The LLM Gamble

Using a large language model (LLM) is likened to playing a slot machine due to the unpredictable yet potentially delightful nature of its responses. The financial model reinforces this metaphor, as users pay for both input and output tokens, with the more expensive output being unpredictable in length and quality. This means users are charged for responses even when they are incorrect or useless, much like losing a gamble. While subscription models abstract this per-use cost, they introduce opaque usage limits that can still make them loss-leaders for providers. This raises questions about the long-term sustainability of a business model that requires customers to pay for unpredictable and often faulty results.