AI is all about Software Engineering (sb.thoughts.ar)

A recent Hacker News discussion highlighted that 73% of AI startups are mainly focused on prompt engineering, raising questions about the long-term viability of such approaches. While rapidly creating prototypes through prompt engineering can seem efficient, the construction of a sustainable AI product necessitates robust software engineering skills due to the inherent non-determinism of AI models. This complexity extends beyond mere functionality; product development increasingly revolves around creating internal frameworks to ensure that AI outputs meet acceptable performance metrics, which can often involve navigating a “multi-class” performance landscape rather than binary testing outcomes. The implications for the AI/ML community are significant. As the field grows, developers must contend with a deluge of choices in model architectures and parameters, leading to a potential “explosion of dimensions” that complicates product engineering. Furthermore, the evolving supply-chain landscape introduces vulnerabilities, making security and dependency management critical considerations. A robust understanding of software engineering principles is essential for addressing these challenges, as successful AI applications will require mastery of advanced concepts like state machines, parallelism, and complex architectural designs, far beyond what is encompassed in basic prompt engineering.