Beautiful essay—and please don’t stop writing. One thought that always runs through my head around this issue is the following: LLMs can generate near-infinite mediocre essays and soon enough, novels. But who’s going to read them? One Google is completely flooded, the market is gone. And we are very, very far from LLMs being able to create great works of art, which one can’t arrive at through local statistical variance around the mean.
At the end of the day, if you were told a decade ago the big technological innovation of the 2020s would take the shape of infinite streams of fairly generic content, would you have cheered? Even if it ultimately leads somewhere greater? Is that something literally anyone wanted, aside from engineers and shareholders?
Thanks Haxion :) I agree. Unless those novels are great, who would read them? And if no one does, how can we know? If the ROI of reading an AI novel is unpredictable, why do it? There's more human books written than anyone can read in a thousand lifetimes.
"Infinite streams of fairly generic content," so well put. I don't even think engineers really wanted that. Shareholders just want money. So it's fair to say no one did want it. It's what we have anyway. How are we so bad at getting what we collectively want by falling victim of our individual incentives that so often poison the commons?
And it’s such a shame, because when the goal is not to try to imitate humans, these architectures have real, fascinating promise. Deepmind’s work is often overhyped (and it’s ridiculous that they can frequently publish in Nature while disclosing so little about what was actually done, making reproducibility impossible), but they have shown that these tools have significant value in assisting scientific discovery. And appropriately siloed, LLMs can be useful or at least entertaining. But dumped into the open internet, it’s just a travesty.
To be a "Ingeniero de letras" as my substack is titled is to recognize that same constant split :). Thank you for maintaining consistency, despite everything. I think the world needs more multifaceted people like us.
And, by the way, it's a pleasure to know that we share the love for Silvio Rodríguez's music, which I sing and play in the guitar. Several years ago, I even wrote about a philosophical interpretation of one of his most famous songs, "La maza", and it became one my most popular post:
Alberto — Clearly you’ve been thinking about this a while. It’s a tough precipice to be on, no? For all of us, I mean. But I’m curious, where do you see it going? Do we succumb to its abilities and lay down our pens, do we fight as a “hand crafted” resistance? Or, is the reality of what has/will happen something else completely? I can’t say. I do feel that Goya was right: the sleep of reason produces monsters
I don't think succumbing is ever the best option. But we should at least be aware to see how things unfold, if anything because many people will realize they need to make a career change. The sooner for that, the better. But no one writes just to be paid for that (why not take an easier route?) so o don't think we should ever lay down our pens. It's very hard to predict what may happen. Even if AGI never happens (I don't know why everyone discards that possibility?) there are many ways AI can change the landscape of work and human relationships.
I must admit that when I first saw the article header my assumption was that you were announcing a second Substack. Have you considered dividing your writing into two different subs?
I think Nick Cave's excellent speech summarizes it well for me, that AI is an anti-life equation. As someone who used to avidly love and indeed worked in early ML, I think the last few months have been a swift realization that at the end of the day, only life, beauty and magic is what gives meaning and that the machine is intolerable and in many ways, demonic: the undead recollection of words to create a mockery of life that has never loved, been born or experienced anything.
It took me awhile to get here and a near complete rejection of my old self, but I think, a good realization.
I've written many times early in my time on this blog that I believe that, on balance, AI is a mistake for humanity at this time. I still believe that. However, I came to realize that...
Whether I'm right about my claim or not, whether I like it or not, and no matter what me or anybody else might say about it... AI is coming. It's like the weather, we can't change it by yelling at it.
Facing the inevitable squarely helps simplify the question considerably. Each of us just needs to decide, are we personally going to try to adapt to this new era, or not?
Oh man, do I ever feel this. I’ve never before been so throughly split down the middle. LLMs are an amazing discovery. I almost envy the folks who don’t feel that way – if I could be content to snark about stochastic parrots, the tension would be resolved. Generative AI still thrills a part of me that I can’t turn off – I’m still the same curious kid who could never stop himself from trying to figure out how everything worked. And yet I also hate with a deep and fiery passion nearly everything we’re actually doing to ourselves with the technologies. The horror of watching inevitable automation sink its fingers deeper and deeper into so many things that make us human and give us meaning. We’ll always have art, yes. We’ll always create, sure, we’ll always have hobbies. Art is a way humans communicate with humans. It’s also a craft, though, and a livelihood, and all kinds of other powerful archetypal things, and while LLM prose is still frankly shit, we’ll tinker away at it, help it encroach further and further. I always try not to focus on despairing outlooks. We miss a lot of beautiful things that way. I can’t shake the conviction though that we’re going to a pretty dark place right now.
Alberto writes, "Companies happily enabling this future (of generative AI) are watching impassively how creativity is dying."
Except that creativity isn't dying. The introduction of new tools like generative AI may mean that the process of expressing our creativity online for profit could be changing. The Substack model of "write words to earn" may be evolving. But creativity is not dying. There's nothing about generative AI that ends our ability to be creative.
Substackers who wish to adapt to this new era have the option to edit their relationship with online publishing a bit. Instead of thinking of the writer and their words as being the hero in the spotlight on stage, the only thing that matters, we can broaden our vision. We can think of ourselves as not being just writers, but as creative crafters of a visitor experience. That visitor experience can include our text, and text from other sources such as guest authors and AI. That user experience can also include images, video, music, voice, animations and other content mediums creatively arranged in a near infinite number of combinations.
And of course each of us also has the option to stick with our original idea of "write words for profit". There's nothing wrong with that.
But such a limited vision of the user experience will indeed face growing competition from more creative, colorful and complete user experiences. The attention span and patience of Internet users is steadily shrinking, and it's likely going to become ever harder to engage an audience with just words. As it is already, only a fraction of our audience even opens the emails.
Thanks in part to my conversations with Alberto, who I thank for being so generous with his time, I've been making a journey from talking about AI, to using AI to create my Substack blog. It's been great fun, and I'm feeling quite creative these days. I don't feel oppressed by changes in this marketplace, I feel liberated by them. After making websites the old fashioned way for 30 long years, it's a delight to find a different way to create online.
The true value of AI, if realized, will not so much be generative as literate -- as in, tools which are able to read closely, tease common threads of meaning, connect where social media algorithms (still the equivalent of a glorified chain letter) fail, and above all -- **curate** content produced by human authors.
My sense from trial and over the past month is that we have crossed a threshold in terms of what is possible with the latest generation of models, but that very few people have noticed.
Why? The act of talking tends to get privileged over the act of listening.
A sane person! I kid but only a little. I can't embrace the dualism your proposing as I'm both a pessimist and occupied with splitting being an author and a publisher at the same time.
I hope things get better but I don't expect them to given my experience of human nature.
"I feel pretty much the same, so I took two decisions: Anything I may write will be with my own neurons and synapses. I'll keep working with AI (it's my job, after all), but I'll remain in the technical field, committing myself to not freeze my creativity by letting some algorithm make the fun part of life for me."
The duality of the AI hype and promises and the dystopian reality is difficult to reconcile I am afraid. Even reporting and curating the news on this has shock me in recent times.
I feel very much the same. The potential, promise, fascinating w technology. And disappointment at what it really means and what I see around me.
Many of us contain multitudes.
Some. like my cousin, not so much.
😂
Beautiful essay—and please don’t stop writing. One thought that always runs through my head around this issue is the following: LLMs can generate near-infinite mediocre essays and soon enough, novels. But who’s going to read them? One Google is completely flooded, the market is gone. And we are very, very far from LLMs being able to create great works of art, which one can’t arrive at through local statistical variance around the mean.
At the end of the day, if you were told a decade ago the big technological innovation of the 2020s would take the shape of infinite streams of fairly generic content, would you have cheered? Even if it ultimately leads somewhere greater? Is that something literally anyone wanted, aside from engineers and shareholders?
Thanks Haxion :) I agree. Unless those novels are great, who would read them? And if no one does, how can we know? If the ROI of reading an AI novel is unpredictable, why do it? There's more human books written than anyone can read in a thousand lifetimes.
"Infinite streams of fairly generic content," so well put. I don't even think engineers really wanted that. Shareholders just want money. So it's fair to say no one did want it. It's what we have anyway. How are we so bad at getting what we collectively want by falling victim of our individual incentives that so often poison the commons?
And it’s such a shame, because when the goal is not to try to imitate humans, these architectures have real, fascinating promise. Deepmind’s work is often overhyped (and it’s ridiculous that they can frequently publish in Nature while disclosing so little about what was actually done, making reproducibility impossible), but they have shown that these tools have significant value in assisting scientific discovery. And appropriately siloed, LLMs can be useful or at least entertaining. But dumped into the open internet, it’s just a travesty.
Glad to know you are a fan of Silvio Rodriguez! I enjoyed more his guitar playing than his voice, but I guess it's a matter of taste.
To be a "Ingeniero de letras" as my substack is titled is to recognize that same constant split :). Thank you for maintaining consistency, despite everything. I think the world needs more multifaceted people like us.
And, by the way, it's a pleasure to know that we share the love for Silvio Rodríguez's music, which I sing and play in the guitar. Several years ago, I even wrote about a philosophical interpretation of one of his most famous songs, "La maza", and it became one my most popular post:
https://arjai.es/2016/05/03/la-maza-de-silvio-rodriguez-una-lectura-filosofica/
Totally agree Javier! Awesome! I was listening to La Maza just yesterday haha love it!
Alberto — Clearly you’ve been thinking about this a while. It’s a tough precipice to be on, no? For all of us, I mean. But I’m curious, where do you see it going? Do we succumb to its abilities and lay down our pens, do we fight as a “hand crafted” resistance? Or, is the reality of what has/will happen something else completely? I can’t say. I do feel that Goya was right: the sleep of reason produces monsters
I don't think succumbing is ever the best option. But we should at least be aware to see how things unfold, if anything because many people will realize they need to make a career change. The sooner for that, the better. But no one writes just to be paid for that (why not take an easier route?) so o don't think we should ever lay down our pens. It's very hard to predict what may happen. Even if AGI never happens (I don't know why everyone discards that possibility?) there are many ways AI can change the landscape of work and human relationships.
I must admit that when I first saw the article header my assumption was that you were announcing a second Substack. Have you considered dividing your writing into two different subs?
I have. But both things are me. I like to have it all in one place because it's the heterogeneity that best reflects who I am.
I think Nick Cave's excellent speech summarizes it well for me, that AI is an anti-life equation. As someone who used to avidly love and indeed worked in early ML, I think the last few months have been a swift realization that at the end of the day, only life, beauty and magic is what gives meaning and that the machine is intolerable and in many ways, demonic: the undead recollection of words to create a mockery of life that has never loved, been born or experienced anything.
It took me awhile to get here and a near complete rejection of my old self, but I think, a good realization.
This is the speech, btw.
https://youtu.be/_Pf4GmQY8Ow?si=y9DUCkZ3TOJWkZez
I've written many times early in my time on this blog that I believe that, on balance, AI is a mistake for humanity at this time. I still believe that. However, I came to realize that...
Whether I'm right about my claim or not, whether I like it or not, and no matter what me or anybody else might say about it... AI is coming. It's like the weather, we can't change it by yelling at it.
Facing the inevitable squarely helps simplify the question considerably. Each of us just needs to decide, are we personally going to try to adapt to this new era, or not?
Oh man, do I ever feel this. I’ve never before been so throughly split down the middle. LLMs are an amazing discovery. I almost envy the folks who don’t feel that way – if I could be content to snark about stochastic parrots, the tension would be resolved. Generative AI still thrills a part of me that I can’t turn off – I’m still the same curious kid who could never stop himself from trying to figure out how everything worked. And yet I also hate with a deep and fiery passion nearly everything we’re actually doing to ourselves with the technologies. The horror of watching inevitable automation sink its fingers deeper and deeper into so many things that make us human and give us meaning. We’ll always have art, yes. We’ll always create, sure, we’ll always have hobbies. Art is a way humans communicate with humans. It’s also a craft, though, and a livelihood, and all kinds of other powerful archetypal things, and while LLM prose is still frankly shit, we’ll tinker away at it, help it encroach further and further. I always try not to focus on despairing outlooks. We miss a lot of beautiful things that way. I can’t shake the conviction though that we’re going to a pretty dark place right now.
Alberto writes, "Companies happily enabling this future (of generative AI) are watching impassively how creativity is dying."
Except that creativity isn't dying. The introduction of new tools like generative AI may mean that the process of expressing our creativity online for profit could be changing. The Substack model of "write words to earn" may be evolving. But creativity is not dying. There's nothing about generative AI that ends our ability to be creative.
Substackers who wish to adapt to this new era have the option to edit their relationship with online publishing a bit. Instead of thinking of the writer and their words as being the hero in the spotlight on stage, the only thing that matters, we can broaden our vision. We can think of ourselves as not being just writers, but as creative crafters of a visitor experience. That visitor experience can include our text, and text from other sources such as guest authors and AI. That user experience can also include images, video, music, voice, animations and other content mediums creatively arranged in a near infinite number of combinations.
And of course each of us also has the option to stick with our original idea of "write words for profit". There's nothing wrong with that.
But such a limited vision of the user experience will indeed face growing competition from more creative, colorful and complete user experiences. The attention span and patience of Internet users is steadily shrinking, and it's likely going to become ever harder to engage an audience with just words. As it is already, only a fraction of our audience even opens the emails.
Thanks in part to my conversations with Alberto, who I thank for being so generous with his time, I've been making a journey from talking about AI, to using AI to create my Substack blog. It's been great fun, and I'm feeling quite creative these days. I don't feel oppressed by changes in this marketplace, I feel liberated by them. After making websites the old fashioned way for 30 long years, it's a delight to find a different way to create online.
The true value of AI, if realized, will not so much be generative as literate -- as in, tools which are able to read closely, tease common threads of meaning, connect where social media algorithms (still the equivalent of a glorified chain letter) fail, and above all -- **curate** content produced by human authors.
My sense from trial and over the past month is that we have crossed a threshold in terms of what is possible with the latest generation of models, but that very few people have noticed.
Why? The act of talking tends to get privileged over the act of listening.
It's a very human problem, actually.
A sane person! I kid but only a little. I can't embrace the dualism your proposing as I'm both a pessimist and occupied with splitting being an author and a publisher at the same time.
I hope things get better but I don't expect them to given my experience of human nature.
Lovely column, thank you.
"I feel pretty much the same, so I took two decisions: Anything I may write will be with my own neurons and synapses. I'll keep working with AI (it's my job, after all), but I'll remain in the technical field, committing myself to not freeze my creativity by letting some algorithm make the fun part of life for me."
The duality of the AI hype and promises and the dystopian reality is difficult to reconcile I am afraid. Even reporting and curating the news on this has shock me in recent times.
Both sides of the same cone.
Go full throttle!
Evolution is not a myth.
“Jean-qui-pleure & Jean-qui-rit”