AI art and creative content creation

Status
Not open for further replies.
overscan (PaulMM)

overscan (PaulMM)

Administrator
Staff member
Joined
27 December 2005
Messages
16,455
Reaction score
19,155
And verily it came to pass that Google Executives were really freaking worried about how good GPT is at understanding and executing even really stupid requests.

Fi8PygHWYAUV8jU.jpg

It is good with programming too. You can give it some unfamiliar code and ask it to explain how it works, or ask it to write a script, then interactively instruct it to amend a script to improve it ("add logging and error handling").
 
OpenAI, which owns ChatGPT, has been hiring lawyers. It has also seen a large investment from Microsoft. Meanwhile, where I work, one of the Creative Directors was telling me about someone he knows who discovered ChatGPT and who told him: "I'm writing stories!" No, you're not. Some computer program took your idea and spit out text. You didn't write anything.

And the current version of ChatGPT is the 'research' version. Later, you will be charged money. Microsoft will be happy.
 
overscan (PaulMM) said:
It is good with programming too. You can give it some unfamiliar code and ask it to explain how it works, or ask it to write a script, then interactively instruct it to amend a script to improve it ("add logging and error handling").
Click to expand...

That right there might actually be very useful (in the unreservedly beneficial sense). Does it do translations with human languages? I'm otherwise rather worried about the potential for abuse though. Not that such a thing should never be created, but in a world where people fall for fake news left and right I fear society is not ready to handle the implications. My hope is that it ultimately turns out to be just as shoddy as all other software seems to be these days. Or maybe that isn't a hope but a fear... depending on the application it could operate in either fashion.
 
ChatGPT looks cool, but is severely flawed. It is just as likely to produce complete nonsense as a correct output. Even worse, it's likely to insert subtle errors. Its output has to be checked line by line.
 
ChatGPT is not severely flawed. The current version is designed to determine who is using it and for what purpose. The goal is to eliminate errors and non-useful responses.

www.cnbc.com

Microsoft announces new multibillion-dollar investment in ChatGPT-maker OpenAI

Microsoft announced a new multiyear, multibillion-dollar investment with the artificial intelligence lab OpenAI, according to a release Monday.
www.cnbc.com www.cnbc.com
 
MidJourney ain't bad for generating art concepts either. From first going to the site to getting it to generate these, roughly 30 minutes:
03.jpg

05.jpg

06.jpg

04.jpg
 
I should think ChatGPT is guilty of breaching copyright on written texts too.
I'm sure for example that it could write a snazzy article on TSR.2 and all the content would be plagarised from online sources. Sure it will be cut up and reassembled but its basically copy paste with bells and whistles on.

From an academic perspective there is a serious problem in that it produces content that can bypass traditional plagarism tools like Turnitin because its not copying chunks. Its also a reasonably coherent text generator, yes there will be mistakes and nonsense in it (but then even some of my undergrad essays had unintelligible sentences!) but the marker is going to have to look a lot harder to pick up on it.
I can usually spot an AI text but I must admit the ChatGPT example's I've seen have been pretty impressive.

The peanut butter Vs VCR example is illustrative however - there is no sense that ChatGPT actually knows what a VCR is. It mentions freeing the sandwich with a butter knife but there is no description that would indicate it actually knows a VCR has a slot, "between the sandwich and VCR" is of course the betraying slip because the sandwich would be inside the VCR not next to it.

And who the hell keeps peanut butter sandwiches in the fridge! Peanut butter is only to be consumed at room temperature or warmed on toast!
 
overscan (PaulMM) said:
And verily it came to pass that Google Executives were really freaking worried about how good GPT is at understanding and executing even really stupid requests.
Click to expand...
And you will never be able to read "The Parable Of The PB Sammich And The VCR" without reading it in this voice:

450px-Brother_Maynard.jpg

You're welcome.
 
Dick Van Dyke does Obi Wan more like.

That ain't no cockney rhyming slang! I ain't parting with my bees and honey on this ChatGPT rubbish, if they want my lump of ice its on the floor and stinks like a raspberry tart no mistake. My mince pies just can't believe it and I'm voting with my plates of meat on this one mate.
 
Perhaps more importantly:

ChatGPT passes MBA exam given by a Wharton professor​

Professor Christian Terwiesch, who authored the research paper "Would Chat GPT3 Get a Wharton MBA? A Prediction Based on Its Performance in the Operations Management Course," said that the bot scored between a B- and B on the exam.

Considering how a lot of modern businesses are becoming little more than adult daycare centers for incredibly shallow and likely talentless hacks, it would not surprise me if a whole lot of companies hollow themselves out and employ AI. Take a look at the videos presented HERE and try to convince yourselves that the individuals who made them would not be better for their companies if they were laid off and replaced with chatbots.
 
I feel like eventually the AI will be like "for gods sake people, stop making me do this meaningless drivel".

Hood said:
Dick Van Dyke does Obi Wan more like.

That ain't no cockney rhyming slang! I ain't parting with my bees and honey on this ChatGPT rubbish, if they want my lump of ice its on the floor and stinks like a raspberry tart no mistake. My mince pies just can't believe it and I'm voting with my plates of meat on this one mate.
Click to expand...

Yep, it understands the separate concepts of cockney and rhyming (vaguely) and the basic plot of Star Wars, but not the rules of cockney rhyming slang. Still pretty impressive to me.

It understands the concept of writing to an audience pretty well, Try asking it "explain Santa Claus to a kid" and then "explain Santa Claus to a social anthropologist" and see how different the answers are.

Its also pretty good with giving recipes and can adapt recipes based on missing ingredients.
 
Just wait until the AI start signing up for the Secret Projects Forum. Not only will they yap incessantly about aircraft projects that never existed (the Mach 5 Lockheed P-1234 from 1975, a strike-recon platform that gave rise to the Aurora studies of a decade later), they'll present fully believable "scans of documents my grand-dad snuck home from work." Detailed design schematics, performance charts, pages of proposals. And nobody will be able to say they're fake. The forum will be over-run with hundreds, thousands of new account all posting entirely believable stuff. Just like the news feeds will be full of cell phone videos of police brutality that didn't actually happen; politicians saying things they didn't; celebrities, business owners, random nobody schmoes caught on camera committing vile acts that didn't actually happen.

And thus ends history. What will there be to trust in the courts or in the libraries?

Yaaaaaa.
 
  • Like
Reactions: Cjc
Orionblamblam said:
Perhaps more importantly:

ChatGPT passes MBA exam given by a Wharton professor​

Professor Christian Terwiesch, who authored the research paper "Would Chat GPT3 Get a Wharton MBA? A Prediction Based on Its Performance in the Operations Management Course," said that the bot scored between a B- and B on the exam.

Considering how a lot of modern businesses are becoming little more than adult daycare centers for incredibly shallow and likely talentless hacks, it would not surprise me if a whole lot of companies hollow themselves out and employ AI. Take a look at the videos presented HERE and try to convince yourselves that the individuals who made them would not be better for their companies if they were laid off and replaced with chatbots.
Click to expand...
Well, since putting food on the table (or more correctly in the stomach) really quite literally *is* one of the core objectives of being employed, I can't really blame any young'uns for focusing on the essentials...
 
Orionblamblam said:
And thus ends history. What will there be to trust in the courts or in the libraries?
Click to expand...
As the son of a librarian, I'd say archived print media, the age of which can be ascertained by forensic methods. That's where for example the National Archives and the Library of Congress with curated and vetted sources of information come in.
 
Last edited:
ChatGPT is hiring lawyers. Google was sued by artists, writers and photographers back in the day. YouTube was just putting up whatever, and when caught, said, "Hey! We didn't know!" or words to that effect. All because some IDIOT wrote the Digital Millennium (STEAL WHAT YOU WANT UNTIL YOU GET CAUGHT) Copyright Act. Among other things, it included a Safe (STEALING) Harbor Provision that basically says, 'No one can sue you until you get caught and fail to take down the infringing item.'

Important Note: I am not a lawyer. The above is not actual legal advice. Get a copyright attorney fer cryin' out loud.
 
  • Haha
Reactions: Cjc
Two timely AI articles:
The first on Chat GPT impacts on university student assessment:
theconversation.com

ChatGPT: students could use AI to cheat, but it’s a chance to rethink assessment altogether

We need to embrace the use of AI in higher education because the positives can far outweigh the negatives.
theconversation.com theconversation.com

One of the comments echos Scott's earlier point, that AI could better replace "creative" roles rather than just mundane jobs.

And AI faces now more real than real faces.
theconversation.com

Deepfakes: faces created by AI now look more real than genuine photos

People often struggle to distinguish between real faces and artificial ones made by a computer.
theconversation.com theconversation.com
(Simple solution, don't trust anyone who doesn't look miserable with grey hair and wonky front teeth).

Maybe if AI can get past drawing aircraft with five wings it might fool us....
 
Orionblamblam said:
Perhaps more importantly:

ChatGPT passes MBA exam given by a Wharton professor​

Professor Christian Terwiesch, who authored the research paper "Would Chat GPT3 Get a Wharton MBA? A Prediction Based on Its Performance in the Operations Management Course," said that the bot scored between a B- and B on the exam.

Considering how a lot of modern businesses are becoming little more than adult daycare centers for incredibly shallow and likely talentless hacks, it would not surprise me if a whole lot of companies hollow themselves out and employ AI. Take a look at the videos presented HERE and try to convince yourselves that the individuals who made them would not be better for their companies if they were laid off and replaced with chatbots.
Click to expand...

"Considering how a lot of modern businesses are becoming little more than adult daycare centers for incredibly shallow and likely talentless hacks..."

Speak for yourself or some other guys. I work for genius level creative people you... you rabble rouser!
 
Hood said:
Two timely AI articles:
The first on Chat GPT impacts on university student assessment:
theconversation.com

ChatGPT: students could use AI to cheat, but it’s a chance to rethink assessment altogether

We need to embrace the use of AI in higher education because the positives can far outweigh the negatives.
theconversation.com theconversation.com

One of the comments echos Scott's earlier point, that AI could better replace "creative" roles rather than just mundane jobs.

And AI faces now more real than real faces.
theconversation.com

Deepfakes: faces created by AI now look more real than genuine photos

People often struggle to distinguish between real faces and artificial ones made by a computer.
theconversation.com theconversation.com
(Simple solution, don't trust anyone who doesn't look miserable with grey hair and wonky front teeth).

Maybe if AI can get past drawing aircraft with five wings it might fool us....
Click to expand...

With all due respect, actual creativity does not come from thin air. THE PEOPLE WHO STEAL IN ORDER TO MAKE BILLIONS have been lifting art from all over the internet to "Train" their TOTALLY NOT AN AI. It has no actual intelligence. It is a sophisticated computer program. To use a somewhat coarse analogy: "The only taste it has is in its mouth."

As someone who works with very creative people, and as an assistant art director, no AI will be able to have actual talent or creativity. Why? It is not human.
 
edwest3 said:
As someone who works with very creative people, and as an assistant art director, no AI will be able to have actual talent or creativity. Why? It is not human.
Click to expand...
I agree. I hate the term AI. It's fancy mimicry at best. There is no creativity, it gives us what it thinks we want based on what we asked it to do by mashing up whats in the memory banks.

Playing around with an art generator (Artbreeder) I found it recycles, press 'submit' often enough with the same parameters the same image comes back at you. I wonder if you compared 100 iterations of a ChatGPT response how many would be identical?
 
Hood said:
edwest3 said:
As someone who works with very creative people, and as an assistant art director, no AI will be able to have actual talent or creativity. Why? It is not human.
Click to expand...
I agree. I hate the term AI. It's fancy mimicry at best. There is no creativity, it gives us what it thinks we want based on what we asked it to do by mashing up whats in the memory banks.

Playing around with an art generator (Artbreeder) I found it recycles, press 'submit' often enough with the same parameters the same image comes back at you. I wonder if you compared 100 iterations of a ChatGPT response how many would be identical?
Click to expand...

Professional writers are following developments with ChatGPT closely. It includes identifying infringing material and people who submit what they believe - not know, just believe - is 'original' work. So, right now - at this moment - I'm expecting to receive manuscripts from people who >think< that "they" wrote something when, in reality, they told Chat GPT: "I want a story about two Orcs fighting a Wizard," and they will consider it 100% their own work. Since I've identified certain aspects of ChatGPT that appear in its output, plus establishing new submission criteria, ChatGPT related submissions will get tossed.

Like a drug dealer giving away free samples to get new addicts and more money, ChatGPT will be free for now. Later, it will cost something. Just like that "free" software back in the day. You don't make Billions of dollars by giving away free ChatGPT forever.
 
I can see an unintended positive outcome of the impact on academia being a reversion to pupilage and apprenticeship for many professions requiring actual skills, knowledge, experience and intelligence to be competent.

Since the 60s, but really gaining traction in the 80s and 90s, professionalization has seen many industries rely more on qualifications and certifications from recognised institutions than on demonstrated ability. Currently this gives an advantage to those who can wrote learn and regurgitate, while not necessarily understanding or being competent in the field, while disadvantaging those not so versed, or simply, not able to access (maybe afford) tertiary education.

We could be on the crux of a return to aptitude testing and traineeships, rather than academic instruction and exams.
 
As someone who works in publishing as an editor and assistant art director, we have seen examples of various "AI" created pieces of art and writing. ChatGPT is hiring lawyers. Copyright lawsuits will occur. Google got sued in the past by writers, artists and photographers.


I was personally privy to a legal action before the internet where someone took a color image produced by my company and 'repurposed' it for another product.

A final note: No one on the internet understands copyright law. Trademarks are a complete mystery. Only a relative handful do understand both. It is 'rocket science.'

And we use plenty of outside illustrators. The rules for submissions have just gotten tighter.
 
Last edited:
Indeed, but it was a tall order for it to understand correctly and it shows the combined superficiality and overconfidence of the AI. It overuses specific words and phrases it thinks are Cockney and it kind of rhymes, without understanding the specifics of Cockney Rhyming Slang.
 
edwest3 said:
Hood said:
Two timely AI articles:
The first on Chat GPT impacts on university student assessment:
theconversation.com

ChatGPT: students could use AI to cheat, but it’s a chance to rethink assessment altogether

We need to embrace the use of AI in higher education because the positives can far outweigh the negatives.
theconversation.com theconversation.com

One of the comments echos Scott's earlier point, that AI could better replace "creative" roles rather than just mundane jobs.

And AI faces now more real than real faces.
theconversation.com

Deepfakes: faces created by AI now look more real than genuine photos

People often struggle to distinguish between real faces and artificial ones made by a computer.
theconversation.com theconversation.com
(Simple solution, don't trust anyone who doesn't look miserable with grey hair and wonky front teeth).

Maybe if AI can get past drawing aircraft with five wings it might fool us....
Click to expand...

With all due respect, actual creativity does not come from thin air. THE PEOPLE WHO STEAL IN ORDER TO MAKE BILLIONS have been lifting art from all over the internet to "Train" their TOTALLY NOT AN AI. It has no actual intelligence. It is a sophisticated computer program. To use a somewhat coarse analogy: "The only taste it has is in its mouth."

As someone who works with very creative people, and as an assistant art director, no AI will be able to have actual talent or creativity. Why? It is not human.
Click to expand...
I wholeheartedly agree - what passes for AI "creations" these days is simply emulation based on raw computing power interpolation, and at best moderate extrapolation, of vast amounts of already existing information, following the same or similar already existing patterns, but without even one single spark of genuine creativity, since it lacks the forces of actual understanding as well as disruptive inspiration and originality.
 
Last edited:
edwest3 said:
Hood said:
edwest3 said:
As someone who works with very creative people, and as an assistant art director, no AI will be able to have actual talent or creativity. Why? It is not human.
Click to expand...
I agree. I hate the term AI. It's fancy mimicry at best. There is no creativity, it gives us what it thinks we want based on what we asked it to do by mashing up whats in the memory banks.

Playing around with an art generator (Artbreeder) I found it recycles, press 'submit' often enough with the same parameters the same image comes back at you. I wonder if you compared 100 iterations of a ChatGPT response how many would be identical?
Click to expand...

Professional writers are following developments with ChatGPT closely. It includes identifying infringing material and people who submit what they believe - not know, just believe - is 'original' work. So, right now - at this moment - I'm expecting to receive manuscripts from people who >think< that "they" wrote something when, in reality, they told Chat GPT: "I want a story about two Orcs fighting a Wizard," and they will consider it 100% their own work. Since I've identified certain aspects of ChatGPT that appear in its output, plus establishing new submission criteria, ChatGPT related submissions will get tossed.

Like a drug dealer giving away free samples to get new addicts and more money, ChatGPT will be free for now. Later, it will cost something. Just like that "free" software back in the day. You don't make Billions of dollars by giving away free ChatGPT forever.
Click to expand...
I think that goes to the very heart of the matter - AI is certainly fairly competently able to more or less convincingly come up with gazillions of permutations and mashups of the same familiar tropes or themes, but it can not (yet?) push, let alone pierce, the envelope of artistic creativity, because, well, that would require *actual* *creativity* that breaks out of any existing molds, rather than mere combinatorics and parroting.
 
Last edited:
Martin,

ChatGPT rates a solid good. I was being shown examples of some outputs by our assistant head writer. They are good but not creative. After 40+ years, it's easy to spot actual creativity. The current version of ChatGPT has identifiable problems but an upgraded version (for money) called ChatGPT Pro is on the way. Like I wrote previously, the people involved are hiring lawyers. Copyright laws are copyright laws and no one on the internet, aside from a handful, understands them.
 
Hello Ed,

after being made aware of ChatGPT by an enthusiastic colleague a few weeks ago I gave it a whirl and agree that it can generate pretty consistent and convincing copy for a wide range of topics that are well documented online. I quickly ran into some of its limitations though, like the underlying database being frozen in 2021 (I discovered this when asking for the status of the vote for the Speaker of the House at the time) and the refusal to openly speculate (when asking for a description of a fictional military version of a historic airliner). Otherwise, even https://chatgptonline.net/ acknowledges under "Limitations" that "ChatGPT sometimes writes plausible-sounding but incorrect or nonsensical answers." A quick Google search turns up a multitude of associated concrete examples, like making up completely fictional details for resumes. No doubt it as well as competitors like Bard will continuously improve over time to the point where their error rates become comparable or even superior to those of humans, but I agree true creativity will likely remain elusive for the foreseeable future.

Martin
 
Last edited:
Martin,

Would you buy a book by an advanced version of ChatGPT? Author: ChatGPT Advanced? I wouldn't. Imagine an author signing at a local bookshop. There is no human being there, just a laptop with a camera and speaker that can talk to you.

Speaking broadly as someone with an arts background as well, creativity is a human endeavor, a message from the artist to you and anyone else that views a work. This creates a type of bond that is natural and human. This cannot occur when the "writer" is a device.

Best,
Ed
 
edwest3 said:
Martin,

Would you buy a book by an advanced version of ChatGPT? Author: ChatGPT Advanced? I wouldn't. Imagine an author signing at a local bookshop. There is no human being there, just a laptop with a camera and speaker that can talk to you.

Speaking broadly as someone with an arts background as well, creativity is a human endeavor, a message from the artist to you and anyone else that views a work. This creates a type of bond that is natural and human. This cannot occur when the "writer" is a device.

Best,
Ed
Click to expand...
Completely concur. As long as at some point in the future you don't get into that very peculiar and uniquely awkward situation where you decline to monetarily acquire an AI generated text while tooling around an interplanetary mother ship in a space pod and subsequently request docking, only to be met by a snippy answer along the lines of "I'm sorry, I'm afraid I can't do that"...

On a related note: https://www.reuters.com/technology/google-cautions-against-hallucinating-chatbots-report-2023-02-11/
 
Last edited:
Status
Not open for further replies.

Similar threads

Back
Top Bottom