It's surprising that AI models trained on general data can accurately predict rare events like hurricanes. The reason is that the physical world is "forgiving"; extreme phenomena are governed by strong physical structures and signatures that AI can learn effectively, even from a limited number of examples.

Ajay Banga argues that the real AI opportunity in emerging markets isn't large, power-hungry models. Instead, it's "Small AI"—localized applications on phones for tasks like medical diagnosis or farming advice. These are more feasible given limitations on electricity, computing power, and data sovereignty.

Counter-intuitively, successful weather and climate AI models are not trained on long-term data. They are trained to predict only the next six hours autoregressively. This surprisingly generalizes to stable rollouts predicting weather patterns for hundreds or thousands of steps into the future, enabling long-term forecasting from short-term training.

Models like Gemini 3 Flash show a key trend: making frontier intelligence faster, cheaper, and more efficient. The trajectory is for today's state-of-the-art models to become 10x cheaper within a year, enabling widespread, low-latency, and on-device deployment.

OpenAI achieved a major reduction in the cost of running its models through purely software and algorithmic improvements, such as quantization and smarter caching. This demonstrates that efficiency innovation can be as impactful as acquiring more hardware, suggesting a path to overcoming compute bottlenecks without relying solely on expensive chips.

Most AI weather models project the Earth onto a flat rectangle, causing simulations to become unstable and "blow up" over long periods. By incorporating the planet's spherical geometry using Fourier Neural Operators, models like ForecastNet remain stable for long-term climate rollouts, effectively becoming climate models.

While AI training requires massive, centralized data centers, the growth of inference workloads is creating a need for a new architecture. This involves smaller (e.g., 5 megawatt), decentralized clusters located closer to users to reduce latency. This shift impacts everything from data center design to the software required to manage these distributed fleets.

Previously, the biggest constraint in AI was compute for training next-gen models. Now, the critical bottleneck is providing enough compute for *inference*—the real-time processing of queries from a rapidly growing user base.

The combination of AI's reasoning ability and cloud-accessible autonomous labs will remove the physical barriers to scientific experimentation. Just as AWS enabled millions to become programmers without owning servers, this new paradigm will empower millions of 'citizen scientists' to pursue their own research ideas.