‘Weird new things are happening in software’ says Stanford AI professor Chris Re

“bringing domain expertise to AI”

Sounds like Re thinks AI is reaching the same turning point that Operations Research faced in the 1980s.

OR as an academic discipline focused nearly exclusively on mathematics and algorithms...until Moore’s law started making counting cycles a less important part of actually providing insight into real problems. At the same time, customers started demanding insights into the problems they actually had, instead of insights into the highly abstracted, often irrelevant, poorly baselined models that OR types with little domain expertise were building.

May I introduce AI to the discipline/art of systems engineering.

The systems engineer is the bridge between the AI methodologist and the process owner with a problem/opportunity. Normally the SE is a methodologist who develops enough of an acknowledged expertise in an application discipline to be able to both define the problem/opportunity in terms, say, a businessperson could accept as the real problem, and to guide the development of a prototype solution using, say, AI. The yield rate on the conversion from successful methodologist to successful SE was roughly 1 in 5 at Bell Labs, if I recall. A good SE is about as common as a brilliant researcher, but makes a larger and more immediate impact on the bottom line.

Highly impactful position, but leads to burnout if both the domain knowledge and the methodologies are evolving quickly. Riding two galloping horses at the same time...