AIs recently passed a fiction-writing Turing test. The lesson I learned from this is that, wow, humans are garbage at both writing and reading flash fiction.
"How many readers are genuinely discerning? (Sadly few, judging by the original blog post.)"
I wonder how much of this is selection effect. If someone read the two stories and said "I don't know if they're AI or not, but I don't care enough to read 6 more mediocre stories", presumably they didn't vote. That's what happened to me when I first saw it going around (though I think after the voting period closed, so I wouldn't have affected the results either way).
I might also add that humans write worse when writing to a prompt than when they write something that inspires them. My secret for prompts is generally to mash them up with an idea that's already been rotating in my mind for a while, but I don't think I could manage that in a flash length.
Basically, if humans write because they have something to say, they can produce something AI can't (because it never has anything to say) but when it comes to prompt fills, maybe they end up doing a lot more like what the AI does.
But generally I agree with this point of yours the most: there is no demand for mediocre writing. The only reason I read it is because I know the person who wrote it. In that case I read it for the same reason I listen when people are talking: I want to be in conversation with them as a person. This is a thing we're doing for our relationship that I wouldn't do purely for the words themselves.
But I don't find any lack of good stuff when I'm reading for quality. Even in a very narrow subgenre (explicit novel-length Kirk/Spock fiction, TOS only) I've read for two years and not run out of good stuff written by humans. Not perhaps bestseller quality, but having enough originality and sincerity to feel worth my time.
I remember once seeing a grant for a poet to live in the Italian countryside for a month; the catch was that the poet had to write solely about the Italian countryside. I understand that some people may relish such challenges, but the exercise would have been frustrating and near-impossible for me. Hmmm...what rhymes with "penne"?
I read these stories independently and I completely agree with your assessment of them. I feel bad about criticizing the work of real humans too harshly, but I still find it hard to believe that #1 was written by a successful professional author. Definitely not going out and buying anything by that person, unless this is some *extremely* anomalous low point in their body of work.
Successful Percy Jackson author Rick Riordan infamously wrote "Nachos after the War", which was very painful to read (and I really enjoyed many of his books!)
Ms Janny Wurts seems to be doing alright for herself. She has over 20 professionally published books, and coauthored a best-selling series. I don't know if she makes enough from her books to live on or if she has a day job, but I'd definitely call her a successful professional author.
I peeked at her highest-rated book on GoodReads and yeah, it's written in the same thesaurus-happy style, only a notch or two less purple than her contest entry. I personally hate it, but there clearly is a market for it, so who cares what you and I think?
[EDIT: Adding context.] Professors gave a set of English majors the first few paragraphs of "Bleak House" and asked them to explain the paragraphs in plain language (they were allowed to look up words). Most students could not, although most of those who could not still thought they would have no trouble reading the rest of the book.
Some interesting commentary here from a (former?) teacher who suggests many people essentially do not form a coherent model of what they read, even if they are people who self-identity as readers and enjoy it:
That study is so ill-designed that it's hard to say it shows anything.
The main problem is that the test for comprehension is "guess the detail I want you to mention". If you don't mention the connection between the literal fog and the vision-obscuring industrial soot, the examiner just says "OK" and fails you, even if you were perfectly aware of it and just decided to discuss something else.
This could perhaps be fixed if the authors published more data than just the final scores, but they didn't.
Another problem is the method seems designed to induce nerves: Yeah I'd find it pretty hard to picture or understand what I'm reading if I had to read it out loud in front of a stone-faced examiner asking ambiguous questions.
It's not necessarily a problem to complain that a student of English literature should know that "whiskers" can refer to a person's facial hair and not just an animal's, or be familiar with Lincoln's Inn, but it makes the study hard to generalise to other populations — I certainly don't feel bad for not knowing those things beforehand.
This is worsened by the difficulty of looking up many of the terms: "Michaelmas term" brings up only its use in schools rather than courts of law, and "Lord Chancellor" brings up his current role, not that he used to be a judge.
All that said, it's true that some of the readers didn't have a clear mental picture of events — and they simply noted the confusion and decided to see if it would become clearer later, which is perfectly sensible.
Yeah I did pretty poorly- I guessed the first 5 were AI and the last 3 were human.
As you say, the easiest AI tells were the parts that made no sense. The most obvious ones to me were these two:
* "Father Aldric barely had time to drop his rosary before the beast was on him [...] Aldric pulled the blade from beneath the altar. [...] "Then come take your tithe, bastard." If the beast is "on him", how is he pulling out a sword and calmly reciting catchphrases?
* "“What do you want?” I asked. The demon traced a finger along the rim of its cup. “A name. Spoken aloud in the right place. That is all.” [...] When I spoke the name, the café windows fogged over. Outside, the rain stopped in midair, each drop quivering as if listening. [...] “Thank you,” it said. “You will sleep tonight.”" So the "right place" was... the cafe where they're currently sitting? What?
Stylistic tells don't seem to be reliable. I thought that #1 was AI due to the same paragraph you called out, and it seems like most respondents did as well, it was voted the "most AI" story overall. But I guess AI style exists because it's copying the average human writer.
I think you're being too harsh on #8. The author presumably was just trying to write a good story about a demon, not trying specifically to write a story that would remain good in the context of 7 other demon stories.
I expect the "does it make sense" method won't work forever. Even the best human-written stories are notorious for having plot holes and inconsistencies, they just happen across a much larger body of work. Doesn't seem like the AI is doing anything fundamentally different here, it just loses coherence on the scale of sentences whereas humans lose it on the scale of chapters.
She knew it was going to be published next to seven other demon stories! I think it's fair to expect an author to keep in mind the context in which a story appears.
“In spite of all this flowery description, we don't get a single novel image, an image which suggests that the author has either observed the world or made anything up out of their head. We have a red-eyed demon with snakelike ability to taste the air, who is wearing a ratty dark purple suit. We have a city with a bus stop and hurrying crowds whose coats get wet in the rain and who wear headphones with glowing power lights.”
I’m a total philistine myself — I spent a lot of time writing fiction as a child, but never got good at it. But I was reminded of something I noticed a few years ago when I read the story at the center of the “Bad Art Friend” scandal. The main character is in the hospital, and at one point the narration says that the pattern on the ceiling of her room looks like constellations, but not in a literal sense.
And I thought: why do we care about that image? Is it a thing in modern literary fiction to go “This needs more similes!” and put in ones that don’t make sense?
Again, I am not a specialist in this area, and the Larson kidney donation story could have been doing something very clever that I just didn’t understand…but I do have a prior at this point that humans can sometimes use metaphors or imagery that seem ill-fitting or bizarre, which would make it less of an AI tell.
The conservative literary Substacker Liza Libes was saying she was editing a novel for publication, and they were having her chop out all the flowery language she actually liked. The modern style is very telegraphic and image-poor. I think the kids get their pictures in bulk from Tiktok and YouTube, so 19th-century page-long descriptions of the way a house looked seem pointless and unnecessary.
> Again, I am not a specialist in this area, and the Larson kidney donation story could have been doing something very clever that I just didn’t understand.
It was not, but at least you can take satisfaction in knowing that it made the author a grand total of $425 despite being workshopped for literal years.
I had that thought as well. I was at an author conversation on a book on AI personhood and he said one possibility of the statistical-prediction aspect of the LLMs making them sound so human is giving them some aspects of consciousness, and the other (which he didn't like so much) is that humans aren't as conscious as we previously thought.
The second (less-desired) possibility was the one I went with--we are just meat machines, most of what we do is pretty predictable, and it's not surprising you can make something sound like us. If I say hello and you say hello back, that's easy to simulate--a conversation about the kids or the weather (or politics) would likely be easy to auto-generate as well given given the appropriate training set.
Similarly, an LLM will probably be able to create a passable 80s pulp scifi novel or 2020s romantasy novel. My understanding is it's right now limited by the size of the context window--it can't actually 'remember' enough to do a novel. But--memory may become cheaper, the algorithms may become more sophisticated about the way they use it, if there's one thing I've seen over the past few decades it's that your brethren in San Francisco are quite creative at overcoming technical difficulties!
(Ironically a combination of this and the modern disdain for publishing male writers is a reason I abandoned my childhood fantasy of using my remaining midlife to write novels--I might get better, but I'm not going to get better than ChatGPT. I sold my remaining books on how to write fiction and used the proceeds to buy a used copy of The Killing Joke.)
One important caveat: you're not going to get the next Cormac McCarthy or Toni Morrison out of an LLM. But how many people read literary fiction anyway?
Consider annoying some filler between the first two paragraphs: I want to see the contest without spoilers.
I got 7 out of 8, which surprised me: I expected to do only a little better than chance. Could easily be coincidence, ofc.
I agree with you that #6 is the best of the lot, and an easy call for human writing. But I completely disagree about second place.
Sure, #5 is full of AI tells. Sure, "I don't believe anyone who liked this story read it with attention." still holds: I skimmed all the stories (except #6, which actually got me to care), and completely missed the licking part.
But it's dreamlike to make it work, when I skim and let the vibe wash over me. The image in my mind is vague, and doesn't make sense when I look closer: yep, that's how dreams are.
I liked the plum thing: sickly-sweet, unwholesome, on the cusp between firm, fresh fruit and rot, alcohol, death. It's ambiguous just like the demon and the ending are.
IMO you're being too harsh on the loose tie: I interpret it as "loosened as if as an afterthought", "loosened like it would be if he did it as an afterthought", because the demon isn't actually careless and scruffy, he's deliberately choosing to look careless and scruffy.
Overall, I enjoyed it, and gave it 4 stars, a solid second place.
In contrast, I thought #7 and #8 were atrocious. (I thought both were AI.)
#7 has all the problems of #5 without the tone to excuse them, plus the extra problems you describe (boring nonplot, nonsensical blocking). Cute premise is cute but not nearly enough for me to be as forgiving as you are.
#8 is the one I got wrong. I'm surprised it was so bad: I'm not Ms Hobb's hugest fan but her books are pretty good. It's tedious, and it has the bad-fanfiction problem of referring to Evory as "the new friend" for no reason. The only appeal is the contrast between the tiny mundane details and the reveal about the demon, but, as you mention, it's not a reveal at all.
Your conclusion confuses me. It seems to conflate books that sell super well with books that discerning readers (such as you) think are the best, but that's clearly not the case: most bestsellers are cliche slop that you and I turn up our noses at.
I mean yes, #6 was the only decent story in the lot, for sure.
I actually rated #1 second-best. Yes, I know; the prose is really bad. But it is at least a competent idea, reasonably structured, which is more than can be said of... well... any of the others. It's not good, no. But it's maybe slightly less bad than the rest.
I did very badly in general. I correctly guessed #6 and #1 (and I was pretty blooming sure about #6). But I got almost all of the others wrong (and I disliked them all to varying degrees).
And yes, you're right; at the moment AI serves basically no function in fiction writing. I touched on the matter in my most recent post, as well: https://andrewcurrall.substack.com/p/on-ai-art
This is why, as an aspiring children’s author, I do not feel threatened by LLMs; so much of what’s already out there is already slop, and if said trash gets replaced by robots, then at least no one’s getting rewarded for contributing to it.
"Most seriously, #7 is structured (in traditional flash fashion) as if there's a satisfying reveal at the end, but there isn’t one."
well it was sUbvErtIng exPecTations
Most seriously (like you say), I agree with what other commenters said, that stories here are too mediocre to grab my attention, regardless of who or what wrote them. I'd like to see an AI-written story that could compete with Asimov's robot mysteries.
"there is no market for new fiction because all the fiction people read is generated by LLMs to their custom specifications"
I think this is more or less happening right now with erotic stories, since those have a lower bar of "literary" quality anyway, and people already want them more customized than they have possibly be (because of how wildly different fetishes there are). There could be an erotic book market apocalypse in less than a year.
Yeah the OP made the rounds last month among concerned/aspiring writers but mostly got laughed at on closer inspection.
Key context here is that Mark Lawrence is the single worst best-selling fantasy author of his generation and it isn't particularly close. What the kids these days call an "industry plant" I think.
"How many readers are genuinely discerning? (Sadly few, judging by the original blog post.)"
I wonder how much of this is selection effect. If someone read the two stories and said "I don't know if they're AI or not, but I don't care enough to read 6 more mediocre stories", presumably they didn't vote. That's what happened to me when I first saw it going around (though I think after the voting period closed, so I wouldn't have affected the results either way).
Yeah I almost quit at the second story, it's just unpleasant to read.
Yeah, I went "I don't know if these are AI; either way they're terrible."
I might also add that humans write worse when writing to a prompt than when they write something that inspires them. My secret for prompts is generally to mash them up with an idea that's already been rotating in my mind for a while, but I don't think I could manage that in a flash length.
Basically, if humans write because they have something to say, they can produce something AI can't (because it never has anything to say) but when it comes to prompt fills, maybe they end up doing a lot more like what the AI does.
But generally I agree with this point of yours the most: there is no demand for mediocre writing. The only reason I read it is because I know the person who wrote it. In that case I read it for the same reason I listen when people are talking: I want to be in conversation with them as a person. This is a thing we're doing for our relationship that I wouldn't do purely for the words themselves.
But I don't find any lack of good stuff when I'm reading for quality. Even in a very narrow subgenre (explicit novel-length Kirk/Spock fiction, TOS only) I've read for two years and not run out of good stuff written by humans. Not perhaps bestseller quality, but having enough originality and sincerity to feel worth my time.
I remember once seeing a grant for a poet to live in the Italian countryside for a month; the catch was that the poet had to write solely about the Italian countryside. I understand that some people may relish such challenges, but the exercise would have been frustrating and near-impossible for me. Hmmm...what rhymes with "penne"?
I read these stories independently and I completely agree with your assessment of them. I feel bad about criticizing the work of real humans too harshly, but I still find it hard to believe that #1 was written by a successful professional author. Definitely not going out and buying anything by that person, unless this is some *extremely* anomalous low point in their body of work.
Successful Percy Jackson author Rick Riordan infamously wrote "Nachos after the War", which was very painful to read (and I really enjoyed many of his books!)
https://web.archive.org/web/20200428135217/https://rickriordan.com/2020/03/nachos-after-the-war-a-heroes-of-olympus-outtake/
Ms Janny Wurts seems to be doing alright for herself. She has over 20 professionally published books, and coauthored a best-selling series. I don't know if she makes enough from her books to live on or if she has a day job, but I'd definitely call her a successful professional author.
I peeked at her highest-rated book on GoodReads and yeah, it's written in the same thesaurus-happy style, only a notch or two less purple than her contest entry. I personally hate it, but there clearly is a market for it, so who cares what you and I think?
Her books are pretty big though. Maybe flash fiction is far off enough from her usual style she stumbled.
I mean, if you brought Shakespeare to the modern day and asked him to write a rap song he'd probably produce some howlers before he figured it out.
"he smirked with a swagger"
AAAAAAAAH
I am reminded of the "They Don't Read Very Well" study: https://muse.jhu.edu/pub/1/article/922346
[EDIT: Adding context.] Professors gave a set of English majors the first few paragraphs of "Bleak House" and asked them to explain the paragraphs in plain language (they were allowed to look up words). Most students could not, although most of those who could not still thought they would have no trouble reading the rest of the book.
Some interesting commentary here from a (former?) teacher who suggests many people essentially do not form a coherent model of what they read, even if they are people who self-identity as readers and enjoy it:
https://www.tumblr.com/prettyboysdontlookatexplosions/783379386552516608/i-appreciated-this-study-they-cant-read-very
That study is so ill-designed that it's hard to say it shows anything.
The main problem is that the test for comprehension is "guess the detail I want you to mention". If you don't mention the connection between the literal fog and the vision-obscuring industrial soot, the examiner just says "OK" and fails you, even if you were perfectly aware of it and just decided to discuss something else.
This could perhaps be fixed if the authors published more data than just the final scores, but they didn't.
Another problem is the method seems designed to induce nerves: Yeah I'd find it pretty hard to picture or understand what I'm reading if I had to read it out loud in front of a stone-faced examiner asking ambiguous questions.
It's not necessarily a problem to complain that a student of English literature should know that "whiskers" can refer to a person's facial hair and not just an animal's, or be familiar with Lincoln's Inn, but it makes the study hard to generalise to other populations — I certainly don't feel bad for not knowing those things beforehand.
This is worsened by the difficulty of looking up many of the terms: "Michaelmas term" brings up only its use in schools rather than courts of law, and "Lord Chancellor" brings up his current role, not that he used to be a judge.
All that said, it's true that some of the readers didn't have a clear mental picture of events — and they simply noted the confusion and decided to see if it would become clearer later, which is perfectly sensible.
Yeah I did pretty poorly- I guessed the first 5 were AI and the last 3 were human.
As you say, the easiest AI tells were the parts that made no sense. The most obvious ones to me were these two:
* "Father Aldric barely had time to drop his rosary before the beast was on him [...] Aldric pulled the blade from beneath the altar. [...] "Then come take your tithe, bastard." If the beast is "on him", how is he pulling out a sword and calmly reciting catchphrases?
* "“What do you want?” I asked. The demon traced a finger along the rim of its cup. “A name. Spoken aloud in the right place. That is all.” [...] When I spoke the name, the café windows fogged over. Outside, the rain stopped in midair, each drop quivering as if listening. [...] “Thank you,” it said. “You will sleep tonight.”" So the "right place" was... the cafe where they're currently sitting? What?
Stylistic tells don't seem to be reliable. I thought that #1 was AI due to the same paragraph you called out, and it seems like most respondents did as well, it was voted the "most AI" story overall. But I guess AI style exists because it's copying the average human writer.
I think you're being too harsh on #8. The author presumably was just trying to write a good story about a demon, not trying specifically to write a story that would remain good in the context of 7 other demon stories.
I expect the "does it make sense" method won't work forever. Even the best human-written stories are notorious for having plot holes and inconsistencies, they just happen across a much larger body of work. Doesn't seem like the AI is doing anything fundamentally different here, it just loses coherence on the scale of sentences whereas humans lose it on the scale of chapters.
She knew it was going to be published next to seven other demon stories! I think it's fair to expect an author to keep in mind the context in which a story appears.
“In spite of all this flowery description, we don't get a single novel image, an image which suggests that the author has either observed the world or made anything up out of their head. We have a red-eyed demon with snakelike ability to taste the air, who is wearing a ratty dark purple suit. We have a city with a bus stop and hurrying crowds whose coats get wet in the rain and who wear headphones with glowing power lights.”
I’m a total philistine myself — I spent a lot of time writing fiction as a child, but never got good at it. But I was reminded of something I noticed a few years ago when I read the story at the center of the “Bad Art Friend” scandal. The main character is in the hospital, and at one point the narration says that the pattern on the ceiling of her room looks like constellations, but not in a literal sense.
And I thought: why do we care about that image? Is it a thing in modern literary fiction to go “This needs more similes!” and put in ones that don’t make sense?
Again, I am not a specialist in this area, and the Larson kidney donation story could have been doing something very clever that I just didn’t understand…but I do have a prior at this point that humans can sometimes use metaphors or imagery that seem ill-fitting or bizarre, which would make it less of an AI tell.
The conservative literary Substacker Liza Libes was saying she was editing a novel for publication, and they were having her chop out all the flowery language she actually liked. The modern style is very telegraphic and image-poor. I think the kids get their pictures in bulk from Tiktok and YouTube, so 19th-century page-long descriptions of the way a house looked seem pointless and unnecessary.
"at one point the narration says that the pattern on the ceiling of her room looks like constellations, but not in a literal sense"
Presumably it also look like a hatstand, and a goat, and the winning 1966 England football team, but not in a literal sense.
> Again, I am not a specialist in this area, and the Larson kidney donation story could have been doing something very clever that I just didn’t understand.
It was not, but at least you can take satisfaction in knowing that it made the author a grand total of $425 despite being workshopped for literal years.
I had that thought as well. I was at an author conversation on a book on AI personhood and he said one possibility of the statistical-prediction aspect of the LLMs making them sound so human is giving them some aspects of consciousness, and the other (which he didn't like so much) is that humans aren't as conscious as we previously thought.
The second (less-desired) possibility was the one I went with--we are just meat machines, most of what we do is pretty predictable, and it's not surprising you can make something sound like us. If I say hello and you say hello back, that's easy to simulate--a conversation about the kids or the weather (or politics) would likely be easy to auto-generate as well given given the appropriate training set.
Similarly, an LLM will probably be able to create a passable 80s pulp scifi novel or 2020s romantasy novel. My understanding is it's right now limited by the size of the context window--it can't actually 'remember' enough to do a novel. But--memory may become cheaper, the algorithms may become more sophisticated about the way they use it, if there's one thing I've seen over the past few decades it's that your brethren in San Francisco are quite creative at overcoming technical difficulties!
(Ironically a combination of this and the modern disdain for publishing male writers is a reason I abandoned my childhood fantasy of using my remaining midlife to write novels--I might get better, but I'm not going to get better than ChatGPT. I sold my remaining books on how to write fiction and used the proceeds to buy a used copy of The Killing Joke.)
One important caveat: you're not going to get the next Cormac McCarthy or Toni Morrison out of an LLM. But how many people read literary fiction anyway?
Consider annoying some filler between the first two paragraphs: I want to see the contest without spoilers.
I got 7 out of 8, which surprised me: I expected to do only a little better than chance. Could easily be coincidence, ofc.
I agree with you that #6 is the best of the lot, and an easy call for human writing. But I completely disagree about second place.
Sure, #5 is full of AI tells. Sure, "I don't believe anyone who liked this story read it with attention." still holds: I skimmed all the stories (except #6, which actually got me to care), and completely missed the licking part.
But it's dreamlike to make it work, when I skim and let the vibe wash over me. The image in my mind is vague, and doesn't make sense when I look closer: yep, that's how dreams are.
I liked the plum thing: sickly-sweet, unwholesome, on the cusp between firm, fresh fruit and rot, alcohol, death. It's ambiguous just like the demon and the ending are.
IMO you're being too harsh on the loose tie: I interpret it as "loosened as if as an afterthought", "loosened like it would be if he did it as an afterthought", because the demon isn't actually careless and scruffy, he's deliberately choosing to look careless and scruffy.
Overall, I enjoyed it, and gave it 4 stars, a solid second place.
In contrast, I thought #7 and #8 were atrocious. (I thought both were AI.)
#7 has all the problems of #5 without the tone to excuse them, plus the extra problems you describe (boring nonplot, nonsensical blocking). Cute premise is cute but not nearly enough for me to be as forgiving as you are.
#8 is the one I got wrong. I'm surprised it was so bad: I'm not Ms Hobb's hugest fan but her books are pretty good. It's tedious, and it has the bad-fanfiction problem of referring to Evory as "the new friend" for no reason. The only appeal is the contrast between the tiny mundane details and the reveal about the demon, but, as you mention, it's not a reveal at all.
Your conclusion confuses me. It seems to conflate books that sell super well with books that discerning readers (such as you) think are the best, but that's clearly not the case: most bestsellers are cliche slop that you and I turn up our noses at.
Ooh, ooh, I have thoughts too.
I mean yes, #6 was the only decent story in the lot, for sure.
I actually rated #1 second-best. Yes, I know; the prose is really bad. But it is at least a competent idea, reasonably structured, which is more than can be said of... well... any of the others. It's not good, no. But it's maybe slightly less bad than the rest.
I did very badly in general. I correctly guessed #6 and #1 (and I was pretty blooming sure about #6). But I got almost all of the others wrong (and I disliked them all to varying degrees).
And yes, you're right; at the moment AI serves basically no function in fiction writing. I touched on the matter in my most recent post, as well: https://andrewcurrall.substack.com/p/on-ai-art
This is why, as an aspiring children’s author, I do not feel threatened by LLMs; so much of what’s already out there is already slop, and if said trash gets replaced by robots, then at least no one’s getting rewarded for contributing to it.
"Most seriously, #7 is structured (in traditional flash fashion) as if there's a satisfying reveal at the end, but there isn’t one."
well it was sUbvErtIng exPecTations
Most seriously (like you say), I agree with what other commenters said, that stories here are too mediocre to grab my attention, regardless of who or what wrote them. I'd like to see an AI-written story that could compete with Asimov's robot mysteries.
"there is no market for new fiction because all the fiction people read is generated by LLMs to their custom specifications"
I think this is more or less happening right now with erotic stories, since those have a lower bar of "literary" quality anyway, and people already want them more customized than they have possibly be (because of how wildly different fetishes there are). There could be an erotic book market apocalypse in less than a year.
> The creature in the space was large and hideous, with features disfigured by [...] a huge erection.
I can't tell whether this writer needs writing lessons or she just needs warm hugs, estrogen, and antiandrogens. Maybe both. xD
Yeah the OP made the rounds last month among concerned/aspiring writers but mostly got laughed at on closer inspection.
Key context here is that Mark Lawrence is the single worst best-selling fantasy author of his generation and it isn't particularly close. What the kids these days call an "industry plant" I think.
TBF Mark Lawrence's flash fiction was the only competent one!
The best sprinters usually aren't the best long-distance runners.