The Model Possibilities Frontier
Resources are finite: every general-purpose model you build costs you specialized models you could have built. Pick your point on the frontier.
Contents
Use it when Deciding whether to invest in one general-purpose model or several specialized ones.
Every company’s resources are finite: compute, memory, even people. Investing in general-purpose models means giving up specialized models you could have built instead. That is the opportunity cost, mapped onto the economist’s production possibilities frontier.
- The red curve is a steep trade-off. Investing in general-purpose models prevents you from producing specialized ones that could have been better for your business.
- The orange curve is a more lenient opportunity cost, reached through hardware improvements, efficiency gains or more resources. It could be the same company later, or a different company that is more capable resource-wise.
Companies with massive investments can build and serve more general-purpose models; smaller niche companies usually build for specific use cases.
Pick your point on the curve. The aim is allocation efficiency: the choice along the frontier that gives the best bang for the buck for your company and your customers. Are you after versatility, or do you know your use cases well enough to optimize for speed and performance? The same trade-off applies beyond AI: broad versus niche marketing, generalist versus specialist hires.