Canta Y No Llores
This is my stand.
We must now face the inevitable and talk about AI.
I think it's obvious to everyone that I don't use AI for writing these blogposts, since 1) the ideas are original and specific, 2) the style is idiosyncratic, unpolished, and usually brief.
I do use AI for the little online widgets I make, and I think some of you would be shocked at how much of the work on those was done by the AI directly, and how fun it is to say "can you make me a tool that does XYZ?" and have it spit out something workable from the start. I also use AI for images, e.g. here.
I know a lot of you have strong feelings against AI use, and I respect that. Personally I have complicated feelings about the impact these tools will have on creative work. Here's my thoughts:
I really do believe that some crafted objects contain something you might call The Divine Spark. There are certain ceramic cups that for reasons quite beyond me feel immediately right and true when you hold them. And I think that even amateur ceramicists sometimes make pieces that have this quality (despite being extremely unprofessional in other ways!), and that some professional ceramicists seem able to produce them consistently, and that I've never yet seen a machine make a piece that has this quality. And I actually do believe that this points at something meaningful about this universe.
On the other hand, a lot of the art we collectively consume every day has long been spark-less, and I don't yet believe that going from human-made-slop to AI-slop is meaningfully worse. The confusing part to me is why we surround ourselves with slop to begin with; I have never been able to explain, to myself or anyone else, why I spend 90% of my art-consuming-time on trash IN A WORLD WHERE THERE IS ALREADY MORE INCREDIBLE ART THAN I COULD EXPERIENCE IN A LIFETIME, if any of you know what's up with this I am begging you to explain me. But in any context where people were already consuming things that lack any connection to a soul, I don't personally see the difference between having it be AI or human soullessness.
For me, the operative question right now is whether AI overall helps make human art better or worse. And I think the answer is.... both? I think that any kind of supportive technology can either enfeeble or enrich, i.e. it can help you get better at doing something naturally by yourself, or can become something you rely on so much that you therefore never get better yourself.
I don't have a great theory for when and why something becomes enriching rather than enfeebling. This has all been said a million times before but the Printing Press has been (generally) good for writers, since it saves us from writing everything out by hand, and therefore let millions of people have meaningful creative experiences they couldn't have had otherwise.
The press had negative consequences as well, in that it reduced the amount that people memorize and story-tell, right? Socrates was right that writing decays the memory and circumvents certain kinds of dialogical relationships, no? Maybe not, I don't know – perhaps there are more standup comics today than there were bards in the ancient world, perhaps everything just becomes more with time. I don't really know that much about the past.
I'm also struck that during every moment of change, the relevant privilege group – that is, the people who had the fortune to already be good at the thing the technology helps with – will, by default, be aghast about the change; moreover, that even after hearing this argument (in general) they will believe that in this case their rights are earned.
So the medieval monarchies truly did believe in the Divine Right of Kings, and/but those monarchs were the extreme beneficiaries of it. And when I mentioned this in a college seminar everyone nodded happily, but then when I said that the same applied to us (the seminar participants) as the winners of the current system, everyone got mad at me. Because in our case it was different, we worked hard to be there and it was entirely meritocratic and blah blah blah blah blah.
And both things could be true at once, there could be good reasons to believe in the merits of the current system, and as the winners of that system you would be super highly incentivized to justify it regardless of its complicated pros and cons.
So I think there's something a bit treacherous about e.g. highly literate people who have spent their life benefiting from being good at spitting out words, assessing the merits of a new technology that is extremely good at spitting out words.
I have certain skills and resources – e.g. the ability to improvise a fairly passable poem – that have now become commoditized and lost 99% of their value overnight. I think this is "the problem of expertise" more broadly: the "experts" are the people most likely to understand the value of something, and also the people who have the most entrenched interest in it, and it's hard to separate the two. I really did work hard on my writing, and I really do think it's bad that the cost of slop-production has fallen to zero, and that thoughtful writers will get drowned out in the process.
And/but the total impact of AI is some combination of "the impact on people who were already good at what AI does" and "the impact on people who were previously not-good at what AI does," and there's a very real chance that the overall outcome is "it's bad for me specifically but good for humans overall," but it'll be hard for the me's of this world to admit that. (It might not be true!, but if it's true it'll be hard to admit it).
The other thing about AI is that it's such a good complement for so many things, and in theory a good complement enables better work from its complementers. I'll write about this more someday but imagine a farmer before the development of mechanization: their output is largely a function of their physical strength, so if you're a 10x strong farmer you can grow 10x more crops than your neighbour.
Then someone invents the tractor and now 1) there's very little advantage to being physically stronger, but 2) there's tons of advantage to being smart and strategic and planning your fields correctly, or whatever. Previously if you were very smart but happened to be weak you couldn't really reap the advantages of your farming-smarts, but now with tractors you finally can, and (what's more) you can now come up with even smarter farming strategies that didn't make sense previously.
And so much of art is like this: previously to be a singer-songwriter you had to be good at writing lyrics and good at writing music and good at singing and good at playing instruments, and now if you're good at any one of these you have an amazing collaborator who can do the others. And honestly I think that's beautiful, and important, and so long as people actually do keep aspiring to be good at at-least-one of these, it's possible that AI will lead to better human art ability. And I know what some of you will say – "you could have just collaborated with another human who had those skills! That is what we're here for" – but as someone who spent a decade trying and very occasionally managing to collaborate with other musicians, the collaboration costs are really really high, and it's obvious to me that 99% of the art that could have existed in the last few decades just never came into being, and if we get that number down to even 90% I think that will be a triumph for the human soul.
For some people, I do understand, the objections are deontological: in that case, obviously the consequences don't matter, it is better to have no art than to use something you believe is wrong. But for those without deontological objections it's surely a question of "what would have happened instead", and so often the alternative to AI art is no art, not human art. It's important to me to figure out how to make sure that the short-term use of AI doesn't lead to long-term atrophy of the human willingness or ability to create, but that's a question of strategy. If we're going down, I plan to go down singing.