This resistance mirrors a global trend, with authors including Sarah Silverman, John Grisham, and George R.R. Martin suing companies like OpenAI and Meta for using their works to train AI technologies. The Singaporean government's approach to seek consent is seen as unusually considerate, but the lack of specifics on copyright protection and compensation remains a concern. The legal status of training LLMs on copyrighted content is still uncertain, and the government's historically uneasy relationship with the arts community adds to the tension.
Key takeaways:
- The Singaporean government's request to use local writers' work to train a large language model (LLM) has met with resistance from the literary community, with concerns over copyright protection and compensation.
- The National Multimodal LLM Programme (NMLP) aims to address the bias of existing LLMs that are influenced by Western societies, and instead train an LLM on local material for more accurate references to Singapore's history, culture, and languages.
- The resistance from Singaporean writers is part of a global trend, with authors and publishers worldwide, including U.S. comedian Sarah Silverman and authors John Grisham and George R.R. Martin, accusing companies like OpenAI and Meta of copyright infringement for using protected work to train AI programs.
- Despite the criticism, the Singapore government's approach of seeking consent is considered rare and unusually considerate of writers' rights. However, the writers want more specifics on how their work will be used and protected, and what compensation they will receive.