Fascinating. I had no idea people are still working on this sweetly optimistic project, like the architects writing a catalog of data taxonomies but for the unfenced territory of humanities knowledge. Sort of a modern Principia Mathematica with the Gödel of GPT4 waiting to blow it away.

Still 42 for Douglas Adams manually assigned token is a nice touch.

much of the AI projects would not really be possible without linked data, so it seems a bit silly to call it optimistic. the two are connected

I have a half hour YouTube video on that topic, maybe you'll find it interesting, but likely too long: https://www.youtube.com/watch?v=WqYBx2gB6vA

It was interesting. I don’t think cheap matters much in a world:where compute is still getting rapidly cheaper. Being certain is good, but the throwmtons of data i to massive models keeps beating everything else. I do think the models need to get some self doubt built in some how but I don’t think the data catalog will be more than post NOC fact checking. Fun times tho.

Edited: It was interesting. I don’t think cheap matters much in a world where compute is still getting rapidly cheaper. Being certain is good, but the throw tons of data into massive models keeps beating everything else. I do think the models need to get some self doubt built in some how but I don’t think the data catalog will be more than post hoc fact checking. Fun times tho. (My mom was an editor, she'd be horrified.)