I agree. I’ve already started to develop a bit of an instinctive ‘ugh’ reaction to random illustrations, even ones without obvious generative model tell-tales or the DALL-E watermark.
It gets worse if you can spot any flaws or they do look bad overall: “so you couldn’t even take another minute to generate a variant without blatant artifacting? That’s how slovenly and careless your work is, and how little you respect your readers?”
if this monkey pox image can be generated by a prompt of 12 redundant words, then that’s another of saying that the image is worth far less than a thousand words—it’s worth less than 12...
It’s only worth less than 12 if you have mental access to state of the model after training. If not, it also includes a bunch of what it learned.
Thanks, I appreciate your perspective. I think I agree directionally but I’m not as negative (also I don’t know what Memphis design is, I’ll look it up).
Overall, I think my original comment’s tone was too negative. It looks like the parent comment retracted, probably because everyone piled on =( .
I think my intent behind my original comment was about the aesthetics and purposes/agenda of Asterisk. I’m pretty sure they want to signal depth/legitimacy, and avoid trendiness.
What is your view about Asterisk’s design, separate from the Dall-E issue?
But I have a lot of questions about the ‘highlight’ feature: aside from the many teething problems Said has already documented, which bugs doubtless will be fixed, I don’t understand what the usecase is, compared to other web annotation systems like Hypothesis—so it stores arbitrary ranges of text, saving them to, I assume, dangerously ephemeral browser LocalStorage where they will be unpredictably erased in a few hours / days / weeks, and unavailable on any other device presumably. Why do I want this? They aren’t clipped to something like Evernote, they aren’t synced to my phone, they aren’t saved anywhere safe, they aren’t posted to social media or visible to anyone else in any way… Is the idea to highlight various parts and then manually copy-paste them?
Maybe? I can’t easily appreciate such a usecase because I always want to save any excerpts I find worth excerpting. Are there a lot of people who want that? If that’s the idea, I guess the “About Highlights” dialogue needs a bit of documentation to explain the intended (and unintended) uses. At least, anyone who doesn’t realize that the annotations are ephemeral, because they aren’t enough of a web dev to understand ‘What you save is stored only on your specific browser locally’ is as much of a bug as it is a feature, is in for a bad time when their annotations inevitably get deleted...
(This is totally piling on here, and AI assisted art and content has a big future)
It’s obvious to everyone that a lot of “2022-style” Dall-E content is gimmicky, not a substitute for illustrators, and going to age poorly.
I agree. I’ve already started to develop a bit of an instinctive ‘ugh’ reaction to random illustrations, even ones without obvious generative model tell-tales or the DALL-E watermark.
It’s comparable to how you feel when you notice that little ‘© Getty Images’ or a Memphis style image, and realize your time & (mental) bandwidth was wasted by the image equivalent of an emoji. It’s not that they look bad, necessarily, but they increasingly signify ‘cheap’ and ‘tacky’. (After all, if this monkey pox image can be generated by a prompt of 12 redundant words, then that’s another of saying that the image is worth far less than a thousand words—it’s worth less than 12...)
It gets worse if you can spot any flaws or they do look bad overall: “so you couldn’t even take another minute to generate a variant without blatant artifacting? That’s how slovenly and careless your work is, and how little you respect your readers?”
It’s only worth less than 12 if you have mental access to state of the model after training. If not, it also includes a bunch of what it learned.
Thanks, I appreciate your perspective. I think I agree directionally but I’m not as negative (also I don’t know what Memphis design is, I’ll look it up).
Overall, I think my original comment’s tone was too negative. It looks like the parent comment retracted, probably because everyone piled on =( .
I think my intent behind my original comment was about the aesthetics and purposes/agenda of Asterisk. I’m pretty sure they want to signal depth/legitimacy, and avoid trendiness.
What is your view about Asterisk’s design, separate from the Dall-E issue?
I like it overall.
But I have a lot of questions about the ‘highlight’ feature: aside from the many teething problems Said has already documented, which bugs doubtless will be fixed, I don’t understand what the usecase is, compared to other web annotation systems like Hypothesis—so it stores arbitrary ranges of text, saving them to, I assume, dangerously ephemeral browser LocalStorage where they will be unpredictably erased in a few hours / days / weeks, and unavailable on any other device presumably. Why do I want this? They aren’t clipped to something like Evernote, they aren’t synced to my phone, they aren’t saved anywhere safe, they aren’t posted to social media or visible to anyone else in any way… Is the idea to highlight various parts and then manually copy-paste them?
I thought that the point was to help with active reading and little more.
Maybe? I can’t easily appreciate such a usecase because I always want to save any excerpts I find worth excerpting. Are there a lot of people who want that? If that’s the idea, I guess the “About Highlights” dialogue needs a bit of documentation to explain the intended (and unintended) uses. At least, anyone who doesn’t realize that the annotations are ephemeral, because they aren’t enough of a web dev to understand ‘What you save is stored only on your specific browser locally’ is as much of a bug as it is a feature, is in for a bad time when their annotations inevitably get deleted...