Not only did this post perfectly articulate one of my many frustrations with AI and its accompanying hype, but it also got that pesky earworm "On Top of Spaghetti" stuck in my head for the first time in three decades :) John: thank you for your always insightful, timely, and trenchant commentary! As an 11th and 12th grade history teacher who grew increasingly horrified over the past 3 years at how AI was rendering some of my favorite assignments obsolete, you've been so helpful at putting these tumultuous times in context.
I'm not too worried because I think the kids will figure this out quicker than the adults.
I talked to a high school student recently and he told me he's had enough with AI already 😂 for the simple reason that the AI he's been training for a year still make simple calculation mistakes within math problems.
But this kid also told me he avoids most social media too, so maybe he's just an outlier?
I think novelty is just the start of m i8y disatisfactions with AI. The fact it is not factually artificial intelligence is next. I've used AI extensively to do dull jobs that I don't want to do such a searching through work for weak verbs and suggesting potential strong verbs. Surprise! It's not very good at that, producing the venerable "usage" as a substitute for the weak verb "use". Still weak, but longer. Well done. For a long time in my life I've taught technology with a business analysis focus, and from the start this has struck me as being a technology with some very real applications, which are mostly specialist, and some very shitty general applications, which are mostly useless as they don't teach anyone anything and in fact short cut the process of learning to where you don't learn anything at all. Before I retired, the management of the educational business I worked for were very excited about AI. I asked them what utility it had for them. After some time the best they could cover was it would write the catalogues. I pointed out that this was a trivial job as essentially we taught the same courses year after year with at most a 20% churn of new courses (even that is debatable as a lot of it is simply recycled old ideas). At best it is a technology looking for a use case. I offer you the myriad robots who seem to have one skill: falling over, and the AI tech in my phone which mostly misspells things from dictation, including inventing new words, and gives me Google summaries of things I don't want summaries of. We can however take a privileged view of the kind of intelligences that WILL find AI endlessly entertaining, by looking at the AI slop generated by the idiots inhabiting the White House. If that makes you feel positive about the future, so be it. Perhaps post-Trump the White House could be rehabilitated into a facility for helping reintroduce people to the real world, books etc.
I still remember when my students first discovered that they could use their computer program to render their writing assignments in any font in any color. "I'm sorry, Pat, but I cannot grade your paper when it's printed in wedding invite scrip in yellow. Please print again in something simple in black." Or the ever popular, "Mr. Greene, don't you want to see what this essay looks like in wingdings?" The novelty could not wear off fast enough.
"Hello Muddah, Hello Fadduh" & "They're Coming to Take Me Away" were the annoying novelty songs of my youth that I enjoyed singing. Hope that remembering them now won't haunt me throughout today.
I love how you use novelty songs as an argument against shiny new initiatives or EdTech. I am already seeing that with AI, these “novelty songs” are designed to be so enticing, many decision-making adults, are pushing AI curriculum slop and AI platforms onto teachers. It seems like the less I use AI, the more easily I can recognize its slop.
It's the rare artist who can make a novelty song that lasts - "Tiger Woods" by Dan Bern fits with the content of some of the songs you mentioned here, but his song "Jerusalem" is really quite brilliant.
Always happy to get another Marlon James novel, but is he planning to finish the Dark Star trilogy?
Your comments on novelty are spot on: dancing bears. (The marvel is not that the bear dances well, but that the bear dances at all.)
It is interesting, though, to try to diagnose what is missing: LLMs don't have a world model and they don't have a theory of mind and they don't have any particular intentionality...but their responses make it seem like they do. (I think this is more or less echoing what you've been saying.)
The problem is that we humans are failing the Turing test: we imagine the LLMs have understanding by virtue of their percieved facility with language. (Like seeing faces in the patterns on burnt toast.)
Great post! People are always seeking new novelty, but old novelty constantly wears off quickly, and neither provide substance. BTW, another insight I stumbled across recently while writing my own post on the subject is that AI is a societal regression from information farming to information foraging.That is, all of these AI agents are happily built on various troves of existing knowledge, but none of these tools provide any ways to contribute and store new knowledge.
I have a small bone to pick with you: Dumb Ditties may have been more impactful than Sora, but it doesn't deserve credit for launching "On Top of Spaghetti" into the public sphere. Tom Glazer recorded the song in 1963. I remember annoying my parents with it in or around 1972.
I remember that early in the days of digital humanities someone described it as “using computers to read books.” While that’s not an entirely fair definition, it’s one that I think about a lot when thinking about machine learning. Are we using the machines? Are we the machines? Are the machines using us? What do we gain from having machines read books? Quite a bit, if you’re talking concordances and full text searching. What do we lose?
(On a separate note, I would be remiss in my duties if I did not mention Disney yet again licensing “its” characters: Disney built its empire on the public domain and now wants to lock that up with mutant contract law and maximalist copyright for all eternity.)
Not only did this post perfectly articulate one of my many frustrations with AI and its accompanying hype, but it also got that pesky earworm "On Top of Spaghetti" stuck in my head for the first time in three decades :) John: thank you for your always insightful, timely, and trenchant commentary! As an 11th and 12th grade history teacher who grew increasingly horrified over the past 3 years at how AI was rendering some of my favorite assignments obsolete, you've been so helpful at putting these tumultuous times in context.
I'm not too worried because I think the kids will figure this out quicker than the adults.
I talked to a high school student recently and he told me he's had enough with AI already 😂 for the simple reason that the AI he's been training for a year still make simple calculation mistakes within math problems.
But this kid also told me he avoids most social media too, so maybe he's just an outlier?
I think novelty is just the start of m i8y disatisfactions with AI. The fact it is not factually artificial intelligence is next. I've used AI extensively to do dull jobs that I don't want to do such a searching through work for weak verbs and suggesting potential strong verbs. Surprise! It's not very good at that, producing the venerable "usage" as a substitute for the weak verb "use". Still weak, but longer. Well done. For a long time in my life I've taught technology with a business analysis focus, and from the start this has struck me as being a technology with some very real applications, which are mostly specialist, and some very shitty general applications, which are mostly useless as they don't teach anyone anything and in fact short cut the process of learning to where you don't learn anything at all. Before I retired, the management of the educational business I worked for were very excited about AI. I asked them what utility it had for them. After some time the best they could cover was it would write the catalogues. I pointed out that this was a trivial job as essentially we taught the same courses year after year with at most a 20% churn of new courses (even that is debatable as a lot of it is simply recycled old ideas). At best it is a technology looking for a use case. I offer you the myriad robots who seem to have one skill: falling over, and the AI tech in my phone which mostly misspells things from dictation, including inventing new words, and gives me Google summaries of things I don't want summaries of. We can however take a privileged view of the kind of intelligences that WILL find AI endlessly entertaining, by looking at the AI slop generated by the idiots inhabiting the White House. If that makes you feel positive about the future, so be it. Perhaps post-Trump the White House could be rehabilitated into a facility for helping reintroduce people to the real world, books etc.
Sorry about any errors in this. Apparently it's too difficult to put an edit button anywhere on the page.
Check under the three-dot menu.
I still remember when my students first discovered that they could use their computer program to render their writing assignments in any font in any color. "I'm sorry, Pat, but I cannot grade your paper when it's printed in wedding invite scrip in yellow. Please print again in something simple in black." Or the ever popular, "Mr. Greene, don't you want to see what this essay looks like in wingdings?" The novelty could not wear off fast enough.
"Hello Muddah, Hello Fadduh" & "They're Coming to Take Me Away" were the annoying novelty songs of my youth that I enjoyed singing. Hope that remembering them now won't haunt me throughout today.
Well, they’re haunting me now!
I love how you use novelty songs as an argument against shiny new initiatives or EdTech. I am already seeing that with AI, these “novelty songs” are designed to be so enticing, many decision-making adults, are pushing AI curriculum slop and AI platforms onto teachers. It seems like the less I use AI, the more easily I can recognize its slop.
It's the rare artist who can make a novelty song that lasts - "Tiger Woods" by Dan Bern fits with the content of some of the songs you mentioned here, but his song "Jerusalem" is really quite brilliant.
Always happy to get another Marlon James novel, but is he planning to finish the Dark Star trilogy?
Your comments on novelty are spot on: dancing bears. (The marvel is not that the bear dances well, but that the bear dances at all.)
It is interesting, though, to try to diagnose what is missing: LLMs don't have a world model and they don't have a theory of mind and they don't have any particular intentionality...but their responses make it seem like they do. (I think this is more or less echoing what you've been saying.)
The problem is that we humans are failing the Turing test: we imagine the LLMs have understanding by virtue of their percieved facility with language. (Like seeing faces in the patterns on burnt toast.)
Great post! People are always seeking new novelty, but old novelty constantly wears off quickly, and neither provide substance. BTW, another insight I stumbled across recently while writing my own post on the subject is that AI is a societal regression from information farming to information foraging.That is, all of these AI agents are happily built on various troves of existing knowledge, but none of these tools provide any ways to contribute and store new knowledge.
I have a small bone to pick with you: Dumb Ditties may have been more impactful than Sora, but it doesn't deserve credit for launching "On Top of Spaghetti" into the public sphere. Tom Glazer recorded the song in 1963. I remember annoying my parents with it in or around 1972.
John, to me personally, the thing about AI is that "On Top of Spaghetti" was a deeply beloved part of my childhood. Confirmed Luddite, is it obvious?
I remember that early in the days of digital humanities someone described it as “using computers to read books.” While that’s not an entirely fair definition, it’s one that I think about a lot when thinking about machine learning. Are we using the machines? Are we the machines? Are the machines using us? What do we gain from having machines read books? Quite a bit, if you’re talking concordances and full text searching. What do we lose?
(On a separate note, I would be remiss in my duties if I did not mention Disney yet again licensing “its” characters: Disney built its empire on the public domain and now wants to lock that up with mutant contract law and maximalist copyright for all eternity.)