Among the various social problems brought about by the rapid development of AI, the problem of copyright infringement is particularly serious. Recently, a joint research team from the State University of New York at Stony Brook and Columbia University published a latest research paper, which verified a crucial hypothesis: "AI that is specially trained (through outstanding copyright-protected creations that have already been released) understands readers' preferences better than professional writers." The key point is that the cost is extremely low.
·The paper is titled "Readers Prefer Outputs of AI Trained on Copyrighted Books over Expert Human Writers". The experimental process is to let the AI receive training on the works of literary giants such as the Nobel Prize for Literature, and then compare the generated works with the works of invited human professional writers, and invite experimenters to judge.
·The research results obtained are surprising. Without indicating the source of the work in advance, more experimenters tend to judge that AI works are better.
·Here comes the problem, according to the current market conditions, the publishing organization invites professional writers to create a work of 100,000 words and the remuneration is about 25,000 US dollars. However, using AI can reduce 99.7% of the cost and achieve better market expectations.
·The conclusion of the study seems to predict,Future publications may no longer be original works made by human brains, but more and more cleverly hidden AI works, and it is difficult to judge whether they infringe copyright. There will be more and more such things.
