The Future of Memory

The Future of Memory is an online art work that looks at artificial intelligence algorithms and automated censorship in China. It brings a poetic intervention into ways users might bypass such automated systems through image, invented language and visual form. The piece explores the boundaries of translation, the ways language changes under algorithmic logic and the limits of machine intelligence.

One content screener at the tech company Bytedance Beijing has notoriously said, “…artificial intelligence is a scalpel, and a human is a machete.” Algorithmic censorship is a blessing and a curse, for how swiftly and surgically models can learn to censor items, yet still hold constraints and inability to anticipate the creativity and playfulness of China’s netizens (online citizens).

For example, when #MeToo became censored in China, users used the combination of a rice bowl and rabbit emojis, because in Chinese, the words are “mi” for rice and “tu” for rabbit. When read aloud the combinations sound like “me too”. The use of homophones to evade censorship in China is widespread, and especially effective given the sheer number of characters in Chinese. The way characters are structured allow entirely new characters with homophonetic sounds to be endlessly formed, into a new language that “looks” Chinese and is legible to someone who reads Chinese, but such characters are entirely made up. Netizens are also evading censorship by posting to places that have enormous economic value, such as Github (for a widespread protest by engineers), where censoring the platform would be economically unfeasible.

The Future of Memory deep dives and reverse engineers how such censorship algorithms work and puts this information onto an interactive, bilingual website. The project creates a new visual-linguistic systems to bypass censorship.

The website offers a translation tool and character generation tool as interventions: of playful censorship evasion through emojis, generating visually compelling, image based memes, suggesting homophone characters (characters that sound the same but are written differently) and creating new homophone characters.

The intervention includes an interactive portion to create new characters, a unique new online language out of the existing buildings blocks of Chinese characters. This is based off existing netizen practices of using homophones and will be accessible to a diverse, broad range of people across the globe.

An online exhibition will be held, where viewers can see these invented, new characters from the censorship resistant language, generated by the tool. Such an exhibition will show the visual and linguistic dimensions of these new characters, as well as the ways language is both a form of visual representation as well as embodiment of meaning. With Chinese characters being ideograms, and illustrative/visual representation, existing characters are similar to what they are trying to represent: the character for sun looks like a sun. These invented characters will also look visually similar to what they represent, with meaning updated for contemporary times.

The Future of Memory plays on language, the constraints of language, and the visual weight of Chinese characters to create a new language: new Chinese characters that are constantly shifting and evolving to bypass the cold logic of algorithmic censorship. I am interested in exploring the line between language and representation as well as the ways new visual-linguistic representation end up shifted by the logic of machines (in this case, censorship algorithms). Other themes that I am looking at in this piece include the provocative role of legality: there is technically no explicit rules on censorship in China, but simply a broad set of vague policies and some scrubbed keywords. How can visual-linguistic systems evolve in this gray area?

While algorithmic censorship is removing emerging media on a day to day basis, existing media and memories from vaccines scandals, to political events, to historical artworks, to the lawyer Zhang Qianfan’s constitutional law book are now being disappeared from the internet entirely, in a shocking amnesia. A distributed, physical network of offline modules will store and archive such disappearing works, news items, photos and memories in this new kind of censorship proofed language of novel, new homophones and emojis. These archived memories will also be available online.

Towards of the end of the residency, I will hold an event in the San Francisco Bay Area to teach users how to make these modules and showcasing the language tool, as well as discussing ways algorithms work within social network such as Weibo and WeChat. Given the intersection of the Bay Area as a site of Chinese diaspora and concentration of VC funding that fuels platforms that use automated censorship, it will be a rich conversation.