Chinese Institutions Use U.S. AI Model Outputs to Train Domestic Rivals

Written by

in

More than 300 Chinese academic papers, patents, corporate disclosures, job postings and government documents show that using U.S. AI model outputs to train domestic systems has moved beyond academic experimentation into industrial applications.

Researchers have used OpenAI’s o1, GPT-4 and GPT-4o to generate training materials, then trained smaller models such as Alibaba’s Qwen2.5-72B, reducing operating costs while retaining some capabilities. U.S. officials believe such large-scale distillation could involve intellectual property theft, but Chinese officials say distillation is a standard AI development technique and have made no specific commitments.

Source:Dataconomy · Read Original

Comments

Leave a Reply

Your email address will not be published. Required fields are marked *