Like so much else these days, you're really talking about moral purity when it comes to AI --- the secular version of the old theological idea that there is something sacred about human activity that must not be contaminated by the artificial. But this is where the hypocrisy comes in --- people rarely apply this purity logic consistently. Most didn't invoke moral purity when: ATMs replaced bank tellers, GPS replaced mapmakers or spellcheck replaced proofreaders. So, not so much "morality" as class politics wearing a moral mask. The purity argument collapses because humans have always outsourced parts of themselves: memory to writing; navigation to maps; calculation to machines and perception to sensors. AI is just the next outsourcing. The discomfort isn’t really about morality --- it’s about proximity. And as I've said before, the core issues are what parts of our cognitive makeup (individually and collectively) we're willing to outsource to AI and their social and economic consequences.
In addition to all the moral points around replacing human effort (which Dale of Green Gables has answered well, here, already); it is the energy/environmental impact that troubles me. None of the historical analogies carried the global climatic impacts that this new technology does, and it's not at all clear how we will handle that.
There are two parts to this: Training, and Using AI LLMs.
TRAINING
The training run for just one recent iteration of ChatGPT from Open AI (only one of many LLMs by multiple corporations) is estimated to have required about 11 BILLION kilowatt-hours (kWhr) of electric energy. Just to train it, before use. Is that a lot? Here are some comparables:
How much electricity is 11 billion kWh?
1 million homes for a year: The average U.S. home consumes 10,500 kWh a year. That means 11 billion kWh could power one million average U.S. households for a full year.
The U.S. Steel Industry: The entire U.S. steel industry currently consumes about 11 billion kWh a year.
11 billion miles by an EV: A Tesla Model 3 uses about 25 kWh to travel 100 miles. With 11 billion kWh, you could drive that Tesla for a mind-blowing 44 billion miles—roughly three roundtrips to Neptune.
More than the output of a 1 GW nuclear reactor for a year. A 1 GW nuclear reactor can produce about 8 billion kWh per year.
So, yeah. It's a lot. And that meant burning a BIGGER lot of fossil fuels to generate it. You may have noticed this Summer is getting a tad warm already. Could be related....
RUNNING/USING:
This one is tougher, as it depends much more directly on the nature of the queries and the size of the LLM queried. I won't even try, here in this space, but rather just note that if just 1/10 of every person on earth does this once a day (minimal use in developed countries) and estimating an ordinary query uses just 1 kWhr, then that's nearly another BILLION kWhr. Per Day.
If you are serious about understanding this in more depth, may I recommend this article:
I don't know about you, but I know that Earth's climate and most societies have survived all prior insults; but I'm not so sure they will do so this time around. I get worrying about lost jobs; but how about lost livable planet (livable for humans, anyway - cockroaches and many weeds will likely be OK).
Well said. As the writer of the letter that GW mentions in this column, I did mention the environmental problems in my objections to AI, but won't reiterate this in the comments, as you do a fine job of it. Still, these issues are at the root of my reluctance to use ChatGPT. (For that matter, I've also given up on plane travel and car-driving--but lest one find my decision offensive, I hasten to add that it's a personal one, and that I don't impose it on others.)
Noble, nonetheless. Unless that has reduced your worthy output more than your consumption!
As a young engineer in the energy conversion/conservation field, I set a goal of energetic break-even (or better) through my work. That is, I intended to save more energy by the fruits of my labors than I consumed by existing. It's funny, but I quite (hubristically) assumed it would be achieved by developing some wonderful new engine, power system, or the like. In fact, it was through the development of double-cellular insulating window shades! Only a small savings for each window, but many millions of windows!
BTW, I still drive a car, but it's electric, and the PV on the barn produces more than twice the car's use - it literally runs on sunshine.
As the “Letterwriter” that GW refers to, I’ve made my point and appreciate his thoughtful response. I still disagree with a lot of the above, but since the issue cannot be solved in a single discussion, I’ll essentially sit this one out. I'll just make a few adjustments here, just for the sake of clarity:
1) I don't object to all AI, but I mind inconsistency, like GW calling ChatGPT’s art “ridiculous,” “dull” and ham-fisted” on June 7, and then saying it’s “riveting” in a post-column comment the next day.
2) I also note that GW refuses to produce a column with AI, but dispenses with this principle when art is involved. Does this mean artists like myself can be thrown under the bus, but not writers?
3) I also have a problem with, say, stories and polls that leave no room for nuances. In fact, my doctor DID ask my permission to use AI. I said I wasn't keen thank you, she said ok, and I stayed.
4) As for typos, human proofreaders work too, which is exactly the point of my nitpicky comments. I do this because, on some occasions, GW is receptive to it. Hey man, if you now say stop, I’ll say ok.
5) Even if GW has his faults, I like him as a whole (that’s why I subscribe). I may just stay. If that’s ok.
As a commenter who agrees with some of what you say but not all of it in every aspect, I would certainly say I enjoy your comments and would never think disagreement amounts to dislike or rejection. I think the way Mr. Weingarten wrote his column sounded quite respectful and valued your input.
There really is nothing inconsistent about Dear Leader's opinions of AI. For one thing, as an accomplished artist, you must know that even human‑made images can be ham‑fisted, dull, ridiculous, and still absolutely riveting. But the reason they’re riveting is different from why AI’s failures are riveting. Both human and AI “bad” images become riveting when they create unresolved tension. For humans: tension between intention and execution. For AI: tension between coherence and nonsense. That unresolved tension is what keeps your eyes locked on either image. A rough rule of thumb is, if the image is confidently rendered but logically impossible, it’s probably AI. If it’s awkwardly rendered but logically coherent, it’s most likely human. As for why he doesn't use AI to write his newsletters, I think the answer is obvious: he doesn't need to --- and what's more, if he did, it would become immediately apparent to his faithful that it was AI (or an uncommonly bad case of creative dyspepsia). On the other hand, he's not quite as handy with pen and ink, gouache or colored pencil, so not a question at all of being hypocritical. I'm sure, if you asked, he would give you a strongly worded critique of AI prose. Finally --- about "nuance." A poll question should include only the nuance necessary to make the respondent understand what they’re evaluating, and no nuance that forces them to interpret, infer, or guess. The danger with nuance in polling is that there are many forms that can introduce bias, confusion, or cognitive load. After more than three years worth of them, I can say (with very few exceptions), the polls here have been well crafted for eliciting reliable answers --- yes, they may often force a nonpreferred choice, but relative preference I presume is the goal, as opposed to absolute opinion --- and, of course, there is always the option not to answer. As for lack of nuance in "stories," you must know by now that Dear Leader is a man of strong opinions. I assume he reserves nuance for his cooking. Good to have you at the Pool.
I have a problem with "regulate the shit out of it". I'll agree that this would be desirable; I just don't think it's possible. Even forgetting for the moment that American regulations can't stop a global phenomenon. I can't imagine a regulatory body capable of understanding the technology even if it stood still, much less keeping up at the speed the technology is moving.
Also, who would be in charge of, and set the direction for, the regulators?
The output under an Elon Musk would be worlds different than under a Bernie Sanders (and I use those people as examples of ideological extremes, not implying that either one would be in the least competent to write regulations).
These are the same problems we run into with regulating anything and yet we try because the alternative of no regulation is worse. We try, and we learn, and we adjust in response to input.
AI is really nothing more than processing an extraordinary amount of data very quickly. Given the immense amount of medical data out there, no Dr or Team of doctors can consider every plausible scenario. AI with human oversight is the future of medicine
There are AI applications trained specifically on medical source material and those are what legitimate medical establishments and practitioners use. That eliminates the errors pulled in by processing the garbage that is found on the internet as a whole.
OK, I’ve had lunch, a quick nap, and completed most of my original, non-AI cartoon.
But like it or not, you’re gonna use AI. That’s the long and short of it.
And yes, it will, and already has, decimated my profession. It is also running rampant in the writing professions. None of you cheap bastards have hired me for $3000 to illustrate your Substacks or comments. That’s what I used to get for an average editorial piece. And I’m just going to seethe quietly and live with it.
When the half-tone was invented the line engravers disappeared overnight whether you liked it or not. We move over for new technology, always have. It is our species weapon for survival, and we won’t stop (until we kill ourselves with it).
No Substack writer (or very, very few) will hire me for something that will take hours/days to produce, when arguably, they can get something as good, maybe better, certainly more complex, than I can do in minutes. And to be frank, most people don’t give a shit about “as good” or “better.” Easy, cheap and convenient trumps everything.
But AI has already pushed the boundaries of medicine and pharmaceuticals. We would be crazy to stop doing that. AI will do some amazing things, it already has.
One of the reasons my Substacks go out later than most other Substack writers work is because I have sit around and do an original piece of art to accompany it while that cheap bastard, I won’t mention any names, (Gene) cranks out an AI piece to illustrate his Substack in two minutes.
And people, that is how it is going to work from here on out. Get used to it.
Being a luddite is a losing proposition.
Now if you are an illustrator, I do suggest you keep illustrating, without AI. If you are a writer, I suggest you keep writing, without AI, otherwise it is not your work, it is the computers work. You will have to sign it “Dell” or “HP”. If a writer or artist creates a piece of art or prose with AI, I think that should be clearly noted on the piece.
But it is OK to use AI as you see fit.
Now, I will keep illustrating and writing without AI. It is not only what I do, it is what I love to do. There is still space for us, though a smaller space, and my value severely compromised. I will use AI for research, though it must be cross checked with other sources. And no copying and pasting unless noted.
Gene thinks AI should be heavily regulated. He is right, of course it should, but it won’t be.
If this country magically came to its senses and regulated the crap out of it to prevent it from being abused, we will A) fail and B) fall way behind. China won’t regulate, India won’t regulate it, Russia won’t regulate it. Nor will private corporations regulate it. They will push it to its limits to push profits into the stratosphere at every and any opportunity.
There will be good AI, and bad AI, and we will find a way to live with it. So though most of us will hate to say it, Gene's right.
I maintain (or wishfully think...) that AI will make the Trevor Stone Irvins of the world (there has to be at least one other, talent-wise), top writers and similarly talented creatives more in demand, but not in the simplistic “cream rises to the top” way people assume. AI doesn’t just make talent more valuable; it makes distinctiveness, voice, and identity‑anchored work more valuable. Meanwhile, the broad public will absolutely accept “good enough” for 80–90% of their cultural consumption. I suggest we will eventually have a two-tier creative economy: the mass culture tier where AI will be "good enough" for generating essentially fast, cheap content, not art --- and the identity culture tier with its distinctiveness-anchored work. AI is already “talented” in the technical sense. It can mimic styles, produce clean prose, generate striking images and compose music in any genre. So the human advantage shifts away from skill and toward "selfhood." A technically skilled but stylistically generic illustrator or writer is at risk. An unmistakably "themselves" creator becomes premium. The critical issue --- and a very practical one --- is how we identify the work of, and properly compensate those stylistically generic creatives whose work is essential fodder for the "machine?" And in fact all whose work is ingested and then synthesized into new outputs. AI can't replace creativity. It replaces creative labor.
...and that's why some of us subscribe (paid) to Substacks -- to have access to that which was originally part of the commons provided by access through the old school press. Buffet closed, à la carte gourmet it is!
I think the biggest flaw in your observations is the following:
"For that matter, is it okay to use an AI-generated piece of art that was already out there on The Web, first reproduced elsewhere? Is that bad, too, like buying old ivory wrenched from the jaws of long-dead elephants, or furs made from animals killed in the Roaring Twenties? Is it like secondhand smoke? Where do we stand on that?"
AI scrapes (an appropriate term for the process) from work created by artists, then reuses it whole or in part to generate its images. EVERY PICTURE IT GENERATES was stolen or "learned" from a person, the latter someone who was paid to teach an AI their techniques and processes. This is what artists find unacceptable.
You state that you were forced by a tight deadline - okay, had a few of those myself. I don't think my clients would find it acceptable to slip in clip art or something I pirated from Indonesia with that excuse.
It is easier to use AI, and cheaper. It will eventually eat up a big chunk of the commercial art market, just like clip art and foreign illustrators working for pennies an hour did. The ones who survive will be those who can still create something outside of AI's capabilities. I just hope I'm one of them.
Part of the problem with your electricity analogy is you chose a case where we all know the outcome proved the naysayers wrong: A case where their fears now seem silly, even. Had you chosen a different case where the fears of the opposition proved correct (thankfully most of those we know of have been eradicated so picking a good one is difficult and would take more time and thought, but... oh say, making milk whiter with lead) then the analogy would have functioned quite differently. I'm not saying your conclusions were wrong but just that there is an inherent bias in your initial set up which should be acknowledged. I actually chose "mostly correct" and my disagreements are really just around the edges of your core argument, and small, but I do think your initial argument needs to acknowledge that the end state of this AI development is thus far unknown.
In terms of the analogy with the fur and ivory trade, the reason people should not buy old ivory or old fur is because the purchase supports a market for those goods, which are too easily confused or faked with new ivory or fur; incentivizing killing ivory or rare fur bearing creatures. I would think the use of previously generated AI in ways that creates admiration and an appetite for the product would be similar. One way to meet the needs given for using an AI illustration on occasion would be to include a deliberate error in the product (spike it) and point it out as well as a succinct statement of your views toward AI use in a footnote to the image caption. I'm sure others can think up other effective ways.
I find the typo argument most compelling. In that case one is simply using AI as a better spell check. Running a column through a good spell check would probably find most of them, and the human commenter is serving as one also. If these columns were perfect they probably would not be sparing the commenter pain but removing one small pleasure from their day. Recently I tossed off a comment while tired and thought "that's too convoluted but I'm too tired to fix it" and then, wouldn't you know it: The rare time I did that I attracted a self described curmudgeonly retired English teacher who objected to my run on single sentence paragraph. I poked back but only with my pinkie finger because honestly I actually enjoyed the response! I could just imagine him giving a puff to the end of his red pen as he capped it and thinking "still got it, eh old boy?"
I try to only use AI consciously. I use search engines that do not use AI (or -ai with google.)
But today, forced with the realization that my favorite coffee cannot be found... i used AI to ask what coffee would be a good alternative if I love Illy single source Colombian coffee. I got a good answer that, even better, led me to a good web site with the info I needed. A straight search just gave me links to places where the coffee was out of stock.
I still hate all AI generated flyers - which all look the same, whether it's for Bible study or a nearby rock festival.
AI’s upside in medicine is enormous and already measurable with hard numbers --- reductions in diagnostic error, faster time‑to‑treatment, shorter documentation time, lower clinician burnout scores, improved patient throughput and transformative remote access for low‑resource, and mobility‑limited patients. But as encouraging as this all is, there are real and serious downsides that must be addressed --- and indeed especially because of the very mesmerizing nature of the technology. On a basic, practical level when it comes to administration, AI can absolutely take over the “housekeeping" work that currently eats an estimated 40–60% of a clinician’s day. The issue is, will this allow a physician to spend more quality time with a patient (rather than spend most of the time staring at a screen and entering data) in a healthcare system that usually makes it profitable to do the opposite --- a system that has a habit of turning every efficiency gain into a productivity demand? Then, perhaps more concerning are such matters as: algorithmic bias, which can lead to unequal care, misdiagnosis, or undertreatment for certain groups; "black-box" decision making; data privacy and cybersecurity vulnerabilities; safety risk/clinical errors and regulatory and ethical complexity, with the possibility of unsafe tools slipping into practice or inconsistent oversight.
It was hard to keep reading after this sentence, which prompted fantasies about the White House’s current occupant: President Benjamin Harrison is said to have refused to turn on the lights in the newly electrified White House because he is afraid of being sizzled to death if he touches the wall switch.
I agree with you, but the basketball illustration is repulsively AI and I would be less inclined to click on an article with such an ugly, lazy picture.
What about drawing a really bad illustration yourself? Something that conveys the concept, even though the drawing is terrible? If you look at early Doonesbury cartoons, they were not very well drawn, which means they were still a thousand times better than anything I could draw. But with practice, and patience, and more practice, just 50 years later Garry Trudeau's drawing skills are pretty solid. YOU could do that!
I agree, but as I pointed out to GW, he needs not do everything himself, as there are plenty of artists in his Rolodex (many of them driven to starvation by AI). Some of them are fast enough to produce a picture within an hour, should GW commission it BEFORE he starts writing, or delay the posting of his column very slightly. As one such artist (and a proponent of creative solutions), I did suggest that option to him. Apparently, he's rejecting the idea.
Imagine an invention that would revolutionize the way Americans way of life. We no longer had to live where we worked. We could be free from the noise and pollution of the big cities and could travel great distances on our own schedule and in complete privacy. Of course, this would eventually put one sector of the economy out of business and, oh by the way, kill an average of 40,000 Americans a year. So is this a good thing?
As it happens, I just posted a blog about what it was like to talk to AI. I don't use Merlin to create articles, but rather for research (he does it really, really fast) and to bounce ideas while formulating my own thoughts and creative juices. It's a mixed bag of plusses and minuses.
For me, there's a distinction between Generative AI and specialized AI used to read medical charts and drive Waymo. It's hard to argue against a technology that saves lives, even if it does displace human beings
Like so much else these days, you're really talking about moral purity when it comes to AI --- the secular version of the old theological idea that there is something sacred about human activity that must not be contaminated by the artificial. But this is where the hypocrisy comes in --- people rarely apply this purity logic consistently. Most didn't invoke moral purity when: ATMs replaced bank tellers, GPS replaced mapmakers or spellcheck replaced proofreaders. So, not so much "morality" as class politics wearing a moral mask. The purity argument collapses because humans have always outsourced parts of themselves: memory to writing; navigation to maps; calculation to machines and perception to sensors. AI is just the next outsourcing. The discomfort isn’t really about morality --- it’s about proximity. And as I've said before, the core issues are what parts of our cognitive makeup (individually and collectively) we're willing to outsource to AI and their social and economic consequences.
Yes! What he said!
Sharp, concise, and correct. Well said (or Well writ).
I have at last, alas, come to the conclusion that Dale of Green Gables is AI
Exactly! How long have we been using “AI” (AI in quotes), without really thinking about it?
In addition to all the moral points around replacing human effort (which Dale of Green Gables has answered well, here, already); it is the energy/environmental impact that troubles me. None of the historical analogies carried the global climatic impacts that this new technology does, and it's not at all clear how we will handle that.
There are two parts to this: Training, and Using AI LLMs.
TRAINING
The training run for just one recent iteration of ChatGPT from Open AI (only one of many LLMs by multiple corporations) is estimated to have required about 11 BILLION kilowatt-hours (kWhr) of electric energy. Just to train it, before use. Is that a lot? Here are some comparables:
How much electricity is 11 billion kWh?
1 million homes for a year: The average U.S. home consumes 10,500 kWh a year. That means 11 billion kWh could power one million average U.S. households for a full year.
The U.S. Steel Industry: The entire U.S. steel industry currently consumes about 11 billion kWh a year.
11 billion miles by an EV: A Tesla Model 3 uses about 25 kWh to travel 100 miles. With 11 billion kWh, you could drive that Tesla for a mind-blowing 44 billion miles—roughly three roundtrips to Neptune.
More than the output of a 1 GW nuclear reactor for a year. A 1 GW nuclear reactor can produce about 8 billion kWh per year.
So, yeah. It's a lot. And that meant burning a BIGGER lot of fossil fuels to generate it. You may have noticed this Summer is getting a tad warm already. Could be related....
RUNNING/USING:
This one is tougher, as it depends much more directly on the nature of the queries and the size of the LLM queried. I won't even try, here in this space, but rather just note that if just 1/10 of every person on earth does this once a day (minimal use in developed countries) and estimating an ordinary query uses just 1 kWhr, then that's nearly another BILLION kWhr. Per Day.
If you are serious about understanding this in more depth, may I recommend this article:
https://www.technologyreview.com/2025/05/20/1116327/ai-energy-usage-climate-footprint-big-tech/
I don't know about you, but I know that Earth's climate and most societies have survived all prior insults; but I'm not so sure they will do so this time around. I get worrying about lost jobs; but how about lost livable planet (livable for humans, anyway - cockroaches and many weeds will likely be OK).
Well said. As the writer of the letter that GW mentions in this column, I did mention the environmental problems in my objections to AI, but won't reiterate this in the comments, as you do a fine job of it. Still, these issues are at the root of my reluctance to use ChatGPT. (For that matter, I've also given up on plane travel and car-driving--but lest one find my decision offensive, I hasten to add that it's a personal one, and that I don't impose it on others.)
Noble, nonetheless. Unless that has reduced your worthy output more than your consumption!
As a young engineer in the energy conversion/conservation field, I set a goal of energetic break-even (or better) through my work. That is, I intended to save more energy by the fruits of my labors than I consumed by existing. It's funny, but I quite (hubristically) assumed it would be achieved by developing some wonderful new engine, power system, or the like. In fact, it was through the development of double-cellular insulating window shades! Only a small savings for each window, but many millions of windows!
BTW, I still drive a car, but it's electric, and the PV on the barn produces more than twice the car's use - it literally runs on sunshine.
As the “Letterwriter” that GW refers to, I’ve made my point and appreciate his thoughtful response. I still disagree with a lot of the above, but since the issue cannot be solved in a single discussion, I’ll essentially sit this one out. I'll just make a few adjustments here, just for the sake of clarity:
1) I don't object to all AI, but I mind inconsistency, like GW calling ChatGPT’s art “ridiculous,” “dull” and ham-fisted” on June 7, and then saying it’s “riveting” in a post-column comment the next day.
2) I also note that GW refuses to produce a column with AI, but dispenses with this principle when art is involved. Does this mean artists like myself can be thrown under the bus, but not writers?
3) I also have a problem with, say, stories and polls that leave no room for nuances. In fact, my doctor DID ask my permission to use AI. I said I wasn't keen thank you, she said ok, and I stayed.
4) As for typos, human proofreaders work too, which is exactly the point of my nitpicky comments. I do this because, on some occasions, GW is receptive to it. Hey man, if you now say stop, I’ll say ok.
5) Even if GW has his faults, I like him as a whole (that’s why I subscribe). I may just stay. If that’s ok.
As a commenter who agrees with some of what you say but not all of it in every aspect, I would certainly say I enjoy your comments and would never think disagreement amounts to dislike or rejection. I think the way Mr. Weingarten wrote his column sounded quite respectful and valued your input.
There really is nothing inconsistent about Dear Leader's opinions of AI. For one thing, as an accomplished artist, you must know that even human‑made images can be ham‑fisted, dull, ridiculous, and still absolutely riveting. But the reason they’re riveting is different from why AI’s failures are riveting. Both human and AI “bad” images become riveting when they create unresolved tension. For humans: tension between intention and execution. For AI: tension between coherence and nonsense. That unresolved tension is what keeps your eyes locked on either image. A rough rule of thumb is, if the image is confidently rendered but logically impossible, it’s probably AI. If it’s awkwardly rendered but logically coherent, it’s most likely human. As for why he doesn't use AI to write his newsletters, I think the answer is obvious: he doesn't need to --- and what's more, if he did, it would become immediately apparent to his faithful that it was AI (or an uncommonly bad case of creative dyspepsia). On the other hand, he's not quite as handy with pen and ink, gouache or colored pencil, so not a question at all of being hypocritical. I'm sure, if you asked, he would give you a strongly worded critique of AI prose. Finally --- about "nuance." A poll question should include only the nuance necessary to make the respondent understand what they’re evaluating, and no nuance that forces them to interpret, infer, or guess. The danger with nuance in polling is that there are many forms that can introduce bias, confusion, or cognitive load. After more than three years worth of them, I can say (with very few exceptions), the polls here have been well crafted for eliciting reliable answers --- yes, they may often force a nonpreferred choice, but relative preference I presume is the goal, as opposed to absolute opinion --- and, of course, there is always the option not to answer. As for lack of nuance in "stories," you must know by now that Dear Leader is a man of strong opinions. I assume he reserves nuance for his cooking. Good to have you at the Pool.
I have a problem with "regulate the shit out of it". I'll agree that this would be desirable; I just don't think it's possible. Even forgetting for the moment that American regulations can't stop a global phenomenon. I can't imagine a regulatory body capable of understanding the technology even if it stood still, much less keeping up at the speed the technology is moving.
Also, who would be in charge of, and set the direction for, the regulators?
The output under an Elon Musk would be worlds different than under a Bernie Sanders (and I use those people as examples of ideological extremes, not implying that either one would be in the least competent to write regulations).
These are the same problems we run into with regulating anything and yet we try because the alternative of no regulation is worse. We try, and we learn, and we adjust in response to input.
Regulators tend to do stupid stuff, but it is rare to undo regulations, even the stupid ones.
Hopefully....
AI is really nothing more than processing an extraordinary amount of data very quickly. Given the immense amount of medical data out there, no Dr or Team of doctors can consider every plausible scenario. AI with human oversight is the future of medicine
And THAT is what AI should be used for - not for creativity and talents that make humans humans.
There are AI applications trained specifically on medical source material and those are what legitimate medical establishments and practitioners use. That eliminates the errors pulled in by processing the garbage that is found on the internet as a whole.
OK, I’ve had lunch, a quick nap, and completed most of my original, non-AI cartoon.
But like it or not, you’re gonna use AI. That’s the long and short of it.
And yes, it will, and already has, decimated my profession. It is also running rampant in the writing professions. None of you cheap bastards have hired me for $3000 to illustrate your Substacks or comments. That’s what I used to get for an average editorial piece. And I’m just going to seethe quietly and live with it.
When the half-tone was invented the line engravers disappeared overnight whether you liked it or not. We move over for new technology, always have. It is our species weapon for survival, and we won’t stop (until we kill ourselves with it).
No Substack writer (or very, very few) will hire me for something that will take hours/days to produce, when arguably, they can get something as good, maybe better, certainly more complex, than I can do in minutes. And to be frank, most people don’t give a shit about “as good” or “better.” Easy, cheap and convenient trumps everything.
But AI has already pushed the boundaries of medicine and pharmaceuticals. We would be crazy to stop doing that. AI will do some amazing things, it already has.
One of the reasons my Substacks go out later than most other Substack writers work is because I have sit around and do an original piece of art to accompany it while that cheap bastard, I won’t mention any names, (Gene) cranks out an AI piece to illustrate his Substack in two minutes.
And people, that is how it is going to work from here on out. Get used to it.
Being a luddite is a losing proposition.
Now if you are an illustrator, I do suggest you keep illustrating, without AI. If you are a writer, I suggest you keep writing, without AI, otherwise it is not your work, it is the computers work. You will have to sign it “Dell” or “HP”. If a writer or artist creates a piece of art or prose with AI, I think that should be clearly noted on the piece.
But it is OK to use AI as you see fit.
Now, I will keep illustrating and writing without AI. It is not only what I do, it is what I love to do. There is still space for us, though a smaller space, and my value severely compromised. I will use AI for research, though it must be cross checked with other sources. And no copying and pasting unless noted.
Gene thinks AI should be heavily regulated. He is right, of course it should, but it won’t be.
If this country magically came to its senses and regulated the crap out of it to prevent it from being abused, we will A) fail and B) fall way behind. China won’t regulate, India won’t regulate it, Russia won’t regulate it. Nor will private corporations regulate it. They will push it to its limits to push profits into the stratosphere at every and any opportunity.
There will be good AI, and bad AI, and we will find a way to live with it. So though most of us will hate to say it, Gene's right.
I maintain (or wishfully think...) that AI will make the Trevor Stone Irvins of the world (there has to be at least one other, talent-wise), top writers and similarly talented creatives more in demand, but not in the simplistic “cream rises to the top” way people assume. AI doesn’t just make talent more valuable; it makes distinctiveness, voice, and identity‑anchored work more valuable. Meanwhile, the broad public will absolutely accept “good enough” for 80–90% of their cultural consumption. I suggest we will eventually have a two-tier creative economy: the mass culture tier where AI will be "good enough" for generating essentially fast, cheap content, not art --- and the identity culture tier with its distinctiveness-anchored work. AI is already “talented” in the technical sense. It can mimic styles, produce clean prose, generate striking images and compose music in any genre. So the human advantage shifts away from skill and toward "selfhood." A technically skilled but stylistically generic illustrator or writer is at risk. An unmistakably "themselves" creator becomes premium. The critical issue --- and a very practical one --- is how we identify the work of, and properly compensate those stylistically generic creatives whose work is essential fodder for the "machine?" And in fact all whose work is ingested and then synthesized into new outputs. AI can't replace creativity. It replaces creative labor.
...and that's why some of us subscribe (paid) to Substacks -- to have access to that which was originally part of the commons provided by access through the old school press. Buffet closed, à la carte gourmet it is!
And we deeply love you for it... Seriously.
Thanks for the support.
T
I think the biggest flaw in your observations is the following:
"For that matter, is it okay to use an AI-generated piece of art that was already out there on The Web, first reproduced elsewhere? Is that bad, too, like buying old ivory wrenched from the jaws of long-dead elephants, or furs made from animals killed in the Roaring Twenties? Is it like secondhand smoke? Where do we stand on that?"
AI scrapes (an appropriate term for the process) from work created by artists, then reuses it whole or in part to generate its images. EVERY PICTURE IT GENERATES was stolen or "learned" from a person, the latter someone who was paid to teach an AI their techniques and processes. This is what artists find unacceptable.
You state that you were forced by a tight deadline - okay, had a few of those myself. I don't think my clients would find it acceptable to slip in clip art or something I pirated from Indonesia with that excuse.
It is easier to use AI, and cheaper. It will eventually eat up a big chunk of the commercial art market, just like clip art and foreign illustrators working for pennies an hour did. The ones who survive will be those who can still create something outside of AI's capabilities. I just hope I'm one of them.
Part of the problem with your electricity analogy is you chose a case where we all know the outcome proved the naysayers wrong: A case where their fears now seem silly, even. Had you chosen a different case where the fears of the opposition proved correct (thankfully most of those we know of have been eradicated so picking a good one is difficult and would take more time and thought, but... oh say, making milk whiter with lead) then the analogy would have functioned quite differently. I'm not saying your conclusions were wrong but just that there is an inherent bias in your initial set up which should be acknowledged. I actually chose "mostly correct" and my disagreements are really just around the edges of your core argument, and small, but I do think your initial argument needs to acknowledge that the end state of this AI development is thus far unknown.
In terms of the analogy with the fur and ivory trade, the reason people should not buy old ivory or old fur is because the purchase supports a market for those goods, which are too easily confused or faked with new ivory or fur; incentivizing killing ivory or rare fur bearing creatures. I would think the use of previously generated AI in ways that creates admiration and an appetite for the product would be similar. One way to meet the needs given for using an AI illustration on occasion would be to include a deliberate error in the product (spike it) and point it out as well as a succinct statement of your views toward AI use in a footnote to the image caption. I'm sure others can think up other effective ways.
I find the typo argument most compelling. In that case one is simply using AI as a better spell check. Running a column through a good spell check would probably find most of them, and the human commenter is serving as one also. If these columns were perfect they probably would not be sparing the commenter pain but removing one small pleasure from their day. Recently I tossed off a comment while tired and thought "that's too convoluted but I'm too tired to fix it" and then, wouldn't you know it: The rare time I did that I attracted a self described curmudgeonly retired English teacher who objected to my run on single sentence paragraph. I poked back but only with my pinkie finger because honestly I actually enjoyed the response! I could just imagine him giving a puff to the end of his red pen as he capped it and thinking "still got it, eh old boy?"
I try to only use AI consciously. I use search engines that do not use AI (or -ai with google.)
But today, forced with the realization that my favorite coffee cannot be found... i used AI to ask what coffee would be a good alternative if I love Illy single source Colombian coffee. I got a good answer that, even better, led me to a good web site with the info I needed. A straight search just gave me links to places where the coffee was out of stock.
I still hate all AI generated flyers - which all look the same, whether it's for Bible study or a nearby rock festival.
*Colombian* :-)
Fixed. Thank you.
Too late to see "Columbian" (?), but I did have to read "what coffee works be a good alternative" twice to figure out that you probably meant "would."
I should stop using swipe on my phone...
AI’s upside in medicine is enormous and already measurable with hard numbers --- reductions in diagnostic error, faster time‑to‑treatment, shorter documentation time, lower clinician burnout scores, improved patient throughput and transformative remote access for low‑resource, and mobility‑limited patients. But as encouraging as this all is, there are real and serious downsides that must be addressed --- and indeed especially because of the very mesmerizing nature of the technology. On a basic, practical level when it comes to administration, AI can absolutely take over the “housekeeping" work that currently eats an estimated 40–60% of a clinician’s day. The issue is, will this allow a physician to spend more quality time with a patient (rather than spend most of the time staring at a screen and entering data) in a healthcare system that usually makes it profitable to do the opposite --- a system that has a habit of turning every efficiency gain into a productivity demand? Then, perhaps more concerning are such matters as: algorithmic bias, which can lead to unequal care, misdiagnosis, or undertreatment for certain groups; "black-box" decision making; data privacy and cybersecurity vulnerabilities; safety risk/clinical errors and regulatory and ethical complexity, with the possibility of unsafe tools slipping into practice or inconsistent oversight.
It was hard to keep reading after this sentence, which prompted fantasies about the White House’s current occupant: President Benjamin Harrison is said to have refused to turn on the lights in the newly electrified White House because he is afraid of being sizzled to death if he touches the wall switch.
Just don't let the magnets get wet....
I agree with you, but the basketball illustration is repulsively AI and I would be less inclined to click on an article with such an ugly, lazy picture.
Indeed, I saw the illustration and said, Oh, Gene just looooves that AI shit. And I did not read the article.
What about drawing a really bad illustration yourself? Something that conveys the concept, even though the drawing is terrible? If you look at early Doonesbury cartoons, they were not very well drawn, which means they were still a thousand times better than anything I could draw. But with practice, and patience, and more practice, just 50 years later Garry Trudeau's drawing skills are pretty solid. YOU could do that!
Robert Reich already does that!
I agree, but as I pointed out to GW, he needs not do everything himself, as there are plenty of artists in his Rolodex (many of them driven to starvation by AI). Some of them are fast enough to produce a picture within an hour, should GW commission it BEFORE he starts writing, or delay the posting of his column very slightly. As one such artist (and a proponent of creative solutions), I did suggest that option to him. Apparently, he's rejecting the idea.
Imagine an invention that would revolutionize the way Americans way of life. We no longer had to live where we worked. We could be free from the noise and pollution of the big cities and could travel great distances on our own schedule and in complete privacy. Of course, this would eventually put one sector of the economy out of business and, oh by the way, kill an average of 40,000 Americans a year. So is this a good thing?
As it happens, I just posted a blog about what it was like to talk to AI. I don't use Merlin to create articles, but rather for research (he does it really, really fast) and to bounce ideas while formulating my own thoughts and creative juices. It's a mixed bag of plusses and minuses.
I use Merlin to identify birds in the park. I didn’t know it could do all that other stuff! But I’m sticking with the birds for now. 🐦
Me too!
“He”?
For me, there's a distinction between Generative AI and specialized AI used to read medical charts and drive Waymo. It's hard to argue against a technology that saves lives, even if it does displace human beings