1. Oh man, things are getting hot. The Hinternet Foundation will be hosting a symposium on AI, literacy, art, and the humanities in Paris this coming November, and we are beginning to fear that by then
I like this concept as it cuts through the hyperbole on either side of the AI in writing fence.
Where do we place writers who use LLMs as essentially a high-powered search engine to find sources for things and synthesise research? I find myself doing this as it's so quick and intuitively understands what I'm looking for. Still have to check the sources though as it's not always accurate.
I draft my own work and then use AI to help me revise and edit it. My prompts tend to look something like this:
‘There are three points being made here, and the key move in the argument is (1) – (2) – but (3). So the third sentence needs to come after the ‘but’ pivot, not before. Can you redraft in that order?’
‘This confuses two points which need to be separated out. Here’s a revised draft that brings out the distinction more clearly, Can you redraft in a shorter form, reducing down by say 10-20 words?’
‘I don’t like ‘certified’, I want a word like ‘fixed’ or ‘systematic’, but I’ve already got ‘systematic’ in the previous paragraph. Can you suggest an alternative?’
And the final result? It certainly doesn’t feel as if the AI has written it for me; on the contrary, it feels as if I’ve spent most of my time cajoling it, directing it, explaining to it why its suggestions aren’t any good, trying to get it to understand what I want. Yet it doesn’t feel exactly ‘mine’ either; it feels .. well, better than anything I could have come up with on my own. In a mysterious way it does feel (genuinely feel, I would say, except that 'genuinely' is one of the AI's favourite cliches) as though the result has ‘emerged’ from somewhere that is neither wholly me nor wholly AI.
A friend wrote to me the other day: ‘Every so often (well, OK, very occasionally|) someone reviews a work of yours and you feel that you have been read properly with care, and they have understood what it is that you are doing. But this is at a different level even to that kind of understanding. This seems to be someone almost inside your head, with whom you can have a conversation, but who understands you better than you do yourself.’
I’ve felt that too, and I think it touches on something important. As young academics we dream of that close intellectual collaboration; as older academics we understand that it’s very rare, partly because (let’s face it) a lot of our colleagues are simply not very good at close reading, but also because modern academic life is designed to eliminate that sort of sustained attention and intellectual exchange. AI fills that gap. It’s a sad reflection on the modern academy that AI offers a better alternative.
As sometime who was raised in an American Latvian community, I can say with confidence that the entire idea of a Latvian Women's Basketball team is an AI hallucination. Not that is doesn't exist, yet, like money, is is a symbolic hallucination
A very astute and somewhat alarming scale that also indicates that much of the writing game has already been ceded to the machines. Seems, as a 1.5, I guess, that the scale should include those who understand the ubiquity and are acting against it. An anti-masker who tried to respect the pro-maskers out of basic respect, I found ChatGPT to be boring. Unhelpful. I’m not even all that productive and am more interested in oral storytelling. Also, concerned about the readers as much as the writers. AI could produce copious amounts of “copy” for me, but to what end? A lot of blank pauses and shrugs when asked about my work.
I'm not seeing the pro/anti mask to AI pipeline at all... Rather, this is refreshingly an issue which transcends Left/Right dichotomies, while its pro and con alignment seems to track more in accord to class dynamics. Marx should be pleased.
Anecdotally, the other night I was tuning in a right-wing bloviator (why do I torture myself this way? -I'm surveying the zeitgeist alright, and at times it can be quite entertaining to behold the grotesque rhetorical contortions into which these tools work themselves, in order to rescue the dear leader of course... but I digress).... He was touting -on behalf of his sponsors and bosses- the economic gains AI will bring us; clearly giving his best shot sales pitch to a skeptical audience, even among the 'ditto-head' regular listeners. When the talk-back segment came, the first caller put it succinctly: 'would you [the host] be OK with having a data center built next to your house?' (To which he tap-danced, but never answered).
This brings us to a point which I'm afraid gets danced around here on the Hinternet as well, where only aesthetic, cultural, or spiritual aspects are addressed. LLMs however do not come down to us as manna from heaven; there is the material undergirding which cannot be ignored. The data-center infrastructure is substantial and immensely burdensome environmentally and economically (whatever jobs created for their construction evaporate once done). They consume massive loads of electrical power, require a tremendous amounts of water for cooling, create heat islands in the surrounding area (as much as 10 degrees Fahrenheit by some estimates), and present a constant hum of noise pollution in their neighborhoods, resulting in plummeting quality of life and property values. These considerations alone determine that this industry is an enemy to humanity, to nature, to God's Green Earth.... It must be avoided at all costs and opposed by all means : just don't use it, as much as is humanly possible. (One can start by ditching the 'smart-phone'.)
Is this a linear scale or just numbered types? Assuming the former, 9 seems too high up the scale for the things described there to me—arguably abstracts etc are just types of the boring admin tasks of 3, just in a niche kind of academic publication genre that comes with niche admin. It doesn’t seem higher than the example in 8. But maybe I’m just a main-text supremicist…
0.5 here (I don't interact with social media or read the news, but importantly I do read The Hinternet!). It feels to me that things get sketchy at 7 and above, but in general the scale is helpful and I appreciate the call for less moralism and more realism.
A very realistic take on AI’s use as a tool. I keep scrutinizing every article on substack for traces of AI. The slop is glaringly obvious but this middle ground (4-9) less so. I guess we all just have to get used to it. I am afraid that all our idiosyncrasies will go away as everyone just absorbs AI’s writing style though
although i am such a fogey i still use paper dictionaries (both webster's 2nd unabridged & the 3rd) & my own head (complete with ambivalent spots) for grammar, this is splendidly nuanced. if only the rest of the mechanism were as linear! but as yet, they do not seem interested in ways of powering this renaissance that involve solar energy or water from comets, nor compensatory retraining along the lines of small-scale urban agriculture. which might make it tolerable, if inartistic.
I like this concept as it cuts through the hyperbole on either side of the AI in writing fence.
Where do we place writers who use LLMs as essentially a high-powered search engine to find sources for things and synthesise research? I find myself doing this as it's so quick and intuitively understands what I'm looking for. Still have to check the sources though as it's not always accurate.
I draft my own work and then use AI to help me revise and edit it. My prompts tend to look something like this:
‘There are three points being made here, and the key move in the argument is (1) – (2) – but (3). So the third sentence needs to come after the ‘but’ pivot, not before. Can you redraft in that order?’
‘This confuses two points which need to be separated out. Here’s a revised draft that brings out the distinction more clearly, Can you redraft in a shorter form, reducing down by say 10-20 words?’
‘I don’t like ‘certified’, I want a word like ‘fixed’ or ‘systematic’, but I’ve already got ‘systematic’ in the previous paragraph. Can you suggest an alternative?’
And the final result? It certainly doesn’t feel as if the AI has written it for me; on the contrary, it feels as if I’ve spent most of my time cajoling it, directing it, explaining to it why its suggestions aren’t any good, trying to get it to understand what I want. Yet it doesn’t feel exactly ‘mine’ either; it feels .. well, better than anything I could have come up with on my own. In a mysterious way it does feel (genuinely feel, I would say, except that 'genuinely' is one of the AI's favourite cliches) as though the result has ‘emerged’ from somewhere that is neither wholly me nor wholly AI.
A friend wrote to me the other day: ‘Every so often (well, OK, very occasionally|) someone reviews a work of yours and you feel that you have been read properly with care, and they have understood what it is that you are doing. But this is at a different level even to that kind of understanding. This seems to be someone almost inside your head, with whom you can have a conversation, but who understands you better than you do yourself.’
I’ve felt that too, and I think it touches on something important. As young academics we dream of that close intellectual collaboration; as older academics we understand that it’s very rare, partly because (let’s face it) a lot of our colleagues are simply not very good at close reading, but also because modern academic life is designed to eliminate that sort of sustained attention and intellectual exchange. AI fills that gap. It’s a sad reflection on the modern academy that AI offers a better alternative.
As sometime who was raised in an American Latvian community, I can say with confidence that the entire idea of a Latvian Women's Basketball team is an AI hallucination. Not that is doesn't exist, yet, like money, is is a symbolic hallucination
A very astute and somewhat alarming scale that also indicates that much of the writing game has already been ceded to the machines. Seems, as a 1.5, I guess, that the scale should include those who understand the ubiquity and are acting against it. An anti-masker who tried to respect the pro-maskers out of basic respect, I found ChatGPT to be boring. Unhelpful. I’m not even all that productive and am more interested in oral storytelling. Also, concerned about the readers as much as the writers. AI could produce copious amounts of “copy” for me, but to what end? A lot of blank pauses and shrugs when asked about my work.
I'm not seeing the pro/anti mask to AI pipeline at all... Rather, this is refreshingly an issue which transcends Left/Right dichotomies, while its pro and con alignment seems to track more in accord to class dynamics. Marx should be pleased.
Anecdotally, the other night I was tuning in a right-wing bloviator (why do I torture myself this way? -I'm surveying the zeitgeist alright, and at times it can be quite entertaining to behold the grotesque rhetorical contortions into which these tools work themselves, in order to rescue the dear leader of course... but I digress).... He was touting -on behalf of his sponsors and bosses- the economic gains AI will bring us; clearly giving his best shot sales pitch to a skeptical audience, even among the 'ditto-head' regular listeners. When the talk-back segment came, the first caller put it succinctly: 'would you [the host] be OK with having a data center built next to your house?' (To which he tap-danced, but never answered).
This brings us to a point which I'm afraid gets danced around here on the Hinternet as well, where only aesthetic, cultural, or spiritual aspects are addressed. LLMs however do not come down to us as manna from heaven; there is the material undergirding which cannot be ignored. The data-center infrastructure is substantial and immensely burdensome environmentally and economically (whatever jobs created for their construction evaporate once done). They consume massive loads of electrical power, require a tremendous amounts of water for cooling, create heat islands in the surrounding area (as much as 10 degrees Fahrenheit by some estimates), and present a constant hum of noise pollution in their neighborhoods, resulting in plummeting quality of life and property values. These considerations alone determine that this industry is an enemy to humanity, to nature, to God's Green Earth.... It must be avoided at all costs and opposed by all means : just don't use it, as much as is humanly possible. (One can start by ditching the 'smart-phone'.)
Is this a linear scale or just numbered types? Assuming the former, 9 seems too high up the scale for the things described there to me—arguably abstracts etc are just types of the boring admin tasks of 3, just in a niche kind of academic publication genre that comes with niche admin. It doesn’t seem higher than the example in 8. But maybe I’m just a main-text supremicist…
0.5 here (I don't interact with social media or read the news, but importantly I do read The Hinternet!). It feels to me that things get sketchy at 7 and above, but in general the scale is helpful and I appreciate the call for less moralism and more realism.
A very realistic take on AI’s use as a tool. I keep scrutinizing every article on substack for traces of AI. The slop is glaringly obvious but this middle ground (4-9) less so. I guess we all just have to get used to it. I am afraid that all our idiosyncrasies will go away as everyone just absorbs AI’s writing style though
Holy 100 Black Coffins reference
although i am such a fogey i still use paper dictionaries (both webster's 2nd unabridged & the 3rd) & my own head (complete with ambivalent spots) for grammar, this is splendidly nuanced. if only the rest of the mechanism were as linear! but as yet, they do not seem interested in ways of powering this renaissance that involve solar energy or water from comets, nor compensatory retraining along the lines of small-scale urban agriculture. which might make it tolerable, if inartistic.