The whole AI workflow sounds incredibly backwards - you give all these creative tasks to computers, which are notoriously bad at creative tasks, and then the drudgery of verification and review goes to humans, which are notoriously bad at verification and review.
Robots replaced most painters a century ago, its called cameras. So what you are talking about has already happened many times over and started a long time ago, its just easier to replace art since it doesn't have a precise requirement.
Or see social media replacing news etc, similar situation. It turns out that manipulating our emotions at scale isn't that hard.
Edit: Look at old things that were made, craftsmen put in extra work to make them beautiful. Modern things machine took all that work and produce standard things, machines took most of the art making that lots of people used to engage with already. Instead today humans mostly do paperwork.
Edit2: Point was that technology always made us move from art to ever more tedious work, only exception was probably farming.
You can absolutely automate ornamentation with cnc. Companies choose not to because there are very few kinds of ornamentation that stand the test of time. The ornate things you see on Antiques Roadshow were, in their time, contemporary with tons and tons of crap that didn’t survive. What you’re seeing is that automation is used to produce crap to exacting standards, which frees craftsmen up to produce ornate pieces for very high cost. People also don’t want to pay for ornament on disposable pieces, which most flat pack furniture is disposable. So it’s far more complex than you make it out to be.
I see a ton of ornamentation in poor old villages and tribes, craftsmen making ornamental pieces has always been a standard part of human culture. Today that part of human culture is mostly lost and only a few hobbyist engage with it or buy it, most are happy with robot made ornaments.
Sure they exist, but they aren't nearly as established or high status as a century ago. Technology did replace artists, some still do art but they are very few and mostly just do it on their personal time instead of being professional artists. You can still draw your own art even with image generators, they didn't change that.
Before automatic photos, automatic music playing etc, people who could play music or draw were highly sought after just to draw and play music everywhere, since there were no alternatives. Technology changed that and today those are dying breeds.
The reason music performances aren't as high status as centuries ago is that it's been democratized. The barrier to entry is lower. Anyone can get a mic and amp and start busking on the street corner. I think that's a good thing for humanity at large.
Practice of art works better IMO as an avenue for creative expression than it does as a status symbol. This can be done even with no audience at all. Perhaps the audience is all consuming AI-generated work.
Status often comes with rarity. There's more art now, and that's a good thing. Some of it has status, but most is just people doing stuff because they can. And that's not a bad thing.
> The reason music performances aren't as high status as centuries ago is that it's been democratized
I'm talking about the median musicians status, not the higher end, that crashed when automatic music playing started to get good. Before then every little village needed to have musicians around for their celebrations and parties, knowing how to play music was respected as a proper profession then unlike now.
That goes for all the ornamental pieces that used to be done by craftsmen as well in villages, things like making nice signs or making celebratory furniture or festival clothing etc, all that artistry that common people used to do is now mass produced by machines that just copies a single design over and over.
So overall I'd say machines so far massively reduced human artistry and cultural creation. The artisans of old didn't just copy either, humans aren't machines that can perfectly copy so they all put in a bit of themselves in all of their work, that wasn't soulless, and it is now mostly lost.
> Before then every little village needed to have musicians around for their celebrations and parties, knowing how to play music was respected as a proper profession then unlike now.
This isn't really right, I suspect. What is more likely is that before then, lots of people knew how to play some musical instruments, and most were rudimentary. Because music was foundational to cohesion.
"Musician" as a profession is really quite new -- as a sibling comment suggests, patronage/donations/tips/busking would have been the only way.
But then also it's fair to say that an entire class of reliable musical instruments produced at any kind of production scale is also quite new.
Essentially all valved brass and wind instruments are less than about 200 years old in design. The first modern classical guitar is also not much more than 200 years old, surprisingly. The first pianoforte is only 320 years old or so.
Many simple folk instruments are this sort of age -- the balaika is at most as old as the piano, the ukulele much younger.
Few truly loud melodic instruments existed much beyond 1550; Amati's violin dates to then. Amazingly the rackett, an instrument often used to portray medieval wind music in films, is younger than the violin, and the crumhorn is not much older.
In the bad old days, the only practical way you could really learn music was to get a patron, or somehow make a living out of it. The barrier to entry was sufficiently high that it was a full-time commitment, one way or another.
Today, it's still possible to be a full-time musician, but by far, most musicians have day jobs. You can do music as a hobby.
On one hand, high fidelity music reproduction lowered the demand for performers, as you note. On the other hand, cheap high quality music equipment lowered the barrier to entry. Today you can play your piano piece on a sub-$1k electric keyboard that's portable and never needs to be tuned. It even sounds good. My opinion is that, on balance, the net human artistic output is way up. I also have no data to support this. But it just feels right.
I also think net human artistic and creative output is up and we just don't realise it -- therefore we do not think clearly of the damage generative AI will do to how we feel about the value of our lives.
My understanding was actually that technology enabled more time for leisure activities, one of those being art. Maybe not in the past few decades, but certainly as we through the Renaissance period and onwards.
Cameras replaced only a very specific kind of painting, one that some people still do as a hobby (realism).
To your second edit, the fantasy of automated robots has always been to move us in the direction of a leisure-based society. Even today, Elon musk is promoting robots and AI as the means to get "universal high income".
The fact that this often does not pan out (though arguably laundry machines, factory clothing and vacuum cleaners have done more than almost anything else in that regard) does not mean that the idea is in any way uncommon.
> Cameras replaced only a very specific kind of painting, one that some people still do as a hobby (realism).
That was by far the most common way for artists to provide for themselves though, cameras did take their jobs.
Similarly, text to image only solves a very specific part of art, art where you aren't very creative but instead draw an image based on a description. Artistry based on coming up with new things to draw will still be there. Sure most jobs artists do today are based on drawing based on descriptions, so its the same situation as with the cameras.
Those are the equivalent to prompt engineer artists, they use the new tool to create new art. Still removed most of the jobs, there isn't nearly as much need for photographers as there were need for painters.
When you consider that the people who develop this stuff consider that a good way to write a song might be to get an AI to generate a complete generic, formulaic song and then have a musician fix it, it's not so hard to understand why they think it might work for the law.
Because that isn't how you write good songs any more than it's likely how you write good legal filings.
Depends on your definition of good. All popular songs have been highly formulaic for a long time. That's what makes them popular, they aren't challenging to listen to and sound roughly like everything else. I'm not going to get into the nitty gritty of music theory, but any serious musician can tell you that writing a good song is striking a balance between boring and novelty. I think AI generated music does in fact get you 95% of the way there. The other 5% is already being in the elite levels of the music industry.
That's even more confused. They never said formulaic songs become popular, they said there is a formula to make popular songs. Those are different statements, and the latter is clearly not true.
A "hook" is a pretty pure example of musicianship, because it has to be "hooky".
Gregory Bateson said that "information is the difference that makes a difference".
For a musician as for music fans, hooks are like this: a thing will stop being hooky when everyone uses it.
It just becomes part of music language. Indeed you could argue that many of the fundamental qualities of popular music are hooky qualities pushed down a layer or two.
Carly Rae Jepsen's "Call Me Maybe" is an example of this. It's full of very obvious, well-understood musical tricks -- syncopation, breakdowns, subtle pacing changes.
But it's also full of maddeningly effective hooks that are incredibly, even deviously well-crafted. People write songs like this to feel proud of writing songs that get into everyone's heads, but also because it gets into their heads.
Since generative tools don't know how to focus on craft -- or even of its existence -- a good hook is something a generative algorithm will always struggle with, because it requires innovation, and because it is a complex, fragile element in itself.
The rest of it, as you say... it's likely quicker for an experienced musician to just do the work rather than keep poking a generative tool until it is right.
I bet you all of this holds true -- innovation, craft -- even to some extent in quite banal legal filings.
> but there is already a significant amount of automation for the mechanical parts.
Right -- and there can be ordinary, boring, individual scrutiny of what those pieces do, and data/code fixes for them.
I mean, it's no small amount of irony that the systems lawyers use are close to the kinds of "expert systems" that dominated AI development after the first AI winter.
It's true. I've integrated ChatGPT into my report writing work flow, and about all it's good for is brainstorming, and quickly formatting ideas. You've got to do the hard work yourself.
TL;DR: There's a lot of really crappy, rushed, "AI" that accumulated the last 18 months. Yet AI is awesome. It's *really* stunning to live it. I cannot stress this enough. You could probably do an GPT wrapper startup today, and as long as you don't rush, roundly beat every single incumbent within a year.
Context:
I'm a sole developer that quit my job at Google 7 months ago, decided to 18 months ago after getting first-hand look at how search x AI was being built.
I just wrapped up 2 days of final benchmarking for release.
3-search-query RAG scores 97% on the USMLE, 6 points above "SOTA": Gemini 1.5 tuned for med that does 4 rounds of N answers each round, then 3 search queries to resolve differences.
How am I possibly roundly beating that? Rushing. Lack of attention to basic details while just getting to the desired outcome, "we finetuned for med and beat GPT-4". And that's *Google*, infinitely resourced.
And Perplexity? AI startup darling? 76%. That's the $40/month version. The free version is at 66%. I'm absolutely stunned, it didn't seem great, I didn't know it was actively much worse than just using ChatGPT.
If that's how Google and Perplexity are going, I can't even imagine the shenanigans that are going down at companies like these. They don't have infinite resources, expertise, or have it as a core competency. There's probably more effort put into who gets to work on the AI thing than working on the AI thing.
(I did legal & med benchmarking, for legal to compare TFA, perplexity free 58%, perplexity pro 67%, llama 8B w/no internet 65%, gpt4o x internet 90%)
