Good Start Labs ran an experiment where an AI was trained on a railroad game. The surprising result: one version of that trained AI later performed better at financial research tasks. This suggests that skills learned in a simulated environment can carry over to a completely different domain.
The crucial factor was not the game itself but how the training was structured. The researchers found that variations in training design led to different outcomes—only one particular approach produced the transfer to finance. This points to the importance of deliberate training methodology over raw game exposure.
Since this is a single source, there are no conflicting accounts to compare. The finding is preliminary, but it raises interesting questions about how abstract problem-solving abilities from games might be repurposed for real-world applications like financial analysis.