On a technical level, DTGI (diffusion-transformer generated imagery) is a subset of CGI (computer generated imagery). All DTGI is also CGI.
Computer is the medium, like painting on a canvas. You can paint, or you can pay someone to paint for you.
On a conceptual level, DTGI is a big change. From painting, or paying a creative to paint, to paying a faceless corp to generate a permutation of paintings by other people (who probably did not consent to that).
If creator is transparent about it, I would appreciate it, but as someone said once I’d prefer to just see the prompt since that is the extent of creative input.
> If creator is transparent about it, I would appreciate it, but as someone said once I’d prefer to just see the prompt since that is the extent of creative input.
I recently went to a small conference focused on AI art. What you said is simply not true for the best AI "infused" works. Many AI art creators use tools that allow you to combine different inputs, including their own "handmade" art, which are later processed by a pipeline that also include AI and require manual selection of intermediate results. There's a lot of manual work and artistic expression involved, and the average person wouldn't be able to replicate the results through a simple prompt.
While it is true that most people enter a low-effort prompt and call the output "art", better artists are going well beyond that. This is why I think it's just a more powerful version of "standard CGI" (image processing and 3D rendering).
If Veo3 was trained on YouTube, as most sources seem to suggest, wouldn’t it mean that Google just turned its “valued” YouTube creators (at least those who were uploading non-adversarial video, which is most of them in all likelihood) into content producers for a tool designed to replace them? What about YouTube Music—is Google headed towards trillion-dollar lawsuits a la Suno and the like?
We could probably (now) trace a purely domain-agnostic path from "I need to understand subject-matter experts" to "It's just data for machine-learning."
Seeing total video automation is incredible. It's the most compelling reason yet to avoid the infinite treadmill.
Citation needed. Humans have created for as long as they could benefit from their creativity (be it reputation, patronage, or capitalism when combined with IP laws). If that way is no more, I would not be so sure.
That's not the case at all. People all around are constantly making art just for the sake of it. Most of it won't ever even be seen by other people. Removing the financial incentive will definitely change the landscape (for the worse, imho) but creativity for its own sake is enough for most people.
Again, citation needed. For as long as humanity existed, creating art always implied [the promise of] sharing it for either utilitarian reasons, or for enjoying the resulting reputation and recognition (we can go all evolutionary biologist about it and say reproduction, etc.), or for financial gain (more recently). Am I missing some point in human history where this was done without ever expecting any of that?
Writing “for the drawer” is generally a thing that happens if one is not yet satisfied or not confident about the quality of own output, or if one is self-censoring, or if one is expecting someone to read it eventually (communication can happen over time and space). I don’t think this is worth in-depth look, as art that was not seen by others for all intents and purposes does not exist.
(All art is communication; if you shouted into the void and made sure no one hears it, did it really take place?)
Interesting. I think writing a journal is a case where there may be an implicit expectation of somebody reading it later (to give a bad analogy, not unlike how a suicide attempt may often carry a desire to convey a point, even though it is often lethal); if there genuinely isn’t, it may be therapy but probably not art. Art is distinct from a random thing or a pretty picture by being an act of self-expression (no self-expression without the other). Generally, there is also an implied amount of effort and rarity of talent, as well as the position of the work within societal/cultural context (again,
something that involves the other).
If art is communication and self-expression to be consumed by others, that may be all the motivation needed. Despite the disincentives, even.
> new data
We'll likely keep seeing it, up until the inflection point between "machines are for people" and "people are for machines." Along a spectrum, the latter being complete cybernetic mechanization of a life.
Exciting opportunity to stop consuming broad social media and narrowing it down to only specific creators that you can trust not to use AI.
If creators are transparent about it, how is AI worse than CGI?
On a technical level, DTGI (diffusion-transformer generated imagery) is a subset of CGI (computer generated imagery). All DTGI is also CGI.
Computer is the medium, like painting on a canvas. You can paint, or you can pay someone to paint for you.
On a conceptual level, DTGI is a big change. From painting, or paying a creative to paint, to paying a faceless corp to generate a permutation of paintings by other people (who probably did not consent to that).
If creator is transparent about it, I would appreciate it, but as someone said once I’d prefer to just see the prompt since that is the extent of creative input.
> If creator is transparent about it, I would appreciate it, but as someone said once I’d prefer to just see the prompt since that is the extent of creative input.
I recently went to a small conference focused on AI art. What you said is simply not true for the best AI "infused" works. Many AI art creators use tools that allow you to combine different inputs, including their own "handmade" art, which are later processed by a pipeline that also include AI and require manual selection of intermediate results. There's a lot of manual work and artistic expression involved, and the average person wouldn't be able to replicate the results through a simple prompt.
While it is true that most people enter a low-effort prompt and call the output "art", better artists are going well beyond that. This is why I think it's just a more powerful version of "standard CGI" (image processing and 3D rendering).
This is boring and wrong moralizing. It's not convincing at all and won't stop the future from coming.
If Veo3 was trained on YouTube, as most sources seem to suggest, wouldn’t it mean that Google just turned its “valued” YouTube creators (at least those who were uploading non-adversarial video, which is most of them in all likelihood) into content producers for a tool designed to replace them? What about YouTube Music—is Google headed towards trillion-dollar lawsuits a la Suno and the like?
We could probably (now) trace a purely domain-agnostic path from "I need to understand subject-matter experts" to "It's just data for machine-learning."
Seeing total video automation is incredible. It's the most compelling reason yet to avoid the infinite treadmill.
Who will provide new data, once the incentives are eliminated?
History is written by those who bother, as they say. Humans will always create.
At the same time, they can also choose to deliver in a binary medium like a game or proprietary WASM container. Or even scans of their handwritings.
Text is now zero cost; ideas may need to be carried along in frictive streams.
> Humans will always create.
Citation needed. Humans have created for as long as they could benefit from their creativity (be it reputation, patronage, or capitalism when combined with IP laws). If that way is no more, I would not be so sure.
That's not the case at all. People all around are constantly making art just for the sake of it. Most of it won't ever even be seen by other people. Removing the financial incentive will definitely change the landscape (for the worse, imho) but creativity for its own sake is enough for most people.
Again, citation needed. For as long as humanity existed, creating art always implied [the promise of] sharing it for either utilitarian reasons, or for enjoying the resulting reputation and recognition (we can go all evolutionary biologist about it and say reproduction, etc.), or for financial gain (more recently). Am I missing some point in human history where this was done without ever expecting any of that?
Writing “for the drawer” is generally a thing that happens if one is not yet satisfied or not confident about the quality of own output, or if one is self-censoring, or if one is expecting someone to read it eventually (communication can happen over time and space). I don’t think this is worth in-depth look, as art that was not seen by others for all intents and purposes does not exist.
(All art is communication; if you shouted into the void and made sure no one hears it, did it really take place?)
> missing
> void
Art as therapy, communicating with oneself. It's a specific case, though.
Writing in a journal is another example.
Interested in hearing your perspective on these.
Interesting. I think writing a journal is a case where there may be an implicit expectation of somebody reading it later (to give a bad analogy, not unlike how a suicide attempt may often carry a desire to convey a point, even though it is often lethal); if there genuinely isn’t, it may be therapy but probably not art. Art is distinct from a random thing or a pretty picture by being an act of self-expression (no self-expression without the other). Generally, there is also an implied amount of effort and rarity of talent, as well as the position of the work within societal/cultural context (again, something that involves the other).
If art is communication and self-expression to be consumed by others, that may be all the motivation needed. Despite the disincentives, even.
> new data
We'll likely keep seeing it, up until the inflection point between "machines are for people" and "people are for machines." Along a spectrum, the latter being complete cybernetic mechanization of a life.
Yes, it's possible no one will want to make anything if there is no compensation.
At the same time, people do not always calculate world states ahead of time; sometimes, it's just to do the thing.
I’m still waiting for the first compelling piece of computer generated literature or cinema.
I’d say art but for many, generated pictures are amazing so I’ll accept that.
But creating compelling narratives seems impossible right now. I wonder why is that. Are the models missing something?
The videos here are short clips. They can’t be made into a full movie or even a full YouTube video.
I was looking a documentary in Netflix (night stalker) and men all transitions between scene were Ai generated.
This is the first time I cant tell AI slop by the pixels :| dark times ahead.