Not to be glib, but this reminds me of days spent reading TV tropes.
I think the author is stretching a little to say noone considered this since Aristotle. Not only is there a lot of overlap with meme theory, people have been doing literary criticism forever.
> people have been doing literary criticism forever.
Do you have any examples? I'm recollecting Augustine, Aquinas, Boethius, Abelard, Plotinus, Maimonides, et al, and nothing like Poetics comes to mind. What am I not remembering?
Not the commenter you responded to, but Northrup Frye at the very least tried to take a systematic approach to literary theory, using Aristotle as a model. I would like to peek into the parallel universe in which his approach in the Anatomy of Criticism (1957) became as widely adopted as, e.g. deconstructionism.
Yes, many instances of modern literary criticism come to mind, such as Ben Jonson's commentary on the works of Shakespeare, his friend, producer, and cast member; Cervantes' analysis embedded in Quixote and Sheridan's analysis embedded in Tristram Shandy; Tolstoy's What Is Art; etc.
My impression remains that during the period between Poetics and, oh, Canterbury Tales, the great thinkers didn't publish anything about Narrative.
None to add to that, though I was more thinking (as another child post) of poets and thinkers who were clearly self-referentially aware of what they were doing and what was going on generally with literary narrative (Swift and Coleridge come to mind).
Edit: oh that child post was you :). Yeah there was for sure a gaping lacuna before Chaucer, and even then you may be right that there was nothing as explicitly on the nose until the 20th century.
TV tropes and Mr Smithsonian are still late to the race in the scheme of things but nearer the pointy end than my original post may have given credit.
Absolutely, it's weird isn't it–talk about an elephant in the room! I'm not sure what would be weirder, if the author never heard of tvtropes.org, or they know of it but write as if it doesn't exist. The author seems to live in some parallel universe, very strange. I think you're too kind though–it makes the article entirely ridiculous.
And then... I read this[0], which talks about the author's work. He's got degrees in literature and neuroscience. And the Smithsonian magazine article was advertising for his book, so he's hardly gonna mention the rabbit-hole that is Tv Tropes or people will never get around to reading his book... So I almost didn't post the above first paragraph.
But that page also says "His most recent work anatomizes the fundamental difference between computer AI and human narrative intelligence; a sample can be found in his 2021 proof in Narrative of why computers will never be able to read (or write) novels."
Ok well...that, I thought, sounds 99.99% likely to be total nonsense, so I had a look.[1] The paper's called Why Computers Will Never Read (or Write) Literature: A Logical Proof and a Narrative..
"In this article, I'll provide a definite answer. And that answer is: never. No computer, no matter how immense its circuitry, will ever be able to extract the know-how from a fairytale that can be gleaned by human children. No machine-learning algorithm, no matter how futuristic its software, will ever author a sonnet or short story. The reason for this is simply that literature encodes a great deal of its thought-stuff in narrative, a mode of communication that requires causal reasoning to process. And while the ability to do causal reasoning is embedded in the architecture of the human brain, computers are hardwired to perform a method of thinking—symbolic logic—that is fundamentally incapable of grasping cause-and-effect."
I found it hard to read.. "As we saw above, this process of scientific learning does not yield absolute truths, and in fact, when we make the mistake of conflating science with truth (in the way that "Enlightenment" thinkers from Thomas Macaulay to Steven Pinker have done) we can veer into smugness, imperialism, and other habits antithetical to the curiosity, open-mindedness, and bias awareness necessary for effective science." (sample paragraph)
Anyway, at the end he gives his proof – although he seems to have changed it, somewhat significantly, from "Will Never" to "Cannot"! :
A Logical Proof That Computers Cannot Read (or Write) Literature
1. Literature has a rhetorical function.
2. Literature's full rhetorical function depends on narrative elements.
3. Narrative elements rely on causal reasoning.
4. Causal reasoning cannot be performed by machine-learning algorithms because those algorithms run on the CPU's Arithmetic Logic Unit, which is designed to run symbolic logic, and symbolic logic can only process correlation.
QED: Computers cannot perform the causal reasoning necessary for learning to use literature.
I imagine by "Cannot" he means "Cannot now, and won't ever" - at least, the title and the abstract say that.
This all strikes me as total garbage.[2] And thus the garbagey flavour of the Smithsonian article, not surprisingly, doesn't seem an isolated aberration.
Bringing this down to the CPU is as absurd as all those 17th century opponents of materialism. "How can rude matter bring about all the ineffable qualia!" Hilarious.
>The paper's called Why Computers Will Never Read (or Write) Literature: A Logical Proof and a Narrative..
That reminds me of a more serious approach of that proof.
Roger Penrose's books, The Emperor's New Mind and the two sequels
The Penrose–Lucas argument based on Gödel's incompleteness theorem. From the theorem we know any mathematical system has sentences that the system itself cannot prove.
So computer algorithms, based on mathematical systems, cannot prove them either. (just like the Halting problem)
But humans can prove them like Gödel did.
Thus human intelligence is noncomputable and inherently superior to AIs.
That would also be true for novels, since novels can contain character talking about such proves.
But, Penrose takes it to the next step. The laws of physics are also a mathematical system and fully simulatable by a computer. But since the computer cannot simulate human intelligence as proved before, this proves that human intelligence is not described by physical laws.
That raises the question, if consciouness is not described by physical laws, ? Essentially it would appear as a random influence to physics (even though it is not actually random, just noncomputable). Now there is a completely random process in physics: quantum wave function collapse. Then he goes on, that consciouness comes from features embedded at the Planck scale in the spacetime geometry.
>4. Causal reasoning cannot be performed by machine-learning algorithms because those algorithms run on the CPU's Arithmetic Logic Unit, which is designed to run symbolic logic, and symbolic logic can only process correlation.
And I just wrote my PhD thesis about causal reasoning algorithms :/
> The Penrose–Lucas argument based on Gödel's incompleteness theorem. From the theorem we know any mathematical system has sentences that the system itself cannot prove. So computer algorithms, based on mathematical systems, cannot prove them either. (just like the Halting problem) But humans can prove them like Gödel did. Thus human intelligence is noncomputable and inherently superior to AIs.
Godel's incompleteness theorems apply to math itself, not to computation. Those observations are not provable but incomputable, they are not provable in the formal system at all. So, humans can't prove them anymore than computers could. The gods themselves couldn't prove them.
What I believe the argument could be is that humans still found them and have reasons to believe they are true (they have an informal proof), even if they can't be formally proven.
However, the Penrose-Lacau argument is quite infamous and is generally not believed in the scientific community. The most obvious way out of Godel's incompleteness theorems conflict with human mathematicians is to simply reject the notion that human mathematicians have proven these things in any sense. Godel's incompleteness theorems don't forbid incorrect but convincing proofs, and informal proofs that can't be formalized could well be just that.
14 comments
[ 0.22 ms ] story [ 40.6 ms ] threadI think the author is stretching a little to say noone considered this since Aristotle. Not only is there a lot of overlap with meme theory, people have been doing literary criticism forever.
Nice article though.
Do you have any examples? I'm recollecting Augustine, Aquinas, Boethius, Abelard, Plotinus, Maimonides, et al, and nothing like Poetics comes to mind. What am I not remembering?
My impression remains that during the period between Poetics and, oh, Canterbury Tales, the great thinkers didn't publish anything about Narrative.
Thanks for the Northrop Frye tip!
Edit: oh that child post was you :). Yeah there was for sure a gaping lacuna before Chaucer, and even then you may be right that there was nothing as explicitly on the nose until the 20th century.
TV tropes and Mr Smithsonian are still late to the race in the scheme of things but nearer the pointy end than my original post may have given credit.
And then... I read this[0], which talks about the author's work. He's got degrees in literature and neuroscience. And the Smithsonian magazine article was advertising for his book, so he's hardly gonna mention the rabbit-hole that is Tv Tropes or people will never get around to reading his book... So I almost didn't post the above first paragraph.
But that page also says "His most recent work anatomizes the fundamental difference between computer AI and human narrative intelligence; a sample can be found in his 2021 proof in Narrative of why computers will never be able to read (or write) novels."
Ok well...that, I thought, sounds 99.99% likely to be total nonsense, so I had a look.[1] The paper's called Why Computers Will Never Read (or Write) Literature: A Logical Proof and a Narrative..
"In this article, I'll provide a definite answer. And that answer is: never. No computer, no matter how immense its circuitry, will ever be able to extract the know-how from a fairytale that can be gleaned by human children. No machine-learning algorithm, no matter how futuristic its software, will ever author a sonnet or short story. The reason for this is simply that literature encodes a great deal of its thought-stuff in narrative, a mode of communication that requires causal reasoning to process. And while the ability to do causal reasoning is embedded in the architecture of the human brain, computers are hardwired to perform a method of thinking—symbolic logic—that is fundamentally incapable of grasping cause-and-effect."
I found it hard to read.. "As we saw above, this process of scientific learning does not yield absolute truths, and in fact, when we make the mistake of conflating science with truth (in the way that "Enlightenment" thinkers from Thomas Macaulay to Steven Pinker have done) we can veer into smugness, imperialism, and other habits antithetical to the curiosity, open-mindedness, and bias awareness necessary for effective science." (sample paragraph)
Anyway, at the end he gives his proof – although he seems to have changed it, somewhat significantly, from "Will Never" to "Cannot"! :
A Logical Proof That Computers Cannot Read (or Write) Literature
1. Literature has a rhetorical function.
2. Literature's full rhetorical function depends on narrative elements.
3. Narrative elements rely on causal reasoning.
4. Causal reasoning cannot be performed by machine-learning algorithms because those algorithms run on the CPU's Arithmetic Logic Unit, which is designed to run symbolic logic, and symbolic logic can only process correlation.
QED: Computers cannot perform the causal reasoning necessary for learning to use literature.
I imagine by "Cannot" he means "Cannot now, and won't ever" - at least, the title and the abstract say that.
This all strikes me as total garbage.[2] And thus the garbagey flavour of the Smithsonian article, not surprisingly, doesn't seem an isolated aberration.
[0] https://projectnarrative.osu.edu/about/current-research/rese...
[1] https://muse.jhu.edu/article/778252
[2] Sorry I don't have more time or the inclination for more than this shallow dismissal. Ars longa, vita brevis.
This is the discussion you mean, I think: https://news.ycombinator.com/item?id=26127165
Bringing this down to the CPU is as absurd as all those 17th century opponents of materialism. "How can rude matter bring about all the ineffable qualia!" Hilarious.
That reminds me of a more serious approach of that proof.
Roger Penrose's books, The Emperor's New Mind and the two sequels
The Penrose–Lucas argument based on Gödel's incompleteness theorem. From the theorem we know any mathematical system has sentences that the system itself cannot prove. So computer algorithms, based on mathematical systems, cannot prove them either. (just like the Halting problem) But humans can prove them like Gödel did. Thus human intelligence is noncomputable and inherently superior to AIs.
That would also be true for novels, since novels can contain character talking about such proves.
But, Penrose takes it to the next step. The laws of physics are also a mathematical system and fully simulatable by a computer. But since the computer cannot simulate human intelligence as proved before, this proves that human intelligence is not described by physical laws.
That raises the question, if consciouness is not described by physical laws, ? Essentially it would appear as a random influence to physics (even though it is not actually random, just noncomputable). Now there is a completely random process in physics: quantum wave function collapse. Then he goes on, that consciouness comes from features embedded at the Planck scale in the spacetime geometry.
>4. Causal reasoning cannot be performed by machine-learning algorithms because those algorithms run on the CPU's Arithmetic Logic Unit, which is designed to run symbolic logic, and symbolic logic can only process correlation.
And I just wrote my PhD thesis about causal reasoning algorithms :/
Sounds interesting. Link?
Godel's incompleteness theorems apply to math itself, not to computation. Those observations are not provable but incomputable, they are not provable in the formal system at all. So, humans can't prove them anymore than computers could. The gods themselves couldn't prove them.
What I believe the argument could be is that humans still found them and have reasons to believe they are true (they have an informal proof), even if they can't be formally proven.
However, the Penrose-Lacau argument is quite infamous and is generally not believed in the scientific community. The most obvious way out of Godel's incompleteness theorems conflict with human mathematicians is to simply reject the notion that human mathematicians have proven these things in any sense. Godel's incompleteness theorems don't forbid incorrect but convincing proofs, and informal proofs that can't be formalized could well be just that.