Adam Chlipala: Expensive AI adoption is a status signal, not optimal engineering

In a LessWrong post, MIT researcher Adam Chlipala contends that modern AI ecosystems are shaped less by pure engineering trade-offs and more by signaling incentives rooted in evolutionary psychology. Drawing parallels to potlatch ceremonies where leaders destroy wealth to display status, he argues that deploying expensive, flashy large language models (LLMs) serves as a costly signal of organizational fitness and individual engineer prowess. This creates a perverse incentive: teams gravitate toward computationally heavy, general-purpose AI (like deep learning) even when cheaper, more reliable alternatives (e.g., symbolic reasoning and formal verification) could solve specific problems faster and with greater guarantees.

Chlipala advocates for a paradigm shift that deliberately avoids AI's hardest challenges—simplifying problems via computer-oriented communication, rearranging physical environments for easier computer vision, or adopting better programming languages—rather than brute-forcing them with LLMs. He suggests that the current status hierarchy rewards “socially legible” hard problems, yet real engineering success may lie in making problems trivial. The post urges professionals to critically evaluate whether their AI choices are driven by genuine efficiency or by the subconscious desire to signal status through expensive, conspicuous technology.