Writers and publishers in Singapore reject a government plan to train AI on their work
When the Singaporean government asked local writers if they would agree to having their work used to train a large language model, it probably did not expect the country’s tiny literary community to react so fiercely.
An email sent late in March said that the National Multimodal LLM Programme (NMLP) aimed to address the bias of existing LLMs that have “disproportionately large influences” from Western societies. Singapore’s own LLM, trained on material produced locally, would have more accurate references to the nation’s history, colloquialisms, and culture and train on widely spoken languages, such as Malay, Mandarin, and Tamil, it said.
“The stages of planning [for the LLM] before writers are even considered as worth consulting do not inspire confidence that my interest will be a priority,” Gwee, author of more than a dozen books, told Rest of World. There is also little clarity on how the works would be protected from being used “for purposes other than what is now claimed as public service towards cultural representation,” he said.
The email initially gave respondents 10 days to respond to a survey. But it had few details on compensation or copyright protection. So Gwee declined to let the LLM train on his works, including the first book written entirely in Singlish — a creole language that is a blend of Singaporean slang and English and is widely spoken in the country.
