5 ms·

It's because this technology isn't (and never has been) intended for the laborers making the stuff. The ultimate aim of these companies is to address the manage

It's because this technology isn't (and never has been) intended for the laborers making the stuff. The ultimate aim of these companies is to address the managerial and c suite people with a product that can replace the labor pool. Honestly, I think the current consumer focus is just an irrational consequence of these companies feeling like they can't leave all that consumer money on the table, but, precisely for the reasons this piece outlines, it makes no sense to market this stuff to people who actually like their craft. These tools are purely intended for superficial people who are not interested in creating things, they are only interested in outputs and end results.

TeMPOraL 2y ago
You aren't wrong, especially about motivations. The "current consumer focus" has had some weird results, too - particularly in the image/video generation space, it's the consumers, or rather the OSS community, that actually did 90% of the work of turning AI research output into product prototypes - which many a startup now just pick, polish and scale up. Not to dismiss that work, either, but that's mostly the old-school SaaS building work; the hot AI part is already there on Github, packaged into ComfyUI nodes for people to play with. Full video is a slightly different story because of compute demands, but image generation is basically the infamous Dropbox comment situation - any fancy image generation task you see offered as a service, a moderately tech-savvy person could replicate for personal use with ComfyUI over a weekend. That applies to interactive portraits too - the nodes are but a click away, models run fast enough for local use on an average PC. I'd like to challenge you on your conclusion, though: > These tools are purely intended for superficial people who are not interested in creating things, they are only interested in outputs and end results. But this is exactly how progress is made. Quoting Alfred North Whitehead[0]: "Civilization advances by extending the number of important operations which we can perform without thinking of them." Where for one craftsman, the process is the most important part of their creative work, another craftsman only cares about the output of the former, because it's an input to their creative work. That doesn't make the latter person "superficial", they're just one level up the dependency graph. -- [0] - https://en.wikipedia.org/wiki/Alfred_North_Whitehead https://en.wikipedia.org/wiki/Alfred_North_Whitehead
- skydhash 2y ago
  > Civilization advances by extending the number of important operations which we can perform without thinking of them. That's for mass production of useful tools. For decades, it's been easy to produce music, write a book, and draw with digital means. The actual truth is that people don't want to invest the time to learn how. Now they can pose as if they've done so, but the result never survives close inspection.
  - TeMPOraL 2y ago
    True, if you evaluate the new in context of the old. Anyone can summon a story with an LLM now, but it ain't gonna be as good as one written by an author pouring their heart and sweat into every word. Same with digital images. However, the exact same technology is just a small bit plumbing away from allowing people to summon immersive stories in interactive 3D environments, and iteratively sculpt them from within, in the same way people in Star Trek did with the Holodeck (sans the actual holograms, but VR will do as a substitute). The building blocks are all there today, and are up to spec - what's missing is some glue and ironing out a kinks; I expect multiple attempts at this succeeding within next year or two. And this is a new kind of creative field, which will breed new mode of creating art (not just by chatting with a computer - there's a need for fine manual touch, too). In the end, some things will be diminished, but new things will become possible. I'm mildly positive about it, but I guess we'll soon whether we'll gain more than we lose.
    - skydhash 2y ago
      What you wrote can be summarized as: 1 Write Prompt 2 ... 3 Have a complete work Let's take step 2. Drawing is easy to learn, but it's very hard to master. Why because, it derives from understanding and muscle controls. People fails to draw a simple cube because they don't understand the thing they see. Yes, they probably know that it has 6 faces that are square, but they haven't learned to see it. Knowledge is important, but it's just the initial step. You then have to understand how it all fits together (extract the rules), then apply that understanding. Same with writing. Everyone has a story, but not everyone can tell it because they don't know how to. You have to learn the techniques, then apply it until you have a complete understanding, then you can create as much as you want to. With no understanding, there's no creation. And generative AI won't help, because it can't feel the world as you and I do.
Xmd5a 2y ago
Alternatively current AIs don't compose well with each other and don't let themselves be decomposed easily. This makes control and user interaction difficult to establish which leads to monolithic one-click input->output products by default. Things that can be stated simply ("replace the banjo with a saxophone", "place the cat to the left of the table") can't really be tackled by composition (comfyui workflows of hell) and if you want something that generalizes, you need to train models specially designed with that composition scheme in mind (layout diffusion for instance).

TeMPOraL 2y ago

You aren't wrong, especially about motivations. The "current consumer focus" has had some weird results, too - particularly in the image/video generation space, it's the consumers, or rather the OSS community, that actually did 90% of the work of turning AI research output into product prototypes - which many a startup now just pick, polish and scale up. Not to dismiss that work, either, but that's mostly the old-school SaaS building work; the hot AI part is already there on Github, packaged into ComfyUI nodes for people to play with. Full video is a slightly different story because of compute demands, but image generation is basically the infamous Dropbox comment situation - any fancy image generation task you see offered as a service, a moderately tech-savvy person could replicate for personal use with ComfyUI over a weekend. That applies to interactive portraits too - the nodes are but a click away, models run fast enough for local use on an average PC. I'd like to challenge you on your conclusion, though: > These tools are purely intended for superficial people who are not interested in creating things, they are only interested in outputs and end results. But this is exactly how progress is made. Quoting Alfred North Whitehead[0]: "Civilization advances by extending the number of important operations which we can perform without thinking of them." Where for one craftsman, the process is the most important part of their creative work, another craftsman only cares about the output of the former, because it's an input to their creative work. That doesn't make the latter person "superficial", they're just one level up the dependency graph. -- [0] - https://en.wikipedia.org/wiki/Alfred_North_Whitehead https://en.wikipedia.org/wiki/Alfred_North_Whitehead

skydhash 2y ago
> Civilization advances by extending the number of important operations which we can perform without thinking of them. That's for mass production of useful tools. For decades, it's been easy to produce music, write a book, and draw with digital means. The actual truth is that people don't want to invest the time to learn how. Now they can pose as if they've done so, but the result never survives close inspection.
- TeMPOraL 2y ago
  True, if you evaluate the new in context of the old. Anyone can summon a story with an LLM now, but it ain't gonna be as good as one written by an author pouring their heart and sweat into every word. Same with digital images. However, the exact same technology is just a small bit plumbing away from allowing people to summon immersive stories in interactive 3D environments, and iteratively sculpt them from within, in the same way people in Star Trek did with the Holodeck (sans the actual holograms, but VR will do as a substitute). The building blocks are all there today, and are up to spec - what's missing is some glue and ironing out a kinks; I expect multiple attempts at this succeeding within next year or two. And this is a new kind of creative field, which will breed new mode of creating art (not just by chatting with a computer - there's a need for fine manual touch, too). In the end, some things will be diminished, but new things will become possible. I'm mildly positive about it, but I guess we'll soon whether we'll gain more than we lose.
  - skydhash 2y ago
    What you wrote can be summarized as: 1 Write Prompt 2 ... 3 Have a complete work Let's take step 2. Drawing is easy to learn, but it's very hard to master. Why because, it derives from understanding and muscle controls. People fails to draw a simple cube because they don't understand the thing they see. Yes, they probably know that it has 6 faces that are square, but they haven't learned to see it. Knowledge is important, but it's just the initial step. You then have to understand how it all fits together (extract the rules), then apply that understanding. Same with writing. Everyone has a story, but not everyone can tell it because they don't know how to. You have to learn the techniques, then apply it until you have a complete understanding, then you can create as much as you want to. With no understanding, there's no creation. And generative AI won't help, because it can't feel the world as you and I do.

Xmd5a 2y ago

Alternatively current AIs don't compose well with each other and don't let themselves be decomposed easily. This makes control and user interaction difficult to establish which leads to monolithic one-click input->output products by default. Things that can be stated simply ("replace the banjo with a saxophone", "place the cat to the left of the table") can't really be tackled by composition (comfyui workflows of hell) and if you want something that generalizes, you need to train models specially designed with that composition scheme in mind (layout diffusion for instance).