Put Gemini and GPT as coaches in Minecraft. Who can understand robot programming? · AuraTracer

SummaryAI generated

Multiple tech media reports indicate that artificial intelligence models such as Gemini and GPT were used in the game Minecraft to assist players with micro-operations as coaches. These experiments focused on the performance of machine learning in game strategies, sparking discussions about AI’s ability to understand and execute complex commands.

Related eventsRELATED EVENTS

2026-09-07 21:02The `GPT-6 Astra model was released, and OpenAI President Brockman confidently declared, “Welcome to the era of AGI.”
2026-09-08 19:21OpenAI released GPT-6 Astra, marking the arrival of the AGI era, with capabilities for autonomous computer operation and scientific research.
2026-09-08 22:40① Nvidia CEO Jensen Huang posted that OpenAI’s GPT-6 Astra, released last week, was trained using approximately 100,000 NV Link 72 clusters, and he believes that General Artificial Intelligence (AGI) has officially arrived; ② GPT-6 Astra can directly operate computers and software to perform complex tasks such as programming, reaching the most advanced level in multiple fields. OpenAI has announced the beginning of the AGI era.

Key entitiesKEY ENTITIES

DeepSeek-VL2-TinyGPT-4.1GeminiNVIDIANational Tsinghua UniversityPhi-3.5

Coverage · reports per dayLANGUAGE SPLIT

Entity relations

SignalsSIGNALS

Keyword heat

Gemini1
GPT-4.11
National Tsinghua University1
NVIDIA1
Phi-3.51
DeepSeek-VL2-Tiny1

All reports (1)SOURCES

↗

雷雷锋网 zh 2026-09-03 18:09

Put Gemini and GPT into Minecraft as coaches—who can understand a robot's micro-management?

# Place Gemini and GPT as coaches in Minecraft. Who can understand robot micro-control? ## Key Conclusions A joint team from National Tsing Hua University and NVIDIA conducted a comprehensive evaluation of mainstream large models such as Gemini 2.0 and GPT-4.1 against embodied intelligence tasks in Minecraft, including milking cows, shearing sheep, and killing spiders. The results showed that **Gemini 2.0-Flash maintained an accuracy rate of over 70% across 10 tasks, being the only model without any shortcomings; models like GPT-4.1 performed mediocrely, while traditional CLIP-based models failed completely.** ## Experimental Design The research team constructed 10 embodied intelligence task scenarios, including: - **Basic control**: Balancing a inverted pendulum (Cart Pole) - **Soft body manipulation**: Straightening ropes, translating water cups - **Spatial perception**: Kicking a ball into the net, clearing blocks into holes, opening drawers...