Hofstadter reading

I think it is pretty obvious that AI translation fails at translating more complicated sentences. While AI has evolved to do many tasks that can best humans like in chess or jeopardy. Those tasks revolve on something that has a definite outcome like the best move or a correct answer. But I don't think translation always has a 'correct answer'. I believe there are many approaches you can take to translation. 

AI translation can't think, or at least how a human does, outside of what is given to it because it has to formulate a situation with what is presented. If you give a google translate a passage with context about its situation, it still won't be able to understand it. Unlike if you give AI a chessboard with pieces, it will be able to find the best move for that turn. There are things between the lines that AI or google translation can't understand. Even words that have multiple meanings or meanings that aren't explicitly written in dictionaries won't be translated correctly. AI in its current state won't be able to translate elegant sentences into another language (I don't even think it can translate Shakespeare into modern English) but maybe in the distant future it can, but maybe by the time that happens AI might have already taken over the world. Who knows. 

Machine translation example:

山路来て なにやらゆかし 菫草

Come on the mountain road

閑さや 岩にしみいる 蝉の声

The voice of a cicada squeezing into a quiet rock

秋深き 隣は何を する人ぞ

What do you do next to me in the deep autumn?

 

Brian