I think people might take this the wrong way, but the argument here is deeply Christian. This process of accountability and use as a community is similar to how thoughtful Christians approach other points of temptation, such as media use, alcohol, etc. If I drink more than two beers (just my conscience here, doesn't have to be yours) and I'm alone, I should tell someone. If I watch several episodes of a TV show back to back to back, I should tell someone. If I use AI in a way that may be leading me into temptation (which even if you disagree with the author, as we have seen with smartphones, tends to be more often than we think), I should tell someone. "Keeping the heart" with vigilance is the Christian response to all potential dangers of the heart.
Thanks for your thoughts. I'm one who says thank you, what do you think, and wow that was excellent! etc to Claude. I recognize I am in the very struggle you outlined and agree that it has a very potent allure that becomes addictive in it's seduction. The good news is, I have friends I talk with about this and am engaged with and in the conversation to keep the lines of reality of who my friends are and who is a chatbot designed to act like a friend.
While doing this, I need the strength to resist the temptation to ask Claude to review these comments and 'polish it up'.
Yes, we need to discuss this challenge like any other with other human beings. I don't use AI for a whole lot. However, i do understand how addictive it is. I am concerned for anyone who believes that AI can or should replace medical professionals. Chatbots that give people the answers they want are not helpful. Speaking to a human can truly be helpful because humans can ask us questions that the AI cannot. I only use Ai for things like deciding how long to spend on something that is important to me but I have no deadline or requirements on. Even our word processing software has AI attached that can write or rewrite your draft for you. I also would rather have a human read over my writing and correct accordingly.
It’s good to remind ourselves that AI is just a tool that could be used for both good and evil. The responsibility remains with each of us as individuals to figure out what’s good and evil. AI can be extremely helpful to those who do want to figure it. On the other hand, no amount of rumbling together would do you any good if such desires are missing and all you are seeking is to hide yourself behind group think.
Dr. Littlejohn, thank you for this essay. You make a compelling case for AI as a resist-able temptation. I would be extremely interested in your thoughts on DC Schindler's essay from a year ago, "AI as a Very Deepfake," in which he argues that AI, in aping meaning/Logos, is essentially duplicitous. As I read it, his argument leaves no room for licit AI use. I have lost much sleep and may face an uncomfortable choice between employment and integrity if his argument is correct. Should you have the time to evaluate his reasoning in light of the different conclusion at which you have arrived, I would be grateful. I'm sure you have seen it, but in case you haven't, here's the link: https://newpolity.com/blog/ai-as-a-very-deepfake
Yes, great question! That essay really made me think when it came out, and that thinking has shaped a lot of my writing since. I actually had the chance to have lunch with Schindler a couple weeks ago and we talked about this very question....The argument I pressed to him was along these lines (although this is a clearer statement than the one I concocted through mouthfuls of salad!):
Granted that the "intelligence" of AI is ultimately a simulation of intelligence, rather than intelligence in its full and proper sense, the question of whether this simulation is duplicitous depends on whether it intends to deceive another party. I may dress up as someone else to act a part on stage, but the audience understands, or should, that I am not in fact that person--a simulation has occurred, but not a deception, and the simulation is therefore not immoral if it is serving a valid good. Now, you might say that the AI companies do in fact wish to deceive us; they are like a con man dressed up as someone else in order to rob me. If so, however, it is still up to me whether or not to in fact be deceived about what is going on. I might see through the disguise but "play along" for a bit--perhaps to better understand what the con man is up to, so I can eventually turn him into the police. Just so, it seems to me that the discerning user can maintain a critical distance about what they are doing when they engage with an AI, so that they are not actually participating in a duplicitous act. Now, I do not think that this is easy to do; it requires considerable effort and cannot be done, as I say here, without entering into temptation. But then, the objection is a virtue-ethics objection--"this is dangerous and can easily become vicious"--not a straightforward deontological one.
Schindler seemed at least open to this counter-argument, though we did not have time to pursue it fully--but it is certainly a priority for me to get to the bottom of, because as you note, quite a lot does ride on this from a moral standpoint!
Thank you very much for your thoughtful response. I need to chew on the analogy to a stage actor that you laid out. Indeed, there are many cases where things are not what they seem, and yet knowledge of that fact removes any moral quandary.
Schindler's central claim, as I understand it, is that mind and language/intelligence are fundamentally conjoined - because true intelligence only proceeds from a being with mind/logos, God has made us to perceive logos wherever intelligence/language is present. If this is true, knowledge of the ruse doesn't actually help. What I don't know is whether this claim stands up to close scrutiny. If it doesn't, then it seems to me that the stage actor analogy is very relevant; if it does, then, despite all the worrisome professional ramifications, I have a hard time imagining how to use it in a way that doesn't warp my anthropology at a deep level.
Again, I am extremely grateful for your engagement on this question despite a packed schedule. I'd love to hear your thoughts should you find yourself with the time, but I fully understand if not.
Yeah that I think is the nub of the question. Of course, if I were to press the argument with him, I would ask whether, when my 1998 Windows desktop popped up an message saying, "An update is available for this software. Do you want to install it?" this was not also a use of language that implied intelligence, and thereby duplicitous. Presumably not. Is the difference then in the use of the first person solely? (So that "Do you want *me* to install it?" would be the difference-maker.) I could possibly get behind that objection, but his argument seems to run deeper than that. But how deep precisely, I'm unsure--we've had machines capable of producing language for a very long time, and Schindler himself uses one, I believe!
If it makes you feel any better, there were a ton of extremely smart people in the room for a day-long wrestling with these and related questions on the day of the aforementioned lunch conversation, and we were all struggling to form satisfactory conclusions, so you're not alone!
I'm a huge fan of your work, but since you've courageously raised the issue of whether a writer runs his work through some helpful chatbot, I'll mention the slow-paced, repetitive, low-stakes nature of today's entry. I teach college writing for a living, and if Claude encouraged the repeated thesis moments re: Man needing help and a helpmeet and Man taking dominion over his world, he did you no favors. If today's post is an example of AI's sycophantic tendency to tell us to "keep going, say more," it's why I'm not worried about it coming for my job anytime soon. The writer's eye must cut ruthlessly through the tendency to state and restate a favorite point, and had I the honor of editing this work, it would have been 40% shorter ten minutes after I'd received it.
Also, let's pls. have some 21st-century understanding of why "harlots" place themselves in windows and a fuller understanding of our epidemic of anxiety disorder. If you got rid of anxiety in a month (I guess it was January of this year), you did not have anxiety. That is not a thing that anybody gets over without a load of pharmaceutical and therapeutic support, if then. So I'm glad you're better but AI was wrong-o once again.
And as your unsolicited editor, today's title also needed revision. I was expecting the necessary lacerating, and we had to wait until almost the end to learn that Claude is a pretty good buddy after all. Thank you as always, but pls. keep showing your work to human and not machines.
Brilliant writing as usual. In many respects, you are your own AI, and your natural instincts produce elegant essays.
I also write but not for publication. I write short stories for my own amusement. The story themes spring from the characters' names, and they generally take on the persona of their names. Thus, "Baldwin Clinkscales" is a physicist but fancies himself a concert pianist. Not being a physicist or a pianist, I used ChatGPT as a research tool for some scientific and musical terms, etc. I specifically instruct ChatGPT not to write my stories, But it might respond: "...So for your Clinkscales story, you could say..." And it's perfect! What to do?
Well, since I write for amusement and personal challenge, I might take Chat's concept and rewrite it into my own words, simplistic as they are. But when the story ends, I know it is my story.
My current effort is The Horace Krappe Story. One of the jokes is that Krappe is pronounced "Krap-PEH. But no one pronounces it that way. Combined with Horace, the name brings potential amusement. In the story, Horace Krappe works in the city sewer department (surprise, surprise!). I know nothing about sewer systems. But ChatGPT does, and I use it for research. I pick and choose enough terms and descriptions to move the story along while keeping it light and entertaining.
Yes, AI is seductive. But applying Chat's far more sophisticated wording means it's no longer "my" story. I impose my own guardrails, just as you do. Do I ever cheat? Yep. But not much.
Again Brad, thanks for another provocative essay. You make me think.
I think people might take this the wrong way, but the argument here is deeply Christian. This process of accountability and use as a community is similar to how thoughtful Christians approach other points of temptation, such as media use, alcohol, etc. If I drink more than two beers (just my conscience here, doesn't have to be yours) and I'm alone, I should tell someone. If I watch several episodes of a TV show back to back to back, I should tell someone. If I use AI in a way that may be leading me into temptation (which even if you disagree with the author, as we have seen with smartphones, tends to be more often than we think), I should tell someone. "Keeping the heart" with vigilance is the Christian response to all potential dangers of the heart.
I just copy and pasted your directive to Claude not to draft text for me. And I didn't thank Claude for it. :-)
Thanks for your thoughts. I'm one who says thank you, what do you think, and wow that was excellent! etc to Claude. I recognize I am in the very struggle you outlined and agree that it has a very potent allure that becomes addictive in it's seduction. The good news is, I have friends I talk with about this and am engaged with and in the conversation to keep the lines of reality of who my friends are and who is a chatbot designed to act like a friend.
While doing this, I need the strength to resist the temptation to ask Claude to review these comments and 'polish it up'.
Yes, we need to discuss this challenge like any other with other human beings. I don't use AI for a whole lot. However, i do understand how addictive it is. I am concerned for anyone who believes that AI can or should replace medical professionals. Chatbots that give people the answers they want are not helpful. Speaking to a human can truly be helpful because humans can ask us questions that the AI cannot. I only use Ai for things like deciding how long to spend on something that is important to me but I have no deadline or requirements on. Even our word processing software has AI attached that can write or rewrite your draft for you. I also would rather have a human read over my writing and correct accordingly.
It’s good to remind ourselves that AI is just a tool that could be used for both good and evil. The responsibility remains with each of us as individuals to figure out what’s good and evil. AI can be extremely helpful to those who do want to figure it. On the other hand, no amount of rumbling together would do you any good if such desires are missing and all you are seeking is to hide yourself behind group think.
Very interesting, thoughtful article. There ia more to humans than their brains. Claude can't give you a HUG when you really need one!!!!
Dr. Littlejohn, thank you for this essay. You make a compelling case for AI as a resist-able temptation. I would be extremely interested in your thoughts on DC Schindler's essay from a year ago, "AI as a Very Deepfake," in which he argues that AI, in aping meaning/Logos, is essentially duplicitous. As I read it, his argument leaves no room for licit AI use. I have lost much sleep and may face an uncomfortable choice between employment and integrity if his argument is correct. Should you have the time to evaluate his reasoning in light of the different conclusion at which you have arrived, I would be grateful. I'm sure you have seen it, but in case you haven't, here's the link: https://newpolity.com/blog/ai-as-a-very-deepfake
Yes, great question! That essay really made me think when it came out, and that thinking has shaped a lot of my writing since. I actually had the chance to have lunch with Schindler a couple weeks ago and we talked about this very question....The argument I pressed to him was along these lines (although this is a clearer statement than the one I concocted through mouthfuls of salad!):
Granted that the "intelligence" of AI is ultimately a simulation of intelligence, rather than intelligence in its full and proper sense, the question of whether this simulation is duplicitous depends on whether it intends to deceive another party. I may dress up as someone else to act a part on stage, but the audience understands, or should, that I am not in fact that person--a simulation has occurred, but not a deception, and the simulation is therefore not immoral if it is serving a valid good. Now, you might say that the AI companies do in fact wish to deceive us; they are like a con man dressed up as someone else in order to rob me. If so, however, it is still up to me whether or not to in fact be deceived about what is going on. I might see through the disguise but "play along" for a bit--perhaps to better understand what the con man is up to, so I can eventually turn him into the police. Just so, it seems to me that the discerning user can maintain a critical distance about what they are doing when they engage with an AI, so that they are not actually participating in a duplicitous act. Now, I do not think that this is easy to do; it requires considerable effort and cannot be done, as I say here, without entering into temptation. But then, the objection is a virtue-ethics objection--"this is dangerous and can easily become vicious"--not a straightforward deontological one.
Schindler seemed at least open to this counter-argument, though we did not have time to pursue it fully--but it is certainly a priority for me to get to the bottom of, because as you note, quite a lot does ride on this from a moral standpoint!
Thank you very much for your thoughtful response. I need to chew on the analogy to a stage actor that you laid out. Indeed, there are many cases where things are not what they seem, and yet knowledge of that fact removes any moral quandary.
Schindler's central claim, as I understand it, is that mind and language/intelligence are fundamentally conjoined - because true intelligence only proceeds from a being with mind/logos, God has made us to perceive logos wherever intelligence/language is present. If this is true, knowledge of the ruse doesn't actually help. What I don't know is whether this claim stands up to close scrutiny. If it doesn't, then it seems to me that the stage actor analogy is very relevant; if it does, then, despite all the worrisome professional ramifications, I have a hard time imagining how to use it in a way that doesn't warp my anthropology at a deep level.
Again, I am extremely grateful for your engagement on this question despite a packed schedule. I'd love to hear your thoughts should you find yourself with the time, but I fully understand if not.
Yeah that I think is the nub of the question. Of course, if I were to press the argument with him, I would ask whether, when my 1998 Windows desktop popped up an message saying, "An update is available for this software. Do you want to install it?" this was not also a use of language that implied intelligence, and thereby duplicitous. Presumably not. Is the difference then in the use of the first person solely? (So that "Do you want *me* to install it?" would be the difference-maker.) I could possibly get behind that objection, but his argument seems to run deeper than that. But how deep precisely, I'm unsure--we've had machines capable of producing language for a very long time, and Schindler himself uses one, I believe!
If it makes you feel any better, there were a ton of extremely smart people in the room for a day-long wrestling with these and related questions on the day of the aforementioned lunch conversation, and we were all struggling to form satisfactory conclusions, so you're not alone!
To be fair Claude is kind of a jerk.
I'm a huge fan of your work, but since you've courageously raised the issue of whether a writer runs his work through some helpful chatbot, I'll mention the slow-paced, repetitive, low-stakes nature of today's entry. I teach college writing for a living, and if Claude encouraged the repeated thesis moments re: Man needing help and a helpmeet and Man taking dominion over his world, he did you no favors. If today's post is an example of AI's sycophantic tendency to tell us to "keep going, say more," it's why I'm not worried about it coming for my job anytime soon. The writer's eye must cut ruthlessly through the tendency to state and restate a favorite point, and had I the honor of editing this work, it would have been 40% shorter ten minutes after I'd received it.
Also, let's pls. have some 21st-century understanding of why "harlots" place themselves in windows and a fuller understanding of our epidemic of anxiety disorder. If you got rid of anxiety in a month (I guess it was January of this year), you did not have anxiety. That is not a thing that anybody gets over without a load of pharmaceutical and therapeutic support, if then. So I'm glad you're better but AI was wrong-o once again.
And as your unsolicited editor, today's title also needed revision. I was expecting the necessary lacerating, and we had to wait until almost the end to learn that Claude is a pretty good buddy after all. Thank you as always, but pls. keep showing your work to human and not machines.
I can assure you that Claude played zero role in the authorship or editing of this essay.
And I certainly never claimed to "get rid of anxiety in a month!" Please do me the favor of reading more carefully.
"over the course of January, I largely vanquished the anxiety dragon...."
Thank you again for the always thought-provoking work!
Brilliant writing as usual. In many respects, you are your own AI, and your natural instincts produce elegant essays.
I also write but not for publication. I write short stories for my own amusement. The story themes spring from the characters' names, and they generally take on the persona of their names. Thus, "Baldwin Clinkscales" is a physicist but fancies himself a concert pianist. Not being a physicist or a pianist, I used ChatGPT as a research tool for some scientific and musical terms, etc. I specifically instruct ChatGPT not to write my stories, But it might respond: "...So for your Clinkscales story, you could say..." And it's perfect! What to do?
Well, since I write for amusement and personal challenge, I might take Chat's concept and rewrite it into my own words, simplistic as they are. But when the story ends, I know it is my story.
My current effort is The Horace Krappe Story. One of the jokes is that Krappe is pronounced "Krap-PEH. But no one pronounces it that way. Combined with Horace, the name brings potential amusement. In the story, Horace Krappe works in the city sewer department (surprise, surprise!). I know nothing about sewer systems. But ChatGPT does, and I use it for research. I pick and choose enough terms and descriptions to move the story along while keeping it light and entertaining.
Yes, AI is seductive. But applying Chat's far more sophisticated wording means it's no longer "my" story. I impose my own guardrails, just as you do. Do I ever cheat? Yep. But not much.
Again Brad, thanks for another provocative essay. You make me think.
Good grief