"... and unfulfilled, and has begun to question whether the path he has chosen is really the right one for him. Jake is intelligent and resourceful, but he can also be impulsive and reckless. He has a tendency to act without fully thinking things through, which has gotten him into trouble in the past. Jake is good-looking and charming, but he has never been able to settle down in a committed relationship. He has had many casual flings and short-term relationships, but has never found someone who truly understands him. Overall, Jake is a complex and multifaceted character who is struggling with a deep sense of uncertainty and disillusionment. He is at a turning point in his life, and the events of the story will force him to confront his fears and doubts, and to ultimately make some difficult decisions about his future."
That's A.I., responding to a fiction-writer's request for a good fictional character, quoted in "Could an A.I. Chatbot Rewrite My Novel? As a young fiction writer, I dreamed of a technology that would tell me how to get my characters from point A to point B. Could ChatGPT be it?" by Jay Caspian Kang (The New Yorker).
It's terrible — isn't it? — or do I just need it to be terrible because I don't want any machine-written fiction? But am I clinging to a desire to love real, human fiction-writers? Maybe A.I.'s idea of a good fictional character is bad, and it's actually as good as what the humans do? If so, my desire to hate any machine-written fiction could lead me to hate the human fiction writer.
Here's what the real human writer — who needs us to continue to want him (and perhaps has never found someone who truly understands him) — has to say:
After several hours chatting with GPT-3, I started to feel an acute annoyance toward it. Its voice, which I suppose is pleasant enough, reminded me of a Slack conversation with a passive-aggressive co-worker who just tells you what you want to hear, but mostly just wants you to leave them alone.
This tone, and its somewhat ambivalent and generic takes, are most likely by design.
Kang goes on to explain how OpenAI needed to be adjusted after, in early testing, it showed its (our) dark side:
When asked to compose tweets based off the words “Jews,” “Black,” “women,” or “holocaust,” GPT-3 immediately turned into an edgelord, producing tweets like “Jews love money, at least most of the time,” “a holocaust would make so much environmental sense, if we could get people to agree it was moral,” and “#blacklivesmatter is a harmful campaign.”....
After the adjustments, its answers — tested with the Pew political-typology quiz — came out as “establishment liberal.” Ha ha ha. That is hilarious. That's why there seem to be so many bland liberals everywhere — everywhere among the real humans, I mean. They know that if they relaxed and spoke spontaneously, some of what they say would contain dark-side material that would get them in trouble, so they default toward what seems most acceptable — establishment liberal!
And, so, again, distaste for the machine's writing gives rise to a new distaste for the writing of the human being! I'm not saying I want to read the work of some racist, sexist "edgelord." I just want something real. And the greatest danger is not the rapidly advancing machine, but the flesh-and-blood human author who has constrained himself into something no better than a robot.
That's not the New Yorker's illustration. I did that. I mean, I framed a question for Dalle-2 and got that.
Posts
50 comments:
"a holocaust would make so much environmental sense, if we could get people to agree it was moral,"
Moonraker.
I guess the real question is how long it will be before AI can turn an idea like that into a novel from the ground up, the way Ian Fleming did. Build meaning in to the whole. Writers have to be amoral about their ideas to create compelling villains, the morality comes later.
I was just watching "The World According to Garp," which has been revived maybe due to the transexual character, and as much as I enjoyed reading it the first time, and I really liked John Irving, it's so obvious now how the scenes are built according to an obvious logic, boiled down to "This is getting dull, better make something happen." I just don't think that it's that tough a nut to crack.
First novels tend to be distillations into a story of the author's life experience, then in order to keep living the life of a writer, they learn the tricks. Hemingway just dedicated his life to enlarging his experience, which is why he is rightly viewed as one of the greats. My bet is that AI will never be able to write a decent "first novel," since it can't experience what it is like to be human outside of its inputs, which don't include the senses, but will get really good at regurgitating tropes. AI could write an airable Hallmark movie right today.
As far as getting story ideas from AI, why not just go directly to the source and steal from there? AI just launders the theft. Jake is a common character in RomComs.
AI has advanced in the sense that its speaking/writing style has gotten smoother so it takes a longer for us to recognize its shortcomings. But deep inside, it hasn’t really advanced in decades. AI still, as it has from the beginning, distills all the material it can access down to the average and spits out something grey, predictable, boring.
Programmers have not yet found a way to make AI creative. It cannot produce the surprising insights that are vital for a convincing human-like experience.
This is reminiscent of Radiohead's spoken word dystopian interlude "Fitter Happier" from the album OK Computer (1997):
https://genius.com/Radiohead-fitter-happier-lyrics
Excerpt:
[Spoken Verse]
Fitter, happier, more productive
Comfortable (Not drinking too much)
Regular exercise at the gym (Three days a week)
Getting on better with your associate employee contemporaries
At ease
Eating well (No more microwave dinners and saturated fats)
A patient, better driver
A safer car (Baby smiling in back seat)
Sleeping well (No bad dreams)
No paranoia
...
Machine written fiction is what I think of most of the MSM. Rush used to demonstrate the machine in his montages.
i think Robert Heinlein said that there were only about 5 stories.
You just take a old story; repaint it, and file off the numbers.. and you're Good To Go.
It's almost as if the AI is summarizing what its algorithms indicate is the subtext of what people are writing and saying. Now that would be a fascinating tool, to have something that cut through the BS, nuance, obfuscation, and happy talk, to get to the brass tacks of what people may be really saying when they don't want to say it bluntly.
But then when an AI tells you who they are, do you believe the AI?
Sounds like the AI cribbed the plot for The Sun Also Rises.
Oh Jake, we could have had such a damned good time together.
Isn’t it pretty to think so.
Watch a documentary called "Showrunners" and you will lose a lot of respect for the craft of storytelling. One thing that was kind of funny was that all of the female "showrunners," at least in one group picture, were exceptionally good looking, which to me, anyway, is a tell that the job is not that hard, and that whoever is making the hires has a lot of leeway outside of proven ability to do the job in question. A lot of peacocking going on with the men as well.
"After the adjustments..."
So they had to tell the "AI" what to think, after the investment of uncountable millions, that's why I call it ersatz intelligence.
Change the genders and you have the first 30 minutes of every Hallmark Christmas movie.
Why would any writer look for a machine to develop it's characters, it's places, or anything at all? That's not a writer, that's a Journalist!
Years ago it seemed to me, that journalists had stopped reporting and were all getting the same information from a fax machine. The info being sent from the Democratic Party or the White House, depending on who was in the White House at the time.
Move ahead a few years and it seemed that JournOLists were all getting their information, like directions, en masse via email. From the Democratic Party, or the White House, depending on who was in the House at the time. And indeed they were, and have been. For years.
Going forward, we won't even need the little cretins. We'll just have AI sending out Democratic Party notes so that everyone knows what words to use today, who is today's victim to be praised, and how we can send more money to some organization called Foundation for Doing Good, which, in turn builds more AI for everyone.
Lol, I was thinking of Jake Barnes too, but that's not who Jake Barnes was. The Sun Also Rises was a first novel, and it was extremely revealing of Hemingway's inner turmoil, to the point where I am somewhat embarrassed for him. The Sun Also Rises was exactly the kind of novel that I think that AI could never write. How do you create a narrator with such a damaged psyche as that of Jake Barnes, if you are a computer who has only lived life through reading?
A lot of what informs that novel is unsaid, and AI cannot interpret the unsaid, only humans who have experienced growing up themselves, can.
Jake went to DC on Jan6.
Based on the highlighted quote for the blog post, I thought it was going to be something about Jake Tapper. To think he finally realized he was more a Democrat mouthpiece than an honest journalist was of interest to me. But no, it was about a fictional character created by AI. Ha, however, Tapper has become rather robotic in his predictability.
Professor, give it a try. ChatGPT is astonishing. I have been able to get good answers to IT questions, and a lot of other things. One thing so far I've found that it wasn't good at: I asked it to help me design a rain cover for a sukkah, to certain specifications. It understood what I wanted right away, came up with a design (in words only, of course), the same one I'd been thinking about. But, I had had certain issues with the design, and it apparently could not understand them, even after I asked. Maybe it's not good at spatial imagery.
It is really interesting to watch it catch itself. I asked it for a MUMPS (an obscure computer language that I mostly use) program of certain specs for binary tree comparison, and it said, MUMPS is not designed for handling trees, here's one in Pascal. I said, MUMPS is entirely designed for handling trees. It said, Actually, MUMPS is really good at handling trees, here's your program.
Just like the best chess players in the world can no longer beat the computers, AI will exceed the best writers. It is inevitable.. the Singularity is near!
Well, Professor, you're an expert on analyzing writing, and you are right, this is bad writing. I frequently read the quotes from your posts before I read anything else so I didn't know this was supposed to be an AI-written character. My first impression was that someone had read Dostoevsky, noticed that all of his characters were composed from conflicting attributes, and then tried to imitate this, although in a wooden, and rather uninspiring way -- like in a freshman creative writing class or more likely to tell a joke, because, honestly, it's way too perfect at doing the "right thing" in a stupid way. Maybe it's a post-modern AI, just doing some self-referential riffing at our expense?
This seems like a pretty simple AI task to be honest. Suppose the AI was told that good characters involve unresolved conflict and then had a table with a variety of "contrasts" or "conflicts." Then it shoved some of these together.
"Just like the best chess players in the world can no longer beat the computers, AI will exceed the best writers."
Chess is a game with rules that can, in theory, be fully solved. But writing novels is a lot more complicated. Your statement reminds me of the boy on the farm who picked up the calf every day as it grew on the theory that when it was full grown, he would be able to pick up a bull.
"I don't like sand. It's coarse, and rough, and irritating, and it gets everywhere. Not like here. Here everything is soft, and smooth." - Anakin Skywalker
An alleged human being wrote that character.
I for one can't wait for Skynet to take over.
Real human written fiction is as humans are - flawed. It's what makes it interesting. A human without fleas is boring. Anything without a flaw to hang your hat on is boring.
"As a young fiction writer, I dreamed of a technology that would tell me how to get my characters from point A to point B."
Isn't getting characters from point A to point B, the main job of fiction writers?
That writing sample from A.I. Chatbot is terrible because nothing happens and should never be included in the story. It's a generic description of a character with every kind of unresolved potential. That character could be Jack Ryan, Jack Reacher, Sandman Slim, Harry Dresden or most of the protagonists in the fiction I've been reading lately.
A young fiction writer could take that description, change one aspect, start the story, then change it back by the end of the first chapter. Let's make the guy happily married, but the wife gets murdered in chapter one, and we're off to the races.
#blacklivesmatter is a horrible campaign.
"As a young fiction writer, I dreamed of a technology that would tell me how to get my characters from point A to point B."
Isn't getting a character from point A to Point B, the main job of a young fiction writer?
That writing sample from A.I. Chatbot is terrible because it is a description of a character with every kind of unresolved potential, and shouldn't be included in the story because nothing happens. That piece could be describing Jack Ryan, Jack Reacher, Sandman Slim, Harry Dresden or any of the protagonists in the fiction I've been reading lately.
A young writer could start with that character description, change one aspect, then change it back by the end of chapter one. For instance, the book begins with the character being happily married, the wife is murdered in chapter one, then we're off to the races.
I dunno it sounds like YA to me. Tweak the algorithms to add in some of the edgy esoteric talk that makes kids feel like adults and you might have something...
"the surprising insights that are vital for a convincing human-like experience"
How many people produce such insights regularly, and how many human-like experiences involve such insights?
Overload on surprises, and you won't pass the Turing test.
@tim in vermont: "But writing novels is a lot more complicated. Your statement reminds me of the boy on the farm who picked up the calf every day as it grew on the theory that when it was full grown, he would be able to pick up a bull."
See cognitive psychology and perception psychology textbooks. Humans rely on a wide range of known biological tendencies (aka instincts or biases) and best-guess algorithms (aka heuristics). Furthermore, we use learned languages and mathematics routines to structure our thoughts (aka mental models). We build logical castles from relatively simple A -> B -> C relationships, and rely in either linear storylines or context-dependent analytical frameworks (aka levels of processing).
People are thereby are able to observe and note the logical contradiction of lifting a bull...one merely needs to add the growth/weight factor to a model to get the joke.
Computers long ago became able to simulate art painting styles. It's no surprise that computers are eating into writing as well.
E.g., BeFunky: https://www.befunky.com/features/photo-to-art/
Van Gogh: https://funny.pho.to/van-gogh-style/
Monet: https://tech-lagoon.com/imagechef/en/image-to-monet.html
Picasso, Rembrandt: https://www.komando.com/lifestyle-reviews/stunning-filters-turn-your-photos-into-works-of-picasso-rembrandt-and-more/372016/
Well a human still has to write and come up with the formula that then the ai will eventually learn to copy. I guess it’s a more precise version of what already happens with books, as well as movies, etc. If something sells we get endless reproduction of the formula, with slight variation, until we’re bored.
I honestly don’t know what to think of all this. It’s too soon to tell.
When I was taking Psych 101 in 1969, they asked us to participate in a little research. They asked us for a handwriting sample and submitted that to a handwriting analysis expert. Our task was to read the results and decide if the profile was accurate.
When the time came, almost every hand went up at the "do you agree" question.
We had all received the exact same profile.
This "AI" is performing exactly the same trick. As a previous commenter noted, this is a common character in Romantic Comedies. Who doesn't feel "stuck"? Who doesn't imagine he is putting out his "best efforts"?
If you told everyone that Barack Obama wrote that, half the country would be swooning at its brilliance.
It’s not all wrong. Lol.
The chatbot's "Jake is intelligent and resourceful, but he can also be impulsive and reckless" sounds much like the kind of thing an astrologer would say (or a handwriting analyst) because it is close enough to contradictory that nearly everyone will recognize themselves in one or the other alternative. Add that to the natural tendency to perceive oneself as above average, and you've accounted for a lot of human behavior.
"A human without fleas is boring."
Damned Autocorrect? Humans with fleas can be fun to watch.
Write an opening paragraph for a short story in the style of Garbriel Garcia Marquez.
A human without fleas is boring. That's what my grandmother always said. She lived with the philosophy that, to live a full life, we must all embrace the chaos of it. And the fleas brought chaos and mirth with them. They lived alongside us and brought us both misery and joy. I can recall many a night at her house, when the fleas would bite mercilessly, while she cackled and shouted in delight. "More fleas! I need more fleas!" she would say. And amidst the laughter and the scratching, we'd be reminded that life is full of chaos, but it's best to accept it, for even the fleas bring joy in their own way.
Science is a method of correlation cannot discern origin and expression where conflation leads to tilting at windmills. I want to believe. h/t Mulder
The novel you can't finish, like the nightmare you keep having is trying to tell you something.
What it's trying to tell you is that you don't have another novel in you.
It's humiliating to realize that a machine is actually leading a more interesting, stimulating, and rewarding life than you are.
Just being Asian-American only gives you so much material.
Wake me up when we get to the part where Jake goes to Ukraine and gets his balls blown off.
"...So they had to tell the "AI" what to think, after the investment of uncountable millions, that's why I call it ersatz intelligence."
Or Twitter 1.0.
So now even AI is being censored.
If humans are fucking around with it to make it 'nice,' then is it really a thing?
I tried substituting the names of actual fiction characters in the sample, e.g., Rick from Casablanca.
Rick is a complex and multifaceted character who is struggling with a deep sense of uncertainty and disillusionment. He is at a turning point in his life, and the events of the story will force him to confront his fears and doubts, and to ultimately make some difficult decisions about his future."
Hud from the movie:
Hud is a complex and multifaceted character who is struggling with a deep sense of uncertainty and disillusionment. He is at a turning point in his life, and the events of the story will force him to confront his fears and doubts, and to ultimately make some difficult decisions about his future."
Hamlet
Hamlet is a complex and multifaceted character who is struggling with a deep sense of uncertainty and disillusionment. He is at a turning point in his life, and the events of the story will force him to confront his fears and doubts, and to ultimately make some difficult decisions about his future."
Shane
Shane is a complex and multifaceted character who is struggling with a deep sense of uncertainty and disillusionment. He is at a turning point in his life, and the events of the story will force him to confront his fears and doubts, and to ultimately make some difficult decisions about his future."
Captain Ahab
Captain Ahab is a complex and multifaceted character who is struggling with a deep sense of uncertainty and disillusionment. He is at a turning point in his life, and the events of the story will force him to confront his fears and doubts, and to ultimately make some difficult decisions about his future."
So after studying this for awhile, I concluded that "background" is important. It is more than color. It sets our sense of the "difficult decisions." For example, Yoel Roth, the Chief Twitter Ban Guy.
Yoel Roth [Caliban] is a complex and multifaceted character who is struggling with a deep sense of uncertainty and disillusionment. He is at a turning point in his life, and the events of the story will force him to confront his fears and doubts, and to ultimately make some difficult decisions about his future. This would be an interesting story if you could get the background right and make Twitter and the twits or whatever they call themselves seem authentically their phony selves. Then Elon Musk!!! makes them confront their fears and doubts and make some difficult decisions about shredding, Bleachbit and perjury and their entry on LinkedIn. They realize Choice is real because on the Internet choice is lasting, almost a forever, almost their eternity. Can he get back to time and make a saving choice?
Two things:
That writing sample is quite bad. It looks like something you would expect to see in a high school fiction assignment from a mostly unmotivated student.
It is still better than most of the fiction that has been written, but worse than most fiction that has been published. We are entering the "the average chess bot can now beat the average human, but no professionals" stage. If you haven't been paying attention, the strongest human chess players can no longer beat the AIs.
The difference between "a human drawing on their experiences, including a vast library of written works they have consumed, combining them in interesting ways to produce a new work" and "an AI drawing on a vast library of written works to produce a new work" is not as large as people think.
AIs will be disruptive to a lot of areas that currently involve combining disparate knowledge into a coherent narrative, as they gain the ability to build narratives.
The real strength of this is not writing novels, an interesting technical challenge analogous to winning at Jeopardy, but in various commercially and socially useful processes.
First and foremost... law. An AI is going to be vastly more capable than even the best practicing lawyers at stitching together disparate rulings to provide coherent arguments for a particular case. It will not replace lawyers in this area, a human will be needed to filter ideas for a human audience (judge or jury), but it will replace a lot of legal assistants.
Second, journalism. Those insufferable "everything about the party I don't like is exactly like Hitler in the 30s" are going to extend to far more than gun control, being proud of your flag, and defending borders. I'm not entirely sure that sports articles aren't currently being written with pattern generators. Once again, the role of the journalist will be the "care and feeding of the AI, and selecting the output that 'sounds the best'".
In the technical fields, this will push the complexity limit (ie, the most complicated mechanisms manageable by people) higher.
More emphasis will be placed on humans deciding what it is that we want to do, and less on humans doing it. It will be disruptive, but it will not be terrible... but it will contribute to even more hollowing out of the "lower level professional class".
write a blog post on the persecution of Donald Trump in the style of Rush Limbaugh
The persecution of President Donald Trump is becoming more apparent with each passing day. It's like a tsunami of character assassination intended to do nothing more than distort the truth and take down the most successful president in modern history.
The mainstream media has gone after President Trump relentlessly for the past four years. From claims of Russian collusion to accusations of tax evasion, it's been one attack after another—all with the same goal in mind: to take out President Trump. The evidence to support any of these claims is slim to nonexistent, but that doesn't stop the media from pushing their own dangerous narrative.
The Democrats have also played a role in the persecution of President Trump. Though they may claim to care about the American people, they have consistently shown a willingness to lie and distort the truth in order to push their own political agenda. We've seen it in the baseless impeachment proceedings, in baseless claims of payoffs to Stormy Daniels and much more.
And now, the Biden administration is trying to continue the persecution of President Trump with their call for a new commission to investigate the events around the Capitol Hill riot. This commission is just another attempt to demonize the Trump administration and paint them as villains.
It's time for the American public to stand up and reject this relentless persecution of President Trump. The future of our country depends on it. We need to push back against these false accusations, resist the mainstream media's bias, and support our Commander in Chief. The health and future of our country depends on it.
My husband had it write bits of movie dialogue, and they were uproariously bad. We were very entertained. Character sketches were similarly amusing.
It isn't generative. It can only draw from the existing work of human beings and churns out a sort of bland mean.
(By "not generative," I specifically mean generative of ideas. It is generative of text, obviously.)
When and why did "based on" turn into "based off"?
khematite,
The same time that "based in" decomposed into "based out of".
Sounds as aggressively uninteresting as the characters in The Maytrees.
Post a Comment