We ran the same Monte Carlo simulation using largely the same assumptions as we did with ChatGPT, although we did specify with Claude 6 million as the max for works (even though it can theoretically be some number under 7 million). The results, not surprisingly, nearly the same.
ChatGPT probabilities of damages amount in Anthropic
- Probability < $1B: 4.69%
- Probability > $1B: 95.31%
- Probability > $10B: 64.28%
- Probability > $50B: 22.29%
- Probability ≥ $100B: 9.68%
Claude probabilities of damages amount in Anthropic
- Probability < $1B: 4.97%
- Probability ≥ $1B: 95.03%
- Probability ≥ $10B: 58.92%
- Probability ≥ $50B: [25.75%]
- Probability ≥ $100B: [14.23%]
But Claude’s probabilities were higher for the 2 highest damages award tiers. A staggering 14.23% for an award over $100 billion! On ChatGPT, it was only 9.68%.
We used 50,000 simulations. The highest award was $872.58 billion!
Here are some of the key figures:
Here’s a pie chart of the various damages awards. Unlike ChatGPT’s results, the highest pie slice was for an award in the range of $1 to $5 billion (20.44% probability), whereas on ChatGPT it was for an award in the range of $10 to $25 billion (25.55% probability).
Related Story