ChatGPT creator acknowledges that it is “impossible” to train his AI without violating copyright

2024-01-11 19:45:55

ChatGPT could not function without violating the Copyright. Over the last few months, companies and artists from multiple sectors have supported this argument in their complaints against the popular application artificial intelligence (AI). The company that created this text generator, OpenAIconfessed this Monday that they are right.

The use of generative AI, such as ChatGPT, worries about possible data leaks and economic losses

The startup led by Sam Altman recognized last Monday that “it would be impossible to train the current models of IA without copyrighted material”, as he advanced The Telegraph. He did so in a document presented to the House of Lords, the UK Parliament, where possible legislation to resolve that problem is being studied.

ChatGPT is a ‘app‘ capable of interacting with users and generating all types of texts based on their requests. That is possible because your system has been trained with large volumes of data extracted from internetmost of them works protected by the so-called copyright. This questioned model has allowed OpenAI, promoted with funding from Microsoftreach a valuation that would exceed 80,000 million dollars.

OpenAI thus recognizes that its model would not work respecting that right. “Since copyright today covers virtually every type of human expression—including blogs, photographs, forum posts, code snippets, software and government documents—it would be impossible to train today’s leading AI models without using copyrighted materials.”

The company assures that not resorting to these works would produce poor AI systems. “Limiting training data to public domain books and drawings created more than a century ago could make for an interesting experiment, but it would not provide AI systems that meet the needs of today’s citizens,” they explained.

From writers Graphic artists, including large media outlets such as ‘The New York Times’, have sued the company for commercially exploiting their creations without consent or prior payment, which they consider a “misappropriation.” However, OpenAI considers that “legally copyright does not prohibit the training” of AI systems.