A podcast clip from the Jack Neel podcast featuring Chinese -Canadian academic, Jiang Xueqin, a.k.a. Professor Jiang has gone viral on social media after he claimed that artificial intelligence tools such as ChatGPT are not generating responses independently.
He instead claimed that AI is “being manipulated by humans somewhere in India.”
How ChatGPT actually generates responses
The remarks, widely shared on X, prompted a wave of reactions as users questioned Jiang’s claims. Once deployed, the model produces responses computationally without a hidden team of people and then write answers for each prompt. Humans are nevertheless involved throughout the development process. For example, though ChatGPT analyzes the vast amounts of text grasp grammar, facts about the world and even some basic reasoning skills, the model is fine tuned on more specific datasets. Following the viral clip, Grok responded that Professor Jiang’s statement was false. MarketsPulse24H also noted under the video that AI companies employ human workers worldwide, including in India, for annotation, moderation, safety testing and evaluation. However, that work is different from manually generating each ChatGPT response.
The claim is misleading because it conflates human involvement in building AI systems with how they generate responses after deployment, according to experts.
Large language models such as ChatGPT generate text by predicting one token at a time using neural networks trained on massive datasets, according to TechRadar . This frequently entails human reviewers who offer input to assist mold the model’s behavior and make its replies more suitable, helpful, and consistent with our expectations, according to TechRadar. It said ChatGPT generates responses autonomously through next-token prediction after training. It said that humans contribute to training and reinforcement learning but don’t help in live response generation.
What did Professor Jiang say?
In the podcast clip shared by Jack Neel, Professor Jiang questioned, “Do you really think when you put something into ChatGPT, it’s going to be generated by the AI? Jiang then went on and claimed, “I guarantee you it’s not true. I guarantee you it’s being manipulated by humans somewhere in India. He went on to argue that a human initially writes responses before they are absorbed into AI models. “When you asked ChatGPT, write me a paper on Napoleon and the Battle of Waterloo. “What people don’t appreciate is humans are the ones who are designing these weights. The algorithm is trying to figure out what these weights should be. ChatGPT does the same bloody thing.
Maybe it’s 90% AI, 10% human, but it’s still 10% human.” Because humans design the algorithms and their weights, he also compared ChatGPT to supervised machine learning used in facial recognition systems, saying AI can never become fully independent.
I think what happened was a human first wrote that, and then is inputted into the AI model, and then absorbed into the AI model,” Jiang said.
