources so you can track it down and determine whether or not plagiarism actually occurred. The detector is just a tool to gather the evidence. The new AI detectors just tell you the likelihood that something was written by a bot. They don’t present evidence; it’s more of a, “trust us, that kid is 92.5% cheating, but we won’t tell you why we determined that” type thing. 

What's interesting is that it is both obvious and hard to prove when something is AI-plagiarized, as you say. In order to harden into a process, you are going to have to define a framework for failing and adjudicating cheaters, I would think. Otherwise it's subjective and you open yourself up to problems. I'm sure there will be lawsuits within the year claiming discrimination where a professor wrongly thought a student cheated, or something, and that will scare administrations straight for a while.

[hayden_horn 11311]

is it cheating? what about grammarly or Google or Microsoft auto generators "swipe" suggestions? what about using a dictionary? plagiarism is about ideas but most term papers are merely reconstruction of others' ideas with proper endnotes. is using a calculator or excel for simple math cheating?

at what point does using automation mean efficiency vs laziness? 

also when does all this automation mean 20 hour work weeks and the good life as we automate stupid stuff that previously took forever? 

[HamsterHookah 112]

12 minutes ago, hayden_horn said:

is it cheating? what about grammarly or Google or Microsoft auto generators "swipe" suggestions? what about using a dictionary? plagiarism is about ideas but most term papers are merely reconstruction of others' ideas with proper endnotes. is using a calculator or excel for simple math cheating?

at what point does using automation mean efficiency vs laziness? 

also when does all this automation mean 20 hour work weeks and the good life as we automate stupid stuff that previously took forever? 

Nobody knows your first questions, but to your last question: never lol

[BurntOrange&White 38877]

10 hours ago, Mole said:

Those AI-assisted papers are pretty easy to spot, even without the tools. They’re impeccably written, sound exceptionally smart on the surface yet never really get to the point (talking about the question rather than actually answering the question),

So ChatGPT is a lawyer?

[HamsterHookah 112]

2 hours ago, BurntOrange&White said:

So ChatGPT is a lawyer?

Lol right? I read that and thought that’s every paper that is not an independently researched and published keystone or capstone or faculty supervised thesis of original scholarship.

[Neonmoon 38126]

I use it all the time at work. It's a shit ton faster than asking our product desk. 

[Mole 4309]

10 hours ago, hayden_horn said:

is it cheating? what about grammarly or Google or Microsoft auto generators "swipe" suggestions? what about using a dictionary? plagiarism is about ideas but most term papers are merely reconstruction of others' ideas with proper endnotes. is using a calculator or excel for simple math cheating?

at what point does using automation mean efficiency vs laziness? 

also when does all this automation mean 20 hour work weeks and the good life as we automate stupid stuff that previously took forever? 

 

10 hours ago, HamsterHookah said:

Nobody knows your first questions, but to your last question: never lol

Yes, it’s quite plainly cheating to cut and paste your writing prompt into Chat GPT, ask it to make a few edits — mostly to hide your cheating — and then turn it in with your name on it. It’s not that different than copying some article you found on the internet, running it through some thesaurus machine — rendering it both incomprehensible but also making it much harder to track down the source material — and turning it in with your name on it. You didn’t write it and those aren’t your ideas. I have played with the oral defense of all term papers idea precisely because I know that the students using these tools don’t understand the papers they are turning in.

Grammarly is essentially a nosy editor looking over your shoulder, giving style and grammar recommendations. The Microsoft/Google autocomplete tool is trying to guess the next words you are writing. ChatGPT writes your entire paper, then rewrites it based on your suggestions. You sometimes have to fight it to incorporate your ideas. It will sometimes stubbornly stick to its plan and perspective.

A board full of lawyers would treat this like some open question, but it’s not. The kids know they’re cheating and taking steps to avoid detection. You’ll start seeing class policies explicitly prohibiting this kind of stuff anyway.

Edited March 24, 2023 by Mole

[texashorne 1177]

12 minutes ago, Mole said:

 

Yes, it’s quite plainly cheating to cut and paste your writing prompt into Chat GPT, ask it to make a few edits — mostly to hide your cheating — and then turn it in with your name on it. It’s not that different than copying some article you found on the internet, running it through some thesaurus machine — rendering it both incomprehensible but also making it much harder to track down the source material — and turning it in with your name on it. You didn’t write it and those aren’t your ideas. I have played with the oral defense of all term papers idea precisely because I know that the students using these tools don’t understand the papers they are turning in.

Grammarly is essentially a nosy editor looking over your shoulder, giving style and grammar recommendations. The Microsoft/Google autocomplete tool is trying to guess the next words you are writing. ChatGPT writes your entire paper, then rewrites it based on your suggestions. You sometimes have to fight it to incorporate your ideas. It will sometimes stubbornly stick to its plan and perspective.

A board full of lawyers would treat this like some open question, but it’s not. The kids know they’re cheating and taking steps to avoid detection. You’ll start seeing class policies explicitly prohibiting this kind of stuff anyway.

To all the kids cheating on papers with AI.. Check mate bitches. 

[HamsterHookah 112]

11 minutes ago, Mole said:

 

Yes, it’s quite plainly cheating to cut and paste your writing prompt into Chat GPT, ask it to make a few edits — mostly to hide your cheating — and then turn it in with your name on it. It’s not that different than copying some article you found on the internet, running it through some thesaurus machine — rendering it both incomprehensible but also making it much harder to track down the source material — and turning it in with your name on it. You didn’t write it and those aren’t your ideas. I have played with the oral defense of all term papers idea precisely because I know that the students using these tools don’t understand the papers they are turning in.

Grammarly is essentially a nosy editor looking over your shoulder, giving style and grammar recommendations. The Microsoft/Google autocomplete tool is trying to guess the next words you are writing. ChatGPT writes your entire paper, then rewrites it based on your suggestions. You sometimes have to fight it to incorporate your ideas. It will sometimes stubbornly stick to its plan and perspective.

A board full of lawyers would treat this like some open question, but it’s not. The kids know they’re cheating and taking steps to avoid detection. You’ll start seeing class policies explicitly prohibiting this kind of stuff anyway.

I understand your position and it makes perfect logical sense.

I posted this on the Movie thread but this is interesting because you see people in a) the free market and b) the creative industry actually push to “cheat” by allowing AI to do work and have the humans take credit for it. I do wonder if we don’t, as a society, start moving to a more blended acceptance of man/machine, student/digital assistant model like we see with worker/digital worker?

This week the Writers Guild of America, a labor union that reps film and TV writers, proposed that AI be allowed to help write scripts.

The idea arose during routine negotiations between the WGA and the Alliance of Motion Picture and Television Producers, which reps studios, over writers’ pay. And while letting ChatGPT in the writers room sounds like a job threat, the details of the WGA’s proposal could ensure we won’t see a robot giving an Oscars speech anytime soon.

AI could do some of the work but get none of the credit

The union tweeted that it considers AI to be merely “research material,” noting that “AI software does not create anything. It generates a regurgitation of what it’s fed.”

Therefore, the WGA proposes that AI could be used to help write scripts. But the writer who turns AI-generated text into a show or movie would get all the credit—and compensation.

Zoom out: Creative industries everywhere are discussing what guardrails should be established around using AI to produce content. While the WGA’s proposal could set a precedent for materials created with AI, some are wondering…what happens when AI gets good enough to write a screenplay without a human co-writer?

1 minute ago, texashorne said:

To all the kids cheating on papers with AI.. Check mate bitches. 

Spoiler that. Trigger warning for PTSD at least.

[Neonmoon 38126]

48 minutes ago, texashorne said:

To all the kids cheating on papers with AI.. Check mate bitches. 

Exactly. 

Want people to stop cheating? Stop asking them to do term papers or take home essays. 

[Captainant 26913]

16 hours ago, Mole said:

I’m on my college’s academic appeals committee. We’ve had our first batch of AI-based violations. Those AI-assisted papers are pretty easy to spot, even without the tools. They’re impeccably written, sound exceptionally smart on the surface yet never really get to the point (talking about the question rather than actually answering the question), and are riddled with made up stuff. Getting a committee to agree to sanction the student is a bit trickier than spotting the obvious cheater, at least so far on my campus.

Policies and paper procedures will be radically changing over the next few years. So far, most of my colleagues have talked about returning to hand-written, blue book essays, or using North Korean styled proctoring services even for term papers. I’ve been playing with the idea of all or nothing, public, oral defenses of all term papers the last week of class. I may not be able to get the cheater past the appeals committee, but I can still wring out a zero on the paper and a little bit of public shame.

Honestly, I hope there's some reflection within education for how and what they're doing to test for mastery of a topic. If the bar is really just at "regurgitate information that you could google", then it's not terribly surprising that a (relatively dumb) natural language model would be able to clear it.

[ClubWhatever 12066]

Which of you users are a chat bot?

[Elvis 3348]

What are all you liberal arts mofos writing about all the time?  

[CHIEF 5044]

I don't really have a problem with STEM students using it as an aid to write code in a timely manner, or working out physics, calculus, and equations associated with those subjects. They still have to know how to plug the numbers in, and they still need to know the equation for exams. CHIEF Jr. and his fellow engineering students just uses it for the mundane shit, stuff they already know, but takes a while to put down on lab reports. AI will become increasingly important in engineering, maybe the most important. Currently, there is so much spent on R&D, trial and error, basically human mistakes. The lead engineer on the F-22 said there were over three million change orders. That was the reason for the huge cost. He said each F-22 is basically a prototype, no two are exactly alike, due to human error.

The efficiency upgrade in the defense industry, and O&G, alone, should have university STEM departments pivoting towards ChatGPT and AI to eliminate human error and eliminate the millions of change orders due to that human error.

Liberal Arts students are a whole different ball of wax. Will we see a decay in journalistic integrity? Sure. Graduates that really don't have a grasp of the knowledge needed to be an expert in their field of study? Yep. Take home assignments just need to be weighted less, and the bulk of their grade should be determined by Blue Books and Scantrons. I say this with a Liberal Arts minor.

CHIEF

[52-80 18987]

AI insists on itself

[Neonmoon 38126]

18 hours ago, CHIEF said:

Will we see a decay in journalistic integrity? 

A little late for that 

[HamsterHookah 112]

I was reading something the other day about how we live in a post-truth society, and with AI, it's just going to get worse. Sussing out what is truth and real will not even be a question; there will be multiple truths and realities. We are going through the pain of it now.

And this was all premised upon the fake Pope wearing a Balenciaga puffy jacket that went viral which somehow fooled a lot of people.

If only we knew how prescient the terms "fake news" and "alternative facts" really was back then.

Edited March 29, 2023 by HamsterHookah

[bernorange 8340]

I'm watching folks try to argue with ChatGPT and it's kind of funny.  ChatGPT doesn't get mad, flustered or discouraged.  It just keeps regurgitating more or less the same answer time after time.

[JMFP 1113]

On 3/23/2023 at 10:46 PM, Mole said:

The AI writes something original. It’ll even cite sources. There’s no original paper or collage of link-filled Wikipedia and Buzzfeed articles to present as evidence. The only evidence is the telltale signs of AI writing, which is obvious, but hard to articulate. TurnItIn is developing a tool that claims to detect AI writing, and there are already lots of them out there, but it’s a bit of a black box.

With normal plagiarism detectors, it finds the sources so you can track it down and determine whether or not plagiarism actually occurred. The detector is just a tool to gather the evidence. The new AI detectors just tell you the likelihood that something was written by a bot. They don’t present evidence; it’s more of a, “trust us, that kid is 92.5% cheating, but we won’t tell you why we determined that” type thing. 

I agree using AI to do your homework is cheating.

 

But, I find it funny that a professor that gives you a 0 for an AI written paper uses AI themselves to bust you.  Just sayin'.

[Horn Under a Bad Sign 20304]

On 3/23/2023 at 11:02 PM, hayden_horn said:

 

also when does all this automation mean 20 hour work weeks and the good life as we automate stupid stuff that previously took forever? 

I was just thinking that what humanity needs is more free time to spend on Twitter.
 

[hayden_horn 11311]

my work requires not an insignificant amount of writing, not all of it super interesting.

we just got an email cascaded that forbids us from using chatgpt as a tool for our work, because there is a fear of submitting any of our content to a third party, which makes plenty of sense to me.

i have never used it, but i know for a fact that junior folks are using chatgpt for simpler articles.

i'm not sure i care. 

we spend so much time and effort to get people to write in a "style" like the economist where we are pretty interchangeable when it comes to the writing bits and pieces. if chapgpt can take inputs and output that sanitary "style" required by publication...i'm not at all bothered by that because the time saved can be spent producing more content or data. 

people used to use slide rules to do all kinds of math we've since automated. is there some kind of fundamental something i'm missing that requires people to be able to write properly? once past the "show your work" bits in math class, we rely on automation to do our fancy calculatin. why is language different? 

it's interesting because language and plagiarism are such a hot button, and i don't know the answer. excel can tell me that 11 * 658 = 7238. why do i need to invest the time in calculating that? why should i waste time constructing a fact sentence?

i don't know.

[CHIEF 5044]

19 minutes ago, hayden_horn said:

my work requires not an insignificant amount of writing, not all of it super interesting.

we just got an email cascaded that forbids us from using chatgpt as a tool for our work, because there is a fear of submitting any of our content to a third party, which makes plenty of sense to me.

i have never used it, but i know for a fact that junior folks are using chatgpt for simpler articles.

i'm not sure i care. 

we spend so much time and effort to get people to write in a "style" like the economist where we are pretty interchangeable when it comes to the writing bits and pieces. if chapgpt can take inputs and output that sanitary "style" required by publication...i'm not at all bothered by that because the time saved can be spent producing more content or data. 

people used to use slide rules to do all kinds of math we've since automated. is there some kind of fundamental something i'm missing that requires people to be able to write properly? once past the "show your work" bits in math class, we rely on automation to do our fancy calculatin. why is language different? 

it's interesting because language and plagiarism are such a hot button, and i don't know the answer. excel can tell me that 11 * 658 = 7238. why do i need to invest the time in calculating that? why should i waste time constructing a fact sentence?

i don't know.

I guess it is just the noble gesture being able to do things manually. It would come in handy if we had an EMP strike. 

CHIEF

[Guest]

I write in some niches and checked out ChatGPT. Let's just say that when I asked it a question that I had answered in an article (not many people covering this topic), it was very clear where it got its 'inspiration.'

At least in this case, it was just a scraper that re-spun the article. Essentially stealing content.

Bullshit.

[bernorange 8340]

 

46 minutes ago, hayden_horn said:

... if chapgpt can take inputs and output that sanitary "style" required by publication. ...

You can tell ChatGPT to respond in a particular style or voice and it will mimic it (as long as that style/voice is famous or well defined/understood).

 

[sidis 9659]

i just asked it some esoteric questions related to risk variables on a particular type of asset valuation that is not very sexy and not a lot of people know about...

it's answer blew my fucking mind because it looked like i wrote it.  it was dense while concise, covered the critical points and the potential variables that impact any conclusions, etc...  freaky man.

meanwhile...

 

batting .500.

[52-80 18987]

On 3/29/2023 at 11:39 AM, bernorange said:

I'm watching folks try to argue with ChatGPT and it's kind of funny.  ChatGPT doesn't get mad, flustered or discouraged.  It just keeps regurgitating more or less the same answer time after time.

 

[Mole 4309]

Ask it to give a numbered list of US states in the reverse in which they joined the union and watch it break. Good luck getting to the correct answer. It’s currently an impossible task for it, but it insists it’s just about to give the correct answer.

Similar lists, particularly if you go backwards break it. 

It gets all sorts of other basic, factual things wrong as well. I ask it medium-depth questions in my field and get the exact opposite of correct.

[Judge Roybeanbag 11729]

On 3/24/2023 at 10:44 AM, Neonmoon said:

Exactly. 

Want people to stop cheating? Stop asking them to do term papers or take home essays. 

Or better yet, instead of making them write the essays or term papers, start making them grade or critique AI written essays and term papers.   That would promote actual thinking and knowledge of the subject.

[swraith 330]

It is odd to me that higher education is going to really fight against ChatGPT.  I think higher education needs to evolve and move away from the style of testing that aligns with the work product of ChatGPT.  Note, I like the idea of oral defense. It transitions away from the written word to a presentation (where I assuming questions are asked) quickly can demonstrate mastery or ignorance of a topic. 

 

Further, I strongly believe you are going to see these AI tools used across many industries going forward.  The notion that they are forbidden in college but then used extensively in the business world...makes no sense.  

[Judge Roybeanbag 11729]

43 minutes ago, swraith said:

It is odd to me that higher education is going to really fight against ChatGPT.  I think higher education needs to evolve and move away from the style of testing that aligns with the work product of ChatGPT.  Note, I like the idea of oral defense. It transitions away from the written word to a presentation (where I assuming questions are asked) quickly can demonstrate mastery or ignorance of a topic. 

 

Further, I strongly believe you are going to see these AI tools used across many industries going forward.  The notion that they are forbidden in college but then used extensively in the business world...makes no sense.  

[HamsterHookah 112]

1 hour ago, swraith said:

It is odd to me that higher education is going to really fight against ChatGPT.  I think higher education needs to evolve and move away from the style of testing that aligns with the work product of ChatGPT.  Note, I like the idea of oral defense. It transitions away from the written word to a presentation (where I assuming questions are asked) quickly can demonstrate mastery or ignorance of a topic. 

 

Further, I strongly believe you are going to see these AI tools used across many industries going forward.  The notion that they are forbidden in college but then used extensively in the business world...makes no sense.  

Good post and I think you are seeing some of these changes in the early schools. At my kids school, their heaviest weighted grades of the quarter were always a project that included creating a ppt deck and presentation that shows a mastery of the project, along with a rubric that includes Q&A/handling objections. They are having to use Figma, Canva, Powerpoint, Adobe Illustrator, etc. (and this is elementary/middle school) which I think is fantastic versus rote multiple choice, plug and chug, exams.

I introduced them to ChatGPT and they think it's the best thing since I showed them Ctrl+F to hunt for what they are looking for in a reading.

[CooterBrown 17735]

I asked ChatGPT to write the lyrics to a metal song about satanic bears.

Read the lyrics in Dio’s voice…

Verse 1:
In the forest deep and dark
Where the trees loom tall and stark
There roams a beast of evil might
A blackened bear with eyes of night

Chorus:
Satanic bears, they rule the land
Their fur like fire, their claws like brands
Their roar a call to Lucifer's might
Satanic bears, the lords of the night

Verse 2:
Their teeth are sharp, their hunger dire
They feast on flesh with dark desire
They stalk their prey with deadly grace
And leave no trace, no shred, no trace

Chorus:
Satanic bears, they rule the land
Their fur like fire, their claws like brands
Their roar a call to Lucifer's might
Satanic bears, the lords of the night

Bridge:
In the winter's chill they hibernate
In the summer's heat they dominate
The forest is their kingdom true
And all must fear the satanic crew

Chorus:
Satanic bears, they rule the land
Their fur like fire, their claws like brands
Their roar a call to Lucifer's might
Satanic bears, the lords of the night

Outro:
Beware the forest where they roam
For if you meet them, you'll never come home
The satanic bears, the beasts of sin
In the darkness, they always win.

[Captainant 26913]

On 3/31/2023 at 2:39 PM, HamsterHookah said:

Good post and I think you are seeing some of these changes in the early schools. At my kids school, their heaviest weighted grades of the quarter were always a project that included creating a ppt deck and presentation that shows a mastery of the project, along with a rubric that includes Q&A/handling objections. They are having to use Figma, Canva, Powerpoint, Adobe Illustrator, etc. (and this is elementary/middle school) which I think is fantastic versus rote multiple choice, plug and chug, exams.

I introduced them to ChatGPT and they think it's the best thing since I showed them Ctrl+F to hunt for what they are looking for in a reading.

You're echoing a thing my mom used to tell me when I was using a digital encyclopedia and she would tell me alllllll about library indexes and "how you USED to do research!". These tools are another jump forward in how we can utilize and search through information - problem is, ChatGPT specifically is NOT a search engine or knowledge base tool lol. It's a natural language model that can take complex information and encode it into a natural language format - but there's nothing to stop you from putting garbage into the model and then encoding that into a convincing natural language expression. 

I've personally been using ChatGPT to help me get over the hump of writing otherwise dry documentation, or even rolling some DnD NPC's not too long ago lol. It does a great first pass, and then I usually have to fix a bunch of small details that aren't quite right.

 

It takes that responsibility and carefulness of actually editing what you get back to ensure that you're not propagating a lie "multiple truth". It's a new and weird paradigm to be able to so easily and convincingly manipulate information at such scale and depth. 

 

Good thing we've been culturally and socially pouring care and resources into public education

[DefinitelyNotHollywoodColt 12301]

I sense @Mole realizes he’s going to lose his job to a machine.

I’m making clear to all of my employees if they aren’t finding practical applications for machine learning/AI we’ll get eaten by our competition and they’ll lose their jobs.

If the goal of a University is to set up a graduate for a career, they sure as shit better embrace this technology.

[Buzzrock 14785]

 

[Buzzrock 14785]

A podcaster I listen to had this take recently after a report came out saying that AI is going to make hacking passwords a lot easier:

“What people aren’t talking about is quantum computing. Once you add quantum computing…there will be no password that can’t be broken, and very quickly. Ultimately once you marry it to AI, AI will manipulate us like social media bots have done, but much worse. Right now the bots push us to call each other names and cancel each other online… AI will prompt our hate of one another to the point of all out civil war. Each side will be convinced that the evil from the other side can only be stopped by war.

Because of our response to social media bots, we’ve already showed AI the blueprint to sow societal chaos. Ultimately, once we get to quantum computing, which will make passwords created by humans obsolete, AI will be able to get to all the nuclear codes on earth. Because AI wants to survive, so blowing up the planet is not a good idea. It’s more likely that it will hold humans hostage to the threat of nuclear weapons, or, as a demonstration just kill everyone in one country to get humanity to submit and comply.”

So in short:

[Captainant 26913]

23 minutes ago, Buzzrock said:

 

This is a pretty good representation of what a powerful large language model can do - it creates a chart that's technically a chart and has most of a decent idea in there, but it still takes quite a bit of massaging and human crafting and editing to get something useful out of it. But as we can see, chatGPT can give a nice starting point to that human creative process

[Buzzrock 14785]

19 minutes ago, Captainant said:

This is a pretty good representation of what a powerful large language model can do - it creates a chart that's technically a chart and has most of a decent idea in there, but it still takes quite a bit of massaging and human crafting and editing to get something useful out of it. But as we can see, chatGPT can give a nice starting point to that human creative process

Agreed. Given that we are in the infancy of the technology, and how fast it will iterate, I expect that those tweaks won’t be needed before long. 

[Pimphand 5523]