The incentives of the model providers are not fully aligned with users.
- The incentives of the model providers are not fully aligned with users.
- Over time, those incentives will become increasingly misaligned.
- Ideally you want your harness to be made by someone other than the model companies.
- The model company wants you to use as many tokens as possible.
- Chatty answers.
- Or to ask a question that you likely won't see for awhile, so when you finally get back to it and answer, you've fallen out of the cache and have to pay the cold rate.
- Also, the model company wants to be able to train on your inputs, to distill your use cases into the model.
- Finally, model companies would like for you to be more engaged.
- It's not that they'll actively create EngagementMaxxing models… it's that they'd have to actively try not to.
- Their incentives are for you to maximize engagements and let the tokens flow.