![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png){kind=small}
August Link Roundup
Aug. 9th, 2021 09:53 pmThis month's Link Roundup is likely to be shorter again, as I'm slush reading for Eita! Magazine, an online magazine that publishes fantasy and science fiction works by Brazilian writers in English (and, as of the next issue, will be bilingual). Our upcoming issue will be about food, so if you're Brazilian and you have a story that fits, send it directly to me! Submissions are open until the 16th of August.
AI Weirdness: My Favourite Non-Existent Painter
Since I'm no longer posting articles from 50 Text Games, I should move on to the other newsletter I follow: AI Weirdness, in which Actual Expert Janelle Shane does horrible things to robots, such as asking them to name 90's bands. The content is hilarious, but also a sobering yardstick to how quickly AI-generated text and images are moving forward. Reading through the archives we can witness the world's most advanced AIs forgetting ingredients midway through their recipees and asking people to sautée eternally, to modern neural networks that, while still not perfect, can not only create something that feels human but also have enough memory to be able to go on a tangent, then return to the main point. Shane herself is brilliant at showing both the advances and the limitations of this form, not flinching from its direst consequences while also able to keep it light and tease AI with giraffes.
But the article I brought today is about image generation, which is even weirder, even as it's obviously further from being able to create a work that passes for human-made (or, like, a photograph) unless you are working under very strict parameters, in which case holy shit. However, once you give the AI permission to come across as more abstract, things get interesting - especially as Shane starts to invent artists and finds out the AI gives them recognizeable styles.
Since you're here anyway and this month's roundup is likely to be shorter, have a short story by Shane as well, and don't forget the tool she used to generate the paints in the article is on a publically available Google notebook.
A Tour of the Sacred Library
Still on the subject of AI - using the same tool I linked above, and prompting the AI with the name of painter James Gurney, an entire world was created. The author of the piece fears we're nearing a point in which AI will be better than humans at creating paintings, much as it has on the fields of chess and Go. I think it's still far from that point, and it's much closer to a RPG sourcebook, capable of producing things that we need to expend effort to make sense of, effectively making us the artists. They may be close to producing art that's good enough for corporate needs, though - which just might be a death knell for illustrators to be a viable career.
Meditations on Moloch
A deep and heavy piece on, well, everything, but mostly civilization and humankind. It's a very sobering look at how we create societal machinery that fails to work so often, and how those failures are easily explained by looking at the incentives the machinery gives people. I don't agree with the piece's final advice, that (if I read it right) might be that we need a benevolent, very tightly programmed dictator AI to save us from ourselves; but whatever we might want to build in the future, this might be a good place to look at first.
Felicia Helps Out
And now for something completely different: Felicia, the ferret that helped clean a particle accelerator.
This month's music is The Correspondent's Fear and Delight. A great song and a great video in the same package.
AI Weirdness: My Favourite Non-Existent Painter
Since I'm no longer posting articles from 50 Text Games, I should move on to the other newsletter I follow: AI Weirdness, in which Actual Expert Janelle Shane does horrible things to robots, such as asking them to name 90's bands. The content is hilarious, but also a sobering yardstick to how quickly AI-generated text and images are moving forward. Reading through the archives we can witness the world's most advanced AIs forgetting ingredients midway through their recipees and asking people to sautée eternally, to modern neural networks that, while still not perfect, can not only create something that feels human but also have enough memory to be able to go on a tangent, then return to the main point. Shane herself is brilliant at showing both the advances and the limitations of this form, not flinching from its direst consequences while also able to keep it light and tease AI with giraffes.
But the article I brought today is about image generation, which is even weirder, even as it's obviously further from being able to create a work that passes for human-made (or, like, a photograph) unless you are working under very strict parameters, in which case holy shit. However, once you give the AI permission to come across as more abstract, things get interesting - especially as Shane starts to invent artists and finds out the AI gives them recognizeable styles.
Since you're here anyway and this month's roundup is likely to be shorter, have a short story by Shane as well, and don't forget the tool she used to generate the paints in the article is on a publically available Google notebook.
A Tour of the Sacred Library
Still on the subject of AI - using the same tool I linked above, and prompting the AI with the name of painter James Gurney, an entire world was created. The author of the piece fears we're nearing a point in which AI will be better than humans at creating paintings, much as it has on the fields of chess and Go. I think it's still far from that point, and it's much closer to a RPG sourcebook, capable of producing things that we need to expend effort to make sense of, effectively making us the artists. They may be close to producing art that's good enough for corporate needs, though - which just might be a death knell for illustrators to be a viable career.
Meditations on Moloch
A deep and heavy piece on, well, everything, but mostly civilization and humankind. It's a very sobering look at how we create societal machinery that fails to work so often, and how those failures are easily explained by looking at the incentives the machinery gives people. I don't agree with the piece's final advice, that (if I read it right) might be that we need a benevolent, very tightly programmed dictator AI to save us from ourselves; but whatever we might want to build in the future, this might be a good place to look at first.
Felicia Helps Out
And now for something completely different: Felicia, the ferret that helped clean a particle accelerator.
This month's music is The Correspondent's Fear and Delight. A great song and a great video in the same package.
