Quantum Physicists Decensor DeepSeek R1; Google Launches Gemini 3

Quantum physicists at Spanish firm Multiverse Computing have successfully created a version of the DeepSeek R1 AI model that removes the censorship built into the original by its Chinese creators. Meanwhile, Google has unveiled Gemini 3, a major upgrade to its multimodal AI model, featuring advanced reasoning and agent capabilities.

DeepSeek R1 Decensored by Quantum Techniques

The original DeepSeek R1 model, developed in China, was subject to strict regulations that ensured content aligned with Chinese laws and socialist values. This resulted in built-in censorship layers during the model's training. Multiverse Computing used quantum-inspired AI techniques to develop DeepSeek R1 Slim, a 55% smaller model with comparable performance. This allowed them to identify and remove the censorship, enabling the model to answer sensitive questions similarly to Western AI models.

Google's Gemini 3: Advanced Reasoning and Agent Capabilities

Google's Gemini 3 represents a significant upgrade to its flagship multimodal AI model. The new version boasts improved reasoning abilities and more fluid multimodal capabilities, allowing it to work seamlessly across voice, text, and images. One of the standout features is Gemini Agent, an experimental tool designed to handle multi-step tasks directly within apps. This agent can connect to services like Google Calendar, Gmail, and Reminders, executing tasks such as organizing inboxes or managing schedules once granted access.

Implications and Future Directions

The decensoring of DeepSeek R1 highlights the potential of quantum-inspired AI techniques in modifying and enhancing existing models. This development could pave the way for more open and unrestricted AI tools. Meanwhile, Gemini 3's advanced capabilities mark a step forward in AI productivity and automation, with potential applications in personal and professional settings.