3 ms·
This seems like an area for AI research. These are the points where a general AI will have a hard time, because language parsing is not enough. It has to “kno
by shireboy 4y ago
This seems like an area for AI research. These are the points where a general AI will have a hard time, because language parsing is not enough. It has to “know” the context -of say MIT’s Green Building vs green building- and be able to determine which is being referenced. Maybe not the best example, but here is one interpretation from Dall-e, which I assume is not accurate: https://labs.openai.com/s/DQQ2lJP0psNqIGLy3XJa4usg https://labs.openai.com/s/DQQ2lJP0psNqIGLy3XJa4usg
- numpad0 4y agoThat is also one of failure modes of translation, sometimes referred to as "engrish". Even human translators can't avoid this type of "mistakes", there just never are full context in texts. 1: https://www.dailymail.co.uk/femail/article-4398258/ https://www.dailymail.co.uk/femail/article-4398258/