When modeling a complex issue like malaria bed nets, don't start with every variable. Begin with a simple model of the 5-6 core drivers. This makes the model easier to understand, hold in your head, and debug. Add complexity later, once the basic dynamics are established and validated.

Beyond supervised fine-tuning (SFT) and human feedback (RLHF), reinforcement learning (RL) in simulated environments is the next evolution. These "playgrounds" teach models to handle messy, multi-step, real-world tasks where current models often fail catastrophically.

The choice between simulation and real-world data depends on a task's core difficulty. For locomotion, complex reactive behavior is harder to capture than simple ground physics, favoring simulation. For manipulation, complex object physics are harder to simulate than simple grasping behaviors, favoring real-world data.

The AI's ability to handle novel situations isn't just an emergent property of scale. Waive actively trains "world models," which are internal generative simulators. This enables the AI to reason about what might happen next, leading to sophisticated behaviors like nudging into intersections or slowing in fog.

Instead of simulating photorealistic worlds, robotics firm Flexion trains its models on simplified, abstract representations. For example, it uses perception models like Segment Anything to 'paint' a door red and its handle green. By training on this simplified abstraction, the robot learns the core task (opening doors) in a way that generalizes across all real-world doors, bypassing the need for perfect simulation.

It's tempting to think you can intuit the few factors a decision hinges on. This is often wrong. Complex systems have non-obvious leverage points. The process of building an explicit model reveals which variables have the most impact—a discovery you can't reliably make with intuition alone.

Theoretical knowledge from articles is insufficient for understanding AI models. True intuition is built through intensive, practical experimentation, such as feeding a model an entire codebase or extensive documentation. Pushing the AI to its limits is the fastest way to learn.

Game engines and procedural generation, built for entertainment, now create interactive, simulated models of cities and ecosystems. These "digital twins" allow urban planners and scientists to test scenarios like climate change impacts before implementing real-world solutions.

Creating realistic training environments isn't blocked by technical complexity—you can simulate anything a computer can run. The real bottleneck is the financial and computational cost of the simulator. The key skill is strategically mocking parts of the system to make training economically viable.