Generated Memories

AI-generated sound animation
1024 x 1024px, 25fps
18'' Loop

"Generated Memories" is the retrospective unfolding by Xiyue Hu upon confronting the belongings of her late relative. These objects mirror the idiosyncratic memories of the era he was living. By taking the collection as a dataset, Xiyue works with the GAN (Generative Adversarial Network) algorithm to generate continuously morphing animations out of the speculation over the information learned from inside the dataset. GAN adjusts the connection of the nerves through positive and negative feedback, simultaneously managing the intensity (weighting) of connectivity. Xiyue approaches positive feedback as memory archiving, while negative feedback as forgetting. Forgetting may not be the antithesis of memorizing. It is the function, complementary to memory, as death is as much a function of life. Through memorizing and forgetting, life anticipates the future, and through death, lives with the present.

Xiyue maps GAN-generated images into digital objects which are placed in a floating cosmic space, reconstructing a vacuous memory site and externalizing memories to process oblivion.

This project is issued in NFT.And exhibited at the virtual group show. On view at:

Editor: Hao Long


1024 x 1024px,25fps

《记忆生成器》是曦月基于对以故亲人的收藏物所展开的追溯。这些物件映射了他所生活的那个年代的特殊记忆。她将这些收藏品作为数据集,使用GAN( 生成对抗网络) 算法,对从数据集里面学得的信息进行猜测,生成一个连续不断变幻的图像。GAN 通过正负反馈来调整神经的连接,同时控制着神经连接的强弱(权重)。曦月将正反视作记忆存档,而将负反馈视为遗忘。遗忘或许不是记忆的反面,而是一种功 能,是对记忆的补充,正如死亡也是生命的一种功能。生命通过记忆和遗忘来预测未来,通过死亡,与当下相处。

曦月将GAN 生成的图像映射成虚拟的数码物件,置入一 个漂浮的宇宙空间中,重构一个真空的记忆场所,将记忆外 置化,以处理遗忘。

该项目以NFT 铸币发行。并通过线上群展展出,查看: