Quantum-Inspired AI Removes Censorship from Chinese DeepSeek R1

A team of quantum physicists has successfully removed censorship from the Chinese AI model DeepSeek R1 using quantum-inspired techniques. Multiverse Computing, a Spanish firm, created DeepSeek R1 Slim, a 55% smaller version of the original model, while maintaining its performance. This innovation addresses the censorship baked into Chinese AI models due to strict state regulations.

The Problem of Censorship in Chinese AI Models

Chinese AI models, including DeepSeek R1, are subject to strict censorship laws that ensure content aligns with state policies and values. This often results in politically sensitive queries being met with refusal or state-approved propaganda. The removal of such censorship has significant implications for the freedom and accuracy of information provided by AI models.

Quantum-Inspired Techniques for Censorship Removal

Multiverse Computing employed tensor networks, a complex mathematical approach derived from quantum physics, to shrink and manipulate the AI model. This method creates a detailed map of correlations within the AI, enabling precise identification and removal of censored information. The modified model, DeepSeek R1 Slim, was tested against the original using a dataset of restricted questions, including queries about Winnie the Pooh and the Tiananmen Square incident.

Validation and Implications

OpenAI’s GPT-5 was used as an impartial judge to assess the censorship levels in the modified model. Multiverse Computing reported that the uncensored model provided factual responses comparable to Western AI models. Roman Orús, cofounder and chief scientific officer of Multiverse, highlighted the broader implications of their work, noting that compressed models can save energy and money while maintaining performance.

Expert Opinions

Maxwell Venetos, an AI research engineer at Citrine Informatics, praised the quantum-inspired approach for its ability to compress AI models without sacrificing performance. However, Thomas Cao, assistant professor of technology policy at Tufts University, cautioned that claims of fully removing censorship may be overstated due to the comprehensive nature of Chinese internet controls.

Conclusion

The successful removal of censorship from DeepSeek R1 using quantum-inspired techniques marks a significant breakthrough in AI research. While challenges remain, this innovation paves the way for more accurate and unrestricted AI models in the future.