资深科技投资者：如果没有Scaling Law的突破，2024年AI就崩了

Core Insights - The launch of Gemini 3 validates the effectiveness of the scaling law in AI development, demonstrating that AI can achieve significant advancements even with limited hardware capabilities [1][2] - The AI industry was at risk of stagnation due to a lack of next-generation computing power, but new reasoning mechanisms have allowed for continued progress [1][2] - The emergence of two new scaling laws has enabled AI to maintain growth during a hardware transition period, filling the gap left by the absence of Nvidia's next-generation chips [3][5] Group 1: Scaling Law and AI Progress - Gemini 3's release is a milestone that confirms the pre-training scaling law remains valid, which is crucial for investors as it indicates that capital expenditures can translate into stronger AI performance [2] - The AI industry's recent growth cannot be solely attributed to hardware advancements; without new reasoning capabilities, a significant downturn was expected starting mid-2024 [2] Group 2: New Mechanisms Driving AI Advancement - The introduction of reasoning capabilities has dramatically increased AI intelligence, with a notable jump from 8% to 95% in performance following the release of a reasoning-capable model by OpenAI [3] - Two new scaling laws have emerged: Reinforcement Learning with Verified Rewards and Test Time Compute, which allow AI to evolve and enhance its performance through logical reasoning and extended computation time [4][5] Group 3: Future Outlook - The AI sector is transitioning from relying solely on hardware to achieving value through logical reasoning and verification, indicating a new phase of growth [5] - Future advancements in AI capabilities are anticipated as these new scaling laws are implemented on more powerful hardware, such as Nvidia's Blackwell models [5]