DOUBLE BLOW
For artists, this is a devastating double blow. Many feel that their work is being taken to create the very tools that then threaten to their livelihoods. A survey conducted last year by the Arts Workers Japan association found that 92 per cent of illustrators feared their works were already being scraped to train AI tools without their permission. Roughly 60 per cent of respondents were also concerned about fewer employment opportunities.
Already, some of the most beloved Japanese characters seem to have been swept into AI training data. Social media has become flooded with artificial versions of them, including machine-gun wielding Hello Kitties or body-builder Pikachus. And a slew of open-source AI models have made it so that almost anyone can train tools on images they upload from their favorite artists to spit out content resembling that style.
And it’s not just still art. Earlier this year, OpenAI teased a first look at its video tool, Sora, rocking the industry. Filmmaker Tyler Perry said he halted a planned US$800 million expansion of his studio in Atlanta after seeing Sora’s “mind-blowing” capabilities. OpenAI, meanwhile, hasn’t publicly shared details of what data it was trained on.
Chief technology officer Mira Murati sidestepped questions about this in an interview earlier this year, saying “I’m actually not sure about that” when asked if videos from YouTube were used. A few months later, Murati took heat for admitting, rather inelegantly, that “some creative jobs maybe will go away”. (She later defended this remark in a lengthy X post).
But these comments from the most influential CTO in the AI sector should be alarming for their lack of transparency about what materials are being used to train its tools – and their frank acknowledgment that this technology will impact artists’ jobs. OpenAI announced it was launching its first Asia office in Tokyo earlier this year, with many suggesting Japan’s hands-off regulation playing a role in that decision.
