I’m with Stephan Heuel on his (concept of the) AI loop. A big part of the evolving AI landscape is the need to find new language to talk about it and to shape ideas. Spicy autocomplete? Mode collapse? Ralph loop? – Etc.
One important function of language is creating categories in order to better understand a phenomenon. Dan Shapiro, CEO of Glowforge, has written up one such categorization of AI use for programming. The reference to the SAE1 Levels for automated driving and the descriptions make it clear that Dan views these as maturity levels through which organizations progress, from Level 0 – Spicy Autocomplete2 to Level 5 – Dark Factory3.
Whether you agree or disagree with Dan’s premise and the promise, it’s an interesting read and the levels describe quite distinct modes of applying AI to programming.
Footnotes
SAE International (from “Society of Automotive Engineers”) is an international association and standards organization. It proposed a classification system of driving tasks for (partly) automated driving.↩︎
Note how there are actually six zero-indexed levels. Maybe the first doesn’t count as AI use? Or it’s simply a call-back to SAE Levels.↩︎
I (re?-)learned that a dark factory is called that because it requires no human presence to run. Lights can stay off.↩︎