Can tech companies learn to love cheaper AI models?

The artificial intelligence industry is re-evaluating its dependence on the largest and most compute-intensive models. Driven by mounting costs and evolving efficiency needs, industry experts like Brian Armstrong predict that a vast majority of AI workloads will shift to cheaper, smaller models that perform equally well for most tasks. This transition threatens the revenue models of major AI labs, as enterprise users prioritize cost-effective performance over simply using the most advanced available technology. While it remains uncertain if businesses will fully adopt this shift, initial tests—such as those by Harvey—demonstrate that inference costs can be reduced significantly without sacrificing quality.

(Source：TechCrunch)