When asked O1, Gemini and DeepSeek what is 18999 ^ 9

OpenAI O1: 322534877887062061968054130143004170999 (used python exec)

Gemini: 3.22534878 * 10^38

Deep Seek: With a lot of reasoning, started off at ~3.21 * 10^38 and optimized its own solution using natural log and came to ~3.223 * 10^38.

Which is the better path for machine learning, solving like humans or machines using their compute capabilities or both