Goldman's Skeptical Point of View on AI Scaling

Goldman Sachs presented a skeptical view on AI scaling, challenging the common belief in scaling laws that say doubling data and compute capacity will double model capability. Their professor from MIT disputed linear or exponential scalability, raising doubts on the continuous improvement of models with increased resources. This skepticism aligns with the uncertainty around whether each subsequent AI model iteration significantly outperforms the previous ones. Goldman's research head echoed similar concerns to Sequoia, emphasizing the necessity for substantial returns from the massive trillion-dollar investments in AI infrastructure to justify the upfront CapEx.