Posts by aaronhertzmann@fediscience.org
 (DIR) Post #AT4zzAQU5VXgMXBHeK by aaronhertzmann@fediscience.org
       2023-02-26T18:27:04Z
       
       0 likes, 0 repeats
       
       @TedUnderwood This paper begins with unsupported false premises that AI are artists or creative or independent from their users. Replace “artificial intelligence” in the text with the more-accurate term “computer software” to see how silly it is.
       
 (DIR) Post #AT50U9KuBrVJxWLS9g by aaronhertzmann@fediscience.org
       2023-02-26T18:32:40Z
       
       0 likes, 0 repeats
       
       @TedUnderwood  This paper begins with unsupported false premises that AI are artists or creative or independent from their users. Replace “artificial intelligence” in the text with the more-accurate term “computer software” to see how silly it is.The paper doesn't define what they mean by "AI." Do the authors even know anything about how these systems work? (They're in a business school.) Or have they completely swallowed the hype?A much better paper on this: https://www.sciencedirect.com/science/article/pii/S2589004220307070
       
 (DIR) Post #AT52Bhlm8xKP9tFJZY by aaronhertzmann@fediscience.org
       2023-02-23T19:24:49Z
       
       0 likes, 1 repeats
       
       Our new paper on indeterminate art that has been accepted to _Psychology of Aesthetics, Creativity, and The Arts_. We found that a word-association methodology predicts peoples' aesthetic preferences. We performed tests using GAN-generated imagery from Artbreeder. https://psyarxiv.com/nd653But, to get to this result, we employed different methodologies that gave different results, providing more reasons to question the uncritical reliance on MTurk evaluations in computer graphics and vision.
       
 (DIR) Post #AYGCz32rqE6GsF0Jto by aaronhertzmann@fediscience.org
       2023-07-31T16:17:10Z
       
       1 likes, 0 repeats
       
       This year at SIGGRAPH, we are presenting two papers that, I believe, represent breakthroughs on the problem of computing occluding contours of smooth surfaces. One of the foundational steps in artistic and non-photorealistic rendering of 3D objects are computing occluding contours, but doing it accurately for smooth surfaces has always been a bafflingly-hard problem. 1/
       
 (DIR) Post #AYGCz4rF6PzQUla43c by aaronhertzmann@fediscience.org
       2023-07-31T16:17:52Z
       
       0 likes, 0 repeats
       
       Here's a new series of blog posts that explains what the problem is and how hard it's been, summarizes our new insights, and describes which methods are now best for which problems—and where there's new opportunities for new research. With post-Spider-Verse Hollywood embracing non-photorealistic animation, I believe these algorithms could finally see wide-spread use.The first post describes the problem and how shockingly hard it's been: https://aaronhertzmann.com/2023/07/31/occluding-contours-part-1.html2/
       
 (DIR) Post #AYGCz6dUUWB60hA6ts by aaronhertzmann@fediscience.org
       2023-07-31T16:19:34Z
       
       0 likes, 0 repeats
       
       Second, a summary of our new insights and approaches:https://aaronhertzmann.com/2023/07/31/occluding-contours-part-2.htmlThird, and finally, which method I think appropriate for different use cases, as well as open research questions. https://aaronhertzmann.com/2023/07/31/occluding-contours-part-3.html3/3
       
 (DIR) Post #AZfr4zSSEswUSyHofI by aaronhertzmann@fediscience.org
       2023-09-10T20:24:34Z
       
       0 likes, 1 repeats
       
       https://www.reddit.com/r/StableDiffusion/comments/16ew9fz/spiral_town_different_approach_to_qr_monster/