DeepL isn’t what is being touted as “AI” this week, though. DeepL is based on older translation technology (by which I mean “far more reliable”).

This is a shell game. Every time there’s a wave of “AI” it’s some new tech that shills sell as the answer to “real” computer intelligence. (It can never possibly be this, of course, because we can’t even define intelligence properly, not to mention making an artificial version of it.) There’s certain levels of hype. There’s a bubble (usually far smaller than this one, of course). Then the bubble pops and we enter the next AI Winter.

The small use cases for which the new technology is actually useful, however, loses the AI monicker and is just called “software”. Like what used to be AI for doing dynamic adjustment on pictures for night shots, HDR, etc. is no longer shilled as AI. It’s just … software that my phone has built in.

So currently “AI” means “LLM” outside of some very specific academic environments. DeepL is just software that (mostly kinda/sorta) works.