Is AI Being Controlled by Humans in India? Professor Jiang's Podcast Claim Explained

A viral podcast clip featuring Chinese-Canadian academic Jiang Xueqin, popularly known as Professor Jiang, has sparked a debate over how artificial intelligence systems such as ChatGPT generate their responses. During a conversation with podcaster Jack Neel, Jiang questioned the idea that AI systems independently produce the information they provide to users. He claimed that artificial intelligence is being “manipulated by humans somewhere in India”, a remark that has since gained attention on social media and triggered questions about whether people are involved in generating AI responses behind the scenes.

Jiang's argument focused on the role humans play in creating and training artificial intelligence. He suggested that information produced by humans eventually becomes part of the material used to train AI systems. Using historical information as an example, he questioned whether an AI system could produce an answer about an event such as the Battle of Waterloo without relying on information that was originally created or documented by people.

The comments have led to some interpretations that human workers in India may be manually writing answers for ChatGPT users. However, there is no evidence presented in the viral clip to establish that people in India are sitting behind computers and composing individual ChatGPT responses in real time. Large language models work through a different process. During training, AI systems process enormous quantities of data and learn statistical patterns through mathematical parameters. Once trained, the model can use those learned parameters to generate a response when a user submits a prompt. The response is produced computationally rather than being manually written by a person for each individual query.

Humans do play an important role in the development of AI. People can be involved in preparing training data, evaluating responses, providing feedback and helping developers improve the behaviour of models. AI companies also use human workers for tasks including data annotation, safety evaluation and other forms of model testing.

India is an important part of the global technology industry, and workers in the country have participated in various technology and AI-related activities. The existence of such work does not demonstrate that Indian workers are secretly generating live ChatGPT answers for users. The key distinction is between training and inference. Human feedback and other human-created information can influence how an AI model is developed and refined during the training and evaluation stages. When a user later asks a question, however, the trained model generates the response using its learned parameters and the information contained in the conversation.

The controversy surrounding Jiang's comments therefore appears to stem from a broader discussion about how much human influence exists within artificial intelligence. While AI systems are developed, trained, evaluated and governed by people, that does not mean a human is necessarily involved in producing every answer a user receives. The viral claim has consequently raised a legitimate question about human involvement in AI development, but the specific suggestion that humans in India are manually manipulating individual ChatGPT responses has not been substantiated by evidence presented in the podcast clip.

As artificial intelligence becomes increasingly common, understanding the difference between human involvement in developing an AI model and human intervention in an individual AI conversation remains important. The two processes are fundamentally different, and confusing them can easily lead to misleading conclusions about how modern AI systems operate.