very few worthwhile tasks?
Follow-up: [very few 2-3 sentence descriptions connect (weblog)](/2023/06/30/very-few-2-3-sentence-descriptions).
What is a "worthwhile task"?
[highlighting added]
\@benedictevans
The more I look at chatGPT, the more I think that the fact NLP didn’t work very well until recently blinded us to the fact that very few worthwhile tasks can be described in 2-3 sentences typed in or spoken in one go. It's the same class of error as pen computing.
On Twitter Jun 29, 2023
Where to start? I think this tweet sort of treats sentences/information **and** "tasks" as overly atomic, decontextualized (and perhaps relies too heavily on "in one go").
My [dissertation](/diss) is about how data engineers heavily (and effectively?) rely on general-purpose search engines to do worthwhile tasks for their jobs. For decades they—and many other sorts of professionals—have navigated their work and the web through incantations much shorter than 2-3 sentences:
Those practices were not handed to them, but developed over time and are woven through their tools, organizational processes, and discourse:
This tweet seems to position people as previously (effectively) believing something about language and tasks that is akin to superficial pronouncements/readings of "_just_ google it". With some reflection we recognize that the 'just' is very context/audience/subject-dependent, indexical, and that communication requires real work[^1]. (Perhaps it is like telling someone embarking on a rigorous trail ultramarathon] to just put one foot in front of another...)
And it also reminds me of "google knows everything until you have an assignment 😂" — general-purpose web search has worked very well for tasks scaffolded within our environment and lives, but when we are thrust into something new new, there may be more work to do (jarring as that can be).
Or maybe I take issue with "can be described"? Is that what a search query or prompt is? A description?