Recently, OpenAI andThe New York TimesThe escalation of the controversy has attracted a lot of attention. The New York TimesAccusing OpenAI of infringing the copyright of its articles, and detailing a series of crimes and evidence in the indictment. However, OpenAI denied this and put forward its own defense. In response to this dispute, there are mixed reviews and opinions.
The New York TimesIn its indictment, OpenAI is accused of using articles to train large models, generating highly identical works, and blaming false informationThe New York Times。The indictment cites a large amount of evidence, including a large amount of training data for GPT-2 and GPT-3The New York TimesThe content of the article. At the same time,The New York TimesIt was also noted that the content generated by GPT was almost identical to its original text, and more than 100 similar examples were listed in detail.
ForThe New York TimesFor articles that are paid to read, users can get the original content by talking to ChatGPT, bypassing the paywall. In addition,The New York TimesIt is also claimed that its**FlowBing, etcSearch enginesTruncated beard.
However, the author thinksThe New York TimesThe accusations are not entirely just. First of all, in terms of using publicly available information on the Internet to train large models, this is in line with the copyright lawFair Use"Scope. Secondly, while GPT generates content withThe New York TimesThe original text is highly similar, but this may be due to the phenomenon of "rumination", rather than OpenAI's deliberate manipulation of the model.
OpenAI: YesThe New York Timesresponded to the allegations and gave explanations on key issues. They said in December last year withThe New York TimesWhen negotiating cooperation, he was suddenly sued. Regarding the question of training data, OpenAI believes that the use of public information is:Fair Usecategory. Although there is a phenomenon of "rumination" in GPT, OpenAI has taken steps to limit it after discovering the problem in July last year, and the cases in the indictment are relatively rare.
OpenAI also responded that it did not deliberately manipulate the model, but blamedThe New York TimesDeliberately bootstrapping**. However,The New York TimesThere has been no response to this so far, which has made the situation even more confusing.
In this controversy, the parties have different views, forming a fierce ** debate. Some ** people have come out in solidarityThe New York Times, arguing that OpenAI should be held responsible for its infringement. But there are also those who thinkThe New York TimesTake the opportunity to knock on OpenAI for financial gain. In addition, AI boss Andrew Ng believes that training using online public information belongsFair Use, but there are also people who are interested in it".Fair Use"Question. Due to differences of opinion, it is still impossible to determine who is right and who is wrong.
Looking back at the whole controversy, I thinkThe New York TimesAlthough the dispute with OpenAI is inevitable, the change of attitude and the complexity of the two sides make it impossible to judge who is right and who is wrong. The New York TimesThe allegations are not comprehensive and fair, and OpenAI's response is plausible. As an outside observer, I can't accurately judge the true situation of the matter, so I can only wait for the outcome of the court trial. This dispute is of great significance to the AI field and copyright protection, and we need to further observe and study it to provide clearer guidance and norms for the future.
In short,The New York TimesThe dispute with OpenAI has attracted a lot of attention, but there are still many controversies and opinions. Only through a court trial can the truth be truly revealed. At the same time, this dispute also reminds us of the need for more in-depth thinking and standardization in AI development and copyright protection.