Nvidia's new AI framework trains an 8B model to manage tools like a pro

Meta has announced plans to significantly cut resources allocated to its metaverse project, a shift from CEO Mark Zuckerberg's earlier vision of the metaverse as the company's future direction. This decision reflects a reevaluation of priorities amid financial pressures and scrutiny regarding the project's viability.

AMD CEO Lisa Su addressed concerns regarding a potential AI bubble during WIRED's Big Interview event, stating that such worries are exaggerated and asserting confidence in the AI chip market's growth. She emphasized that from her perspective, the AI sector is not experiencing a bubble.

Cambricon Technologies has announced plans to more than triple its production of AI chips by 2026, aiming to capture market share from Huawei Technologies in China and to fill the gap left by Nvidia's exit from the market. This strategic move comes as the company seeks to enhance its domestic AI chip manufacturing capabilities amid increasing competition.

Nvidia and OpenAI have yet to finalize their anticipated $100 billion deal, leaving the future of their collaboration uncertain. This delay raises questions about the strategic alignment and operational readiness of both companies in the rapidly evolving AI landscape.

Jensen Huang, CEO of Nvidia, expressed skepticism about whether China would still desire the company's advanced H200 AI chips, even if U.S. export restrictions were relaxed. This statement reflects the complex dynamics between the U.S. and China regarding technology and trade.

A comparative study has been conducted on the object counting capabilities of two multi-modal large language models, GPT-4o and GPT-5, focusing on their performance in zero-shot scenarios using only textual prompts. The evaluation was carried out on the FSC-147 and CARPK datasets, revealing that both models achieved results comparable to state-of-the-art methods, with some instances exceeding them.

A recent paper has introduced a quantifiable framework for defining Artificial General Intelligence (AGI), proposing that AGI should match the cognitive versatility of a well-educated adult. This framework is based on the Cattell-Horn-Carroll theory and evaluates AI systems across ten cognitive domains, revealing significant gaps in current AI models, particularly in long-term memory storage.

Anthropic and OpenAI have recently showcased their respective AI models, Claude Opus 4.5 and GPT-5, highlighting their distinct approaches to security validation through system cards and red-team exercises. Anthropic's extensive 153-page system card contrasts with OpenAI's 60-page version, revealing differing methodologies in assessing AI robustness and security metrics.