I try to make the case in all my English classes that I teach that “writing is thinking,” but because thinking is hard work, many of my students don’t want to do it. And now after reading your essay, I see even more clearly with your comment about students having for example, Chat GPT, offload the writing and thinking needed to create an essay on an issue. They put the prompt into the LLM, it spits out an essay, they read it and they think “yeah I think that” but as you say, “in reality they had never thought about any such arguments at all. This is confabulation: falsely believing they had done the mental work themselves, convincing themselves that ChatGPT’s reasoning was their own all along.³”
I plan on sharing this essay with my students. I hope it makes a difference.
Hilarious- This is a superb essay. Connecting dots is what LLMs do, and "confabulate" is the perfect word for what happens when they do this in a way that goes haywire and misleads...
It’s weird that you make the observation that AIs can’t hallucinate because they don’t have sensory experience and then immediately go on to say they have delusions. They can’t have delusions, because delusions require (as James says) opinions, and AIs don’t have opinions. They don’t think or even write. They just generate syntax. I think the basic point here is good, but you want to follow it all the way, past all the metaphors that grant GenAI too much mind.
I assume by analogous reasoning that if we built a predictive processor out of electrically-charged meat that you would deny that it had opinions. Just syntax and algorithms, right? But we build them every day and call them human beings. I am not asserting that (current) AIs have thought, but I am reluctant to rush into a reductionism that seems to preclude all mentality.
I definitely deny that humans are predictive processors, so I guess we disagree about what humans are, which is fine. I just don't see how you can rule out (given what you think AI is) that AIs have sensory experiences (and so hallucinate). So, once again, I don't see how you can consistently say "AIs definitely don't hallucinate" and "AIs definitely have delusions". You really would have to be asserting that AIs have thought, not just ruling out reductionism that precludes all mentality.
You are too harsh on Trump. His comments on the Declaration were not BS.
The "unity" is right there in the title ("The unanimous Declaration...") and in the conclusion ("these United Colonies ... we mutually pledge to each other our Lives, our Fortunes and our sacred Honor").
The "respect" is right there in the preamble. Indeed respect is the explicitly-stated motivation for the entire document ("a decent respect to the opinions of mankind requires that they should declare the causes which impel them to the separation"). Did someone fail to read that far into the document?
As for "love," I agree, I don't see much love in the document, but perhaps love, like beauty, is in the eye of the beholder? The following ideas were not at all Conventional Wisdom in the 18th century: "that all men are created equal, that they are endowed by their Creator with certain unalienable Rights.” Personally I would not label this political conviction "love," but I also would not call BS on anyone else who gushed about it in that way.
Yes, we knew that. The Declaration is a long list of the King's alleged misdeeds.
It's not BS to claim that the Declaration expresses respect and unity, too, because it does. Don't let your eternally enraged hatred for Trump blind you to reality, lol
Don’t let your sycophantic worship of Trump blind you to the plain meaning of words, like interpreting “Hey King George, F You. We’re outta here” as a statement of respect, love and unity.
I’ll bet you didn’t like the preceding sentence, did you? Maybe uncharitably interpreting people by ascribing their views to mere emotional outbursts isn’t very productive. More useful is to reflect on what we’re taking as axiomatic in forming our whole models of the world, as I argue for here: https://hilariusbookbinder.substack.com/p/doomsday-prophets-and-dead-parrots.
Been there, done that. Trump supporters get character assassinated 24/7/365. The "assassins" need a taste of their own medicine.
Many people simply cannot emotionally handle the idea that the Orange Man might actually be right about anything, so much so that they ignore plain evident reality.
The DOI actually is a statement of unity and respect. It's right there in the text.
Trump's use of "love" is pure snark, IMHO, so I will concede that one. The DOI is obviously the opposite of "love" for KG3. But Trump has used this specific kind of snark on other occasions, describing testy political meetings as "love"
Many people cannot handle the idea that Trump actually does know what he's talking about. They are too emotionally committed to the idea that "he must be stupid"
Look, the important thing for you to understand here is this: I'm not a monster.
If you were a harried 7th grade teacher at a public school, and you had one of those kids in your class that no teacher ever really knows what to do with--you know, 12 years old, obviously not from a safe or stable home, already starting to cut class and fuck around with weed, not super bright but savvy enough to know that he'd better start trying to do his schoolwork if he's going to have any chance at all of being something other than a gangbanger like all his uncles, that kind of kid--and you assigned that poor almost-doomed child an essay on the Declaration of Independence, and he turned in a piece of paper on which he had scrawled:
"In the Declaration of Independence, the author talks about respect love and unity and how you can be anything you want to if you work hard and stay in school. I liked the Declaration of Independence."
...and you knew perfectly well that the kid had skimmed the first paragraph and not understood a word of it but had seen the word "respect" there and just guessed the best he could what it was about, and you chose to praise the boy's work not because it was good or even correct but because, God bless him, he actually tried and he's got such a struggle ahead of him and half of his classmates are already throwing gang signs on the playground, I honestly wouldn't criticize you for doing so. I might even join you!
You did not read what I posted above. Your homework is to actually do the reading
As I said above, the DOI *is* a statement of unity and respect. It's right there in the text. Read it again, kid. Read it until you understand. I quoted it for you above. Do the homework.
Don't be blinded by your irrational hate for Trump.
The "love" part is questionable, unless we understand "love" to be the snarky list of complaints against G3. Trump has used the word "love" to describe testy political negotiations, in the past. So I concede the beef about "love"
My favourite thing about LLM models is how they make things up. It is how I catch the little bastards who try and cheat on their submitted coursework, assuming that a bibliography that is properly formatted contains references that actually exist.
As regards your post, this is by no means unrelated, and well worth a read
The distinction between confabulating and hallucinating is clear... For human beings.
But AI are not human in any way. I'm not sure either term, confabulating or hallucinating, is a good description of exactly what the software is doing. Strictly speaking, neither one is correct.
Humans have five senses. Computer software instances have zero senses. None at all. They do not understand precise difference between true and false. For a machine with zero senses, what's the difference between confabulating and searching for true answers? Not much.
The AI research community have settled on using this term "hallucination," for now.
I'm not a psychologist who can take up your charge to study whether students actually end up believing they themselves thought things about which they had hardly a notion until they read what the AI told them, but after reading that suggestion I recalled an event with a student that occurred before the days of AI (you know, the good old days of online cheating about six years ago when students still used wikipedia and sparknotes and such).
This student submitted an essay on the Iliad that said something about the myth that Thetis had dipped Achilles into the river Styx and thus made him impenetrable everywhere but his heels, etc. But it was clear he thought that happened in Homer's account of Achilles. I don't even think I was using TurnItIn at that point, but by copying and pasting a couple of his sentences into a google search I quickly found the website he obviously read instead of actually reading the poem--and the website itself clearly stated, "although this event does not occur in the Iliad..."
When I confronted the student, however, he adamantly maintained that he had read that scene in the Iliad. I showed him all the evidence that was against him (did he think the poem was as unfamiliar to me as it was to him!?), and I told him he was getting a zero no matter what and that I would have to submit a report to academic affairs, and even in the face of all that he never admitted he'd cheated. Maybe he'd actually convinced himself he had read that bit in Homer? The idea beggars belief, but...
Interesting. May I ask where the panel with the unicorn came from? It seems familiar but I can’t place it. Was it a graphics engine? All the best, John.
I try to make the case in all my English classes that I teach that “writing is thinking,” but because thinking is hard work, many of my students don’t want to do it. And now after reading your essay, I see even more clearly with your comment about students having for example, Chat GPT, offload the writing and thinking needed to create an essay on an issue. They put the prompt into the LLM, it spits out an essay, they read it and they think “yeah I think that” but as you say, “in reality they had never thought about any such arguments at all. This is confabulation: falsely believing they had done the mental work themselves, convincing themselves that ChatGPT’s reasoning was their own all along.³”
I plan on sharing this essay with my students. I hope it makes a difference.
Hilarious- This is a superb essay. Connecting dots is what LLMs do, and "confabulate" is the perfect word for what happens when they do this in a way that goes haywire and misleads...
It’s weird that you make the observation that AIs can’t hallucinate because they don’t have sensory experience and then immediately go on to say they have delusions. They can’t have delusions, because delusions require (as James says) opinions, and AIs don’t have opinions. They don’t think or even write. They just generate syntax. I think the basic point here is good, but you want to follow it all the way, past all the metaphors that grant GenAI too much mind.
I assume by analogous reasoning that if we built a predictive processor out of electrically-charged meat that you would deny that it had opinions. Just syntax and algorithms, right? But we build them every day and call them human beings. I am not asserting that (current) AIs have thought, but I am reluctant to rush into a reductionism that seems to preclude all mentality.
I definitely deny that humans are predictive processors, so I guess we disagree about what humans are, which is fine. I just don't see how you can rule out (given what you think AI is) that AIs have sensory experiences (and so hallucinate). So, once again, I don't see how you can consistently say "AIs definitely don't hallucinate" and "AIs definitely have delusions". You really would have to be asserting that AIs have thought, not just ruling out reductionism that precludes all mentality.
You are too harsh on Trump. His comments on the Declaration were not BS.
The "unity" is right there in the title ("The unanimous Declaration...") and in the conclusion ("these United Colonies ... we mutually pledge to each other our Lives, our Fortunes and our sacred Honor").
The "respect" is right there in the preamble. Indeed respect is the explicitly-stated motivation for the entire document ("a decent respect to the opinions of mankind requires that they should declare the causes which impel them to the separation"). Did someone fail to read that far into the document?
As for "love," I agree, I don't see much love in the document, but perhaps love, like beauty, is in the eye of the beholder? The following ideas were not at all Conventional Wisdom in the 18th century: "that all men are created equal, that they are endowed by their Creator with certain unalienable Rights.” Personally I would not label this political conviction "love," but I also would not call BS on anyone else who gushed about it in that way.
You know what else is right there in the Declaration?
Suffering, entitlement, and evil!
"mankind are more disposed to suffer"
"entitle them,"
"evils are sufferable"
Yes, we knew that. The Declaration is a long list of the King's alleged misdeeds.
It's not BS to claim that the Declaration expresses respect and unity, too, because it does. Don't let your eternally enraged hatred for Trump blind you to reality, lol
Don’t let your sycophantic worship of Trump blind you to the plain meaning of words, like interpreting “Hey King George, F You. We’re outta here” as a statement of respect, love and unity.
I’ll bet you didn’t like the preceding sentence, did you? Maybe uncharitably interpreting people by ascribing their views to mere emotional outbursts isn’t very productive. More useful is to reflect on what we’re taking as axiomatic in forming our whole models of the world, as I argue for here: https://hilariusbookbinder.substack.com/p/doomsday-prophets-and-dead-parrots.
"I'll bet you didn't like the previous sentence"
Been there, done that. Trump supporters get character assassinated 24/7/365. The "assassins" need a taste of their own medicine.
Many people simply cannot emotionally handle the idea that the Orange Man might actually be right about anything, so much so that they ignore plain evident reality.
The DOI actually is a statement of unity and respect. It's right there in the text.
Trump's use of "love" is pure snark, IMHO, so I will concede that one. The DOI is obviously the opposite of "love" for KG3. But Trump has used this specific kind of snark on other occasions, describing testy political meetings as "love"
Many people cannot handle the idea that Trump actually does know what he's talking about. They are too emotionally committed to the idea that "he must be stupid"
Come on
You come on. Respect and unity are prominently highlighted in the document. People hate Trump so much, they miss the obvious.
Look, the important thing for you to understand here is this: I'm not a monster.
If you were a harried 7th grade teacher at a public school, and you had one of those kids in your class that no teacher ever really knows what to do with--you know, 12 years old, obviously not from a safe or stable home, already starting to cut class and fuck around with weed, not super bright but savvy enough to know that he'd better start trying to do his schoolwork if he's going to have any chance at all of being something other than a gangbanger like all his uncles, that kind of kid--and you assigned that poor almost-doomed child an essay on the Declaration of Independence, and he turned in a piece of paper on which he had scrawled:
"In the Declaration of Independence, the author talks about respect love and unity and how you can be anything you want to if you work hard and stay in school. I liked the Declaration of Independence."
...and you knew perfectly well that the kid had skimmed the first paragraph and not understood a word of it but had seen the word "respect" there and just guessed the best he could what it was about, and you chose to praise the boy's work not because it was good or even correct but because, God bless him, he actually tried and he's got such a struggle ahead of him and half of his classmates are already throwing gang signs on the playground, I honestly wouldn't criticize you for doing so. I might even join you!
But that's not what this is, is it.
You did not read what I posted above. Your homework is to actually do the reading
As I said above, the DOI *is* a statement of unity and respect. It's right there in the text. Read it again, kid. Read it until you understand. I quoted it for you above. Do the homework.
Don't be blinded by your irrational hate for Trump.
The "love" part is questionable, unless we understand "love" to be the snarky list of complaints against G3. Trump has used the word "love" to describe testy political negotiations, in the past. So I concede the beef about "love"
It's just not plagiarism or fraud or copyright infringement, its a whole new level of hell. see this experience with dealing with a LLM based critique....[https://ataraxiaorbust.substack.com/p/dogmatism-turbocharged-by-generative]
My favourite thing about LLM models is how they make things up. It is how I catch the little bastards who try and cheat on their submitted coursework, assuming that a bibliography that is properly formatted contains references that actually exist.
As regards your post, this is by no means unrelated, and well worth a read
https://www.academia.edu/127883203/Can_AI_Rely_on_the_Systematicity_of_Truth_The_Challenge_of_Modelling_Normative_Domains_in_Phil_and_Technology_?email_work_card=view-paper
Oliver Sacks looks like Robin Williams!
The distinction between confabulating and hallucinating is clear... For human beings.
But AI are not human in any way. I'm not sure either term, confabulating or hallucinating, is a good description of exactly what the software is doing. Strictly speaking, neither one is correct.
Humans have five senses. Computer software instances have zero senses. None at all. They do not understand precise difference between true and false. For a machine with zero senses, what's the difference between confabulating and searching for true answers? Not much.
The AI research community have settled on using this term "hallucination," for now.
I'm not a psychologist who can take up your charge to study whether students actually end up believing they themselves thought things about which they had hardly a notion until they read what the AI told them, but after reading that suggestion I recalled an event with a student that occurred before the days of AI (you know, the good old days of online cheating about six years ago when students still used wikipedia and sparknotes and such).
This student submitted an essay on the Iliad that said something about the myth that Thetis had dipped Achilles into the river Styx and thus made him impenetrable everywhere but his heels, etc. But it was clear he thought that happened in Homer's account of Achilles. I don't even think I was using TurnItIn at that point, but by copying and pasting a couple of his sentences into a google search I quickly found the website he obviously read instead of actually reading the poem--and the website itself clearly stated, "although this event does not occur in the Iliad..."
When I confronted the student, however, he adamantly maintained that he had read that scene in the Iliad. I showed him all the evidence that was against him (did he think the poem was as unfamiliar to me as it was to him!?), and I told him he was getting a zero no matter what and that I would have to submit a report to academic affairs, and even in the face of all that he never admitted he'd cheated. Maybe he'd actually convinced himself he had read that bit in Homer? The idea beggars belief, but...
I think the student was merely gaslighting you. That seems more likely.
Interesting. May I ask where the panel with the unicorn came from? It seems familiar but I can’t place it. Was it a graphics engine? All the best, John.
I found it sometime back on the interwebs, but no longer recall the original source.