SG literary community push back on Govt efforts to train AI language model

SINGAPORE: Members of Singapore’s literary community have been unhappy with efforts from the government to use their works in training a large language model (LLM), machine learning designed to comprehend and generate human language text.

The Infocomm Media Development Authority’s (IMDA) Southeast Asia-focused LLM, the National Multimodal LLM Programme (NMLP), had been described in December 2023 as “a base model with regional context that can understand Singapore’s and the region’s unique linguistic characteristics and multilingual environment”.

In April, however, writers expressed concerns over copyright and compensation concerns. A survey was sent by IMDA on March 28, giving the writers until Apr 7 to provide consent for their works to be used for NMLP, though the deadline was later extended to Apr 15.

“If implemented without due consideration and safeguards, AI software will start assimilating material that would otherwise be copyrighted, and adversely impact the livelihood of existing authors and publishers,” The Straits Times quoted Ethos publisher Ng Kah Gay as saying.

Journalist Charmaine Lim, who wrote the ST piece, noted in a LinkedIn post that using authors’ works without their permission not only violates copyright laws, “it would also deprive writers of all kinds of their livelihood and compensation for the hard work that went into creating their written pieces.”

“That’s barely scratching the surface of all the other complications with LLMs – plagiarism, stealing someone’s written voice and publishing fake works in their name, protecting the dataset, the sheer ethics of it all, etc,” she added.

Earlier this week, in a piece in tech publication Rest of World, journalist and writer Nicholas Yong discussed the matter further, noting the literary community’s fierce reaction to IMDA’s request.

He spoke to noted author and Singlish doyen Gwee Li Sui, who declined for his works to be used for training the NMLP.

“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,” Mr Yong quotes the prolific writer as saying.

Mr Yong also placed the resistance put up by Singapore writers in the larger context across the globe where other artists have similarly declined to have their works used for AI training programs, adding that such well-known authors as John Grisham and George R R Martin have said that their copyrights have been infringed by OpenAi in training ChatGPT.

He also noted, however, the rarity of writers’ consent being sought in such instances, quoting Nuurrianti Jalli, an assistant professor at Oklahoma State University as saying that “ the Singapore government’s approach stands out as unusually considerate of writers’ rights. But writers understandably also want to know specifics.”

Interestingly, over Facebook on Mar 10, Mr Gwee posted that Mr Yong’s Rest of World piece had itself been hijacked by AI on Medium, in a piece titled, “Writers and publishers in Singapore reject government’s AI training proposal amid concerns.”

“Nicholas Yong, AI just stole the news article you wrote,” Mr Gwee wrote. /TISG

Read also: Artificial intelligence in Singapore schools: Friend or Foe?