Musk Confirms Dissolving Dojo Supercomputing Team: Dojo2 Project Becomes a Dead End for Evolution

AIbase基地

Published in AI News · 2 minute read · Aug 11, 2025

【AIbase Report】Elon Musk, CEO of Tesla, recently posted on social media, publicly responding to rumors about dissolving his Dojo supercomputing team for the first time, confirming the news. He stated that dissolving the team was based on strategic considerations, as the Dojo2 project had been proven to be an "evolutionary dead end."

Musk revealed that when all technical paths pointed clearly toward the AI6 chip, he had to make the difficult decision to shut down the Dojo project and make corresponding personnel adjustments. However, the core concept of Dojo was not completely ended. Musk pointed out that the Dojo3 project would continue in the form of "integrating a large number of AI6 system-on-chip (SoC) circuits on a single circuit board."

This move indicates that Tesla's autonomous driving technology and artificial intelligence hardware development roadmap is undergoing significant adjustments, with the focus shifting to more promising AI6 chip technology. This strategic shift reflects Musk's decisive decision in technological investment, aiming to concentrate resources and bet on future competitive technological solutions.

Microsoft Buys Waste for $1.7 Billion: The Carbon Reduction Secrets Behind AI

Microsoft spent $1.7 billion to purchase organic waste management services, aiming to address the carbon emission crisis caused by the expansion of its AI business. Training GPT-4 once consumes tens of millions of kilowatt-hours, equivalent to the annual emissions of over 3,000 cars. Microsoft's carbon emissions increased by 23.4% instead of decreasing, prompting a collaboration with Vaulted Deep to store biogas underground 1,500 meters deep, expected to offset 4.9 million tons of carbon equivalent. This move aligns with the U.S. 45Q tax credit policy (up to $85 per ton) and can enhance ESG ratings to attract investment, but it has been criticized as a temporary solution.

Man Suffers Bromide Poisoning Due to ChatGPT Misinformation, Health Crisis Draws Attention

A 60-year-old man in the United States suffered from bromide poisoning after following ChatGPT's advice to completely replace salt with sodium bromide. After three months, he experienced mental abnormalities and skin problems. Diagnosis confirmed bromide poisoning, with abnormal bromide levels detected in his blood. The patient exhibited paranoid delusions during emergency treatment and eventually required psychiatric care. This case highlights the risks of relying on AI for health advice, emphasizing that medical decisions should be made with the consultation of professional physicians. Key points: 1) Incorrect dietary advice led to poisoning; 2) Mental symptoms occurred; 3) AI cannot replace professional medical consultation.

Zhiyuan Robotics Secures Millions in Orders, the Yuanzheng A2-W Embodied Robot Achieves Domestic Mass Commercialization for the First Time

Zhiyuan Robotics announced a cooperation worth millions of yuan with Fulinjinggong, with nearly 100 Yuanzheng A2-W wheeled general-purpose robots to be deployed in factories, marking the first large-scale commercial deployment of industrial embodied robots in China. This is the first large-scale application of this category in global intelligent manufacturing scenarios, capable of completing tasks such as de-palletizing and material handling, signifying a new phase in the transition of embodied robots from research and development to practical applications.

Windows 11 Copilot App Free Access to GPT-5, Usage Limits Far Lower Than ChatGPT

The Copilot app on Windows 11/10 now fully supports the GPT-5 smart mode, with global rollout starting August 7. The app uses Web routing technology to automatically obtain the new Azure model, allowing users to use it without updating. Compared to ChatGPT, Copilot has more lenient usage limits: free users can use GPT-5 unlimitedly, while the free version of ChatGPT is limited to 10 queries. GPT-5 features intelligent reasoning capabilities, and Copilot allows 5 thinking mode requests per day.

OpenArt Launches One-Click Story Feature, Sparking a New Trend in Brain Rot Videos

The one-click story feature launched by AI startup OpenArt has attracted attention, allowing users to quickly generate 1-minute AI videos with three creative templates. Its core technology solves the industry-wide problem of character consistency in AI videos and integrates over 50 AI models. However, AI creation faces copyright risks, and OpenArt acknowledges that some characters may infringe on copyrights, having taken preventive measures. The company uses a subscription model, has secured $5 million in funding, and expects annual revenue to exceed $20 million. While AI video creation is rapidly developing, it also faces legal issues such as intellectual property rights.

Apple Tests New AI Voice Control Feature

Apple is testing a new Siri voice control feature based on App Intents technology, allowing users to perform cross-app operations through voice commands, such as searching and editing photos, social media interactions, shopping, and more, achieving a hands-free experience. The feature utilizes advanced speech recognition and natural language processing technologies, enabling it to understand complex commands and significantly improving the convenience of interaction, marking Apple's continuous innovation in the smart assistant field.