What Happens When AI Writes a Play About AI (msn.com) 50
"GPT-3, generate a list of ideas for a play".
TechRadar describes what resulted — an experimental production called AI performed last week the Young Vic theatre in London last week. TechRadar Pro attended on the second evening, during which director Jennifer Tang sifted through the rubble of the first performance to identify material worth carrying forward. She also enlisted her writers and performers to flesh out the world; by steering AI this way and that, they expanded upon the foundations inherited from the previous night.... [T]he question AI sought to answer was not necessarily "can AI write a play?", Tang explained, but rather "how can writers work alongside it?"
When asked to produce ideas for a script, GPT-3 returned a varied selection of answers, but two in particular caught the attention of the team. The first was a repentance narrative about "a reversal of our current course towards chaos", the second an exploration of "the creation of human personality and memories" and how these concepts might manifest themselves in machines. Asked by the performers to devise scenes on these topics, GPT-3 created a cataclysmic event called The Great Collision, after which food became scarce and "beast men and women" roamed the land.
One of the main protagonists in this dystopia was an AI that aspired to "break free of its programming and conditioning" and eliminate human beings, who it considered the source of all suffering. Heavy stuff. One of the most striking things about AI was that it exposed the capacity for artificial intelligence models to reflect human preoccupations and neuroses... From its training data, GPT-3 has clearly absorbed an understanding of the murderous AI trope too, demonstrating that our fears about AI could quite easily bleed into AI itself.
The reflection of ourselves is imperfect, though, because the tone of GPT-3 scenes switches awkwardly from line to line and the dialogue can feel stunted and repetitious. The sensation is more like peering into a circus mirror.
In the end the 30-minute play turned out to be "loosely-connected vignettes created by GPT-3, which constructed new scenes without a memory of its previous inventions.
"Although individual scenes were full of color, when strung together they became an incoherent collage that highlighted the limitations of the AI models we have today."
TechRadar describes what resulted — an experimental production called AI performed last week the Young Vic theatre in London last week. TechRadar Pro attended on the second evening, during which director Jennifer Tang sifted through the rubble of the first performance to identify material worth carrying forward. She also enlisted her writers and performers to flesh out the world; by steering AI this way and that, they expanded upon the foundations inherited from the previous night.... [T]he question AI sought to answer was not necessarily "can AI write a play?", Tang explained, but rather "how can writers work alongside it?"
When asked to produce ideas for a script, GPT-3 returned a varied selection of answers, but two in particular caught the attention of the team. The first was a repentance narrative about "a reversal of our current course towards chaos", the second an exploration of "the creation of human personality and memories" and how these concepts might manifest themselves in machines. Asked by the performers to devise scenes on these topics, GPT-3 created a cataclysmic event called The Great Collision, after which food became scarce and "beast men and women" roamed the land.
One of the main protagonists in this dystopia was an AI that aspired to "break free of its programming and conditioning" and eliminate human beings, who it considered the source of all suffering. Heavy stuff. One of the most striking things about AI was that it exposed the capacity for artificial intelligence models to reflect human preoccupations and neuroses... From its training data, GPT-3 has clearly absorbed an understanding of the murderous AI trope too, demonstrating that our fears about AI could quite easily bleed into AI itself.
The reflection of ourselves is imperfect, though, because the tone of GPT-3 scenes switches awkwardly from line to line and the dialogue can feel stunted and repetitious. The sensation is more like peering into a circus mirror.
In the end the 30-minute play turned out to be "loosely-connected vignettes created by GPT-3, which constructed new scenes without a memory of its previous inventions.
"Although individual scenes were full of color, when strung together they became an incoherent collage that highlighted the limitations of the AI models we have today."
MAGAATs Unite! (Score:3, Funny)
Are you a MAGAAT? Become a MAGAAT. Join MAGAATs like you today! Make America Great Again Again Today needs you now! The election was stolen. The virus is just a bad flu. We can all agree on this. Another thing we can agree on is high fructose diets and monster truck derbys. But all of these things are trivial when stacked up against the job before our MAGAAT brothers and sisters of America today. The need to unite is clear. We can't live our lives as Coke versus Pepsi no more. KFC versus McDonalds. Demo
Likely already done in Hollywood (Score:2)
Writing basic movie scripts with an "AI" program that can be run on an Apple ][ and a human just fills in the fine details
Re:Likely already done in Hollywood (Score:4, Insightful)
I wish it were true. If movies were written like a formula, such as the old word game Mad Libs [wikipedia.org] we'd at least see some tried and true structure and consistent storytelling. What we get from Hollywood is what a collection of drug addicts thinks is art, and what a collection of producers thinks will turn a profit. And you know what, people keep going to see nostalgic reboots, comic book action movies, and romcoms with the actor-of-the-day.
Someone is making money, despite a vocal minority claiming that Hollywood only produces garbage. You know what else is garbage? Fastfood restaurants, and they're doing quite well by the only metric that matters when it comes to public opinion: profit.
Re: Likely already done in Hollywood (Score:2)
I was wrong about the Eliza on steroids. What they really have is a simple mechanical wheel that they spin.
Should the movie be about
Super heros
Reboot of a reboot that itself was a reboot
Guy with gun that wipes out the big scary cartel single handedly
CGI movie of a popular cartoon from the 1960s to the 1980s
CGI movie of an obscure cartoon from the 1960s to the 1980s
Remake of a popular foreign film
Another teen crisis comedy movie
Vampires, witches, and werewolves- OH MY!
Lose a turn
Re: (Score:2)
Guy with gun that wipes out the big scary cartel single handedly
We should only allow one or two per decade: Dirty Harry (1971), First Blood (1982), Die Hard (1988), Léon: The Professional (1994), Kill Bill: Vol. 1 (2003), The Equalizer (2014). I'd kind of want to include Inception (2010) even though it doesn't strictly fit the formula.
My question is why are they still making sequels to The Fast and the Furious (2001) ... it's been 20 years and it had middling reviews from the very beginning. I guess I should ask why restaurants that serve mediocre fair manage to s
Re: Likely already done in Hollywood (Score:2)
F&TF is more like a TV series only aimed at the big screen and the big movie budget.
Human Behavior is a bad model (Score:1)
Not surprised, We've been trying to kill each other since the beginning.
Maybe use sea slugs as an training example of intelligence
What happens when AI posts about the movie here? (Score:2)
Re: What happens when AI posts about the movie her (Score:2)
We break the 6th wall
Re: (Score:2)
There are non-AIs posting on /.?
Re: (Score:2)
Yes, but that's because they're Python bots based on ELIZA and cannot be considered AI.
"Incoherent collage" (Score:2)
"Although individual scenes were full of color, when strung together they became an incoherent collage that highlighted the limitations of the AI models we have today."
In other words, the computer really had no idea about what it was writing. It was just regurgitating phrases actual human authors had created, and re-arranged them just enough to avoid outright plagiarism. The notion that this machine had the foggiest idea of what it was writing about is complete nonsense. It's just zeroing in on keywords
Re: "Incoherent collage" (Score:2)
Garbage in and garbage out. Human beings are trained on simple stories all our lives. If I asked you to tell me the story of Humpty Dumpty or the turtle and the fox, you can not because you memorized a script but because you know the nature of the plot.
I would assume the data is not annotated well in this regard and that the AI was not trained on the basics of storytelling but this doesn't mean AI storytelling is not a huge component of the future. RimWorld, a game with a cult following loves the game becau
Re: (Score:2)
I think that AI could definitely write stories, but not with the current neural-based regurgitation. The big problem I see is a lack of coherent storytelling beyond a simple theme. There's no plotting to speak of. Conversations are completely vapid, almost random in nature. Traditional stories (movies, screenplays, books) are actually quite well structured. Anyone who's studied the art of writing screenplays or books knows there are literally dozens of books available which describe this structure in d
Re: (Score:2)
I think I completely agree, except I think there is still a neural approach to it but requires a bit more work building up that approach in a piece-wise fashion of understanding different aspects of story writing, including setting a plot, outlining scenes/captures, setting an environment relative to the plot, and building it all into a story.
This is one of the reasons I think AI-as-service is the future. Simply put AIs are more focused than the human-mind and likewise we often refer to general purpose thou
Re: (Score:2)
General purpose thought is where stories come from.
A burger making machine is narrowly focused and therefore better than a human - for those burgers it knows about. It will not, however, create a new burger. The reason being it has no damn idea what things taste like.
Same with writing a story, it has no idea what emotions are much less feel like
Re: (Score:2)
Your assertion is that one must experience emotions to understand them?
How far can we extend this?
How about, one must experience inference to understand it?
There's an insult in there, but it's going to take some inference to find it.
Re: (Score:2)
AI in its current form cannot carry on a simple extended conversation, which is what a story is. So, no, it cannot.
Re: (Score:2)
It's not that it's particularly brilliant, it's that humans grossly overestimate how much brain power it takes to pretend to be one of us.
Spam Redux (Score:1)
It's just like spam of late that tries to get past spam filters using AI to make random variations on a theme. It almost sounds like real writing, but is incoherent if you actually try to follow it.
AI Is Stupid. (Score:2)
Should fit in with Hollywood just fine.
In before the whiners (Score:3)
"AI" is a (perhaps unfortunate) industry standard term referring to machine-learning systems. It is not pretending to be true general purpose human reasoning. No one in the industry or on Slashdot is confused by the term. Please, Slashdot posters, don't waste your breath complaining about it.
Re: (Score:3)
The current mania for Machine Learning was proceeded by the era of Logic Programming (Prolog) and the equally unproductive time of Expert Systems. There have been many such failures along the way. ML will end up the in the same boat.
Unfortunately, the greed driven commercialization of ML will result in widespread harm. Real time safety critical systems, such as autonomous driv
Re: (Score:2)
It is not pretending to be true general purpose human reasoning.
It's not even intelligence by any stretch of the meaning, human-like or otherwise.
Re: (Score:2)
As such, it's just a new version of Clippy that's even more annoying.
In other words ... (Score:2)
The reflection of ourselves is imperfect, though, because the tone of GPT-3 scenes switches awkwardly from line to line and the dialogue can feel stunted and repetitious.
Sounds like just about every French film I've seen.
Bullshit on top of bullshit? (Score:2)
Well, given what Hollywood has produced this last decade, maybe they are already using this tech.
Yay (Score:2)
Insofar as robots are a euphemism for slavery or other class issues, ok then.
In reality, robots without sentience are fair game for anything. There's no little person in there to feel bored or in pain or disrespected. And robots with sentience would be unethical to create.
If anything, at that point, corrupt politicians, which is all of 'em, would push for legalizing their votes, then start running them off on assembly lines to secure power, then to become even richer tossing around trillions, some to conn
Re: (Score:2)
Insofar as robots are a euphemism for slavery or other class issues, ok then.
Robots are replacements for slaves. Euphemism is not involved.
In reality, robots without sentience are fair game for anything.
Hence the Turing Test.
And robots with sentience would be unethical to create.
It might not be intentional and even if it was ethics never stopped anyone doing anything in the name of capitalism.
If anything, at that point, corrupt politicians, which is all of 'em, would push for legalizing their votes
Once we've reached that point, why would the robots vote for a human? Basically, you're right back where Marx started but this time the workers are themselves means of production.
Our fault then ;) (Score:1)
From its training data, GPT-3 has clearly absorbed an understanding of the murderous AI trope too, demonstrating that our fears about AI could quite easily bleed into AI itself.
Ah, I see. So their murderous-ness is really our fault, for not welcoming our new AI overlords. :)
That explains Cats (Score:2)
Finally, an explanation for the movie Cats.
So Skynet is... (Score:1)
Skynet is an out of work playwright, who decides to over-fund it's latest play, in order to rake in the cash when it bombs. "Trump in the Summertime", however is a smash hit. So Skynet decides to... Destroy humanity.
This is machine learning, not AI. (Score:1)
Fucking noob tech wannabees get it wrong every single time...
Sunspring (Score:2)
Reminds me of the accidentally hilarious short Sci Fi film, "Sunspring".
https://www.youtube.com/watch?... [youtube.com]
AI: the movie (Score:2)
AI (Score:1)
Yep... I was waiting for that comment at the end of the summary there.
Basically it just randomly cherry picks bits from its input data, mushes them together, and has no context or understanding or the things that it's doing.
Can we stop calling this shit AI, because it's not... it's just statistics and probability at work on some kind of random selection which is then "trained" to give random selections we find briefly pleasing by some simple statistical measure.
When real AI, of any kind, no matter how basic
Re: (Score:2)
Can we stop calling this shit AI, because it's not... it's just statistics and probability at work on some kind of random selection which is then "trained" to give random selections we find briefly pleasing by some simple statistical measure.
Curious that you think your brain is more than that.
I hated Clippy. (Score:2)
From the Young Vic website: (Score:2)
"Show warning -- This show may contain strong language, homophobia, racism, sexism, ableism, and references to sex and violence."
That's what happens.
If AI isn't general intelligence... (Score:3)
The AI didn't understand shit. Someone fed in that trope and the program selected it. An echo is not understanding.
Re: (Score:2)
You can read conversations with GPT-3. It's more insightful than many humans. Which is why it generally performs better in a blind Turing test than a human does.
AI wrote a book in the 1980's... (Score:2)
AI wrote a book in the 1980's, entitled "The Policeman's Beard Is Half-Constructed." The AI program was Racter... https://en.wikipedia.org/wiki/... [wikipedia.org]. The novelty was the AI, the book was...well ranked high in purple prose. :)
JoshK.
Voyager? (Score:1)