One of my favorites (and perhaps most confusing) is in Japanese scenes with old men. Japanese does not need to use any pronouns in a sentence as the different points can just be understood, however, Japanese also has like 50+ different pronouns to choose from should they want to pick one (often with different genders, politeness levels and other historical contexts attached). Anyway some old men use the term 儂 to refer to themselves and this is pronounced わし or "washi". But there is a much more common word for "washi" being 鷲 which means "eagle". And since both of those kanji characters are not well known among Japanese speakers they tend to simplify them using just the kana or phonetic "alphabet" as わし. And thus machine translation gets these scenes containing わし wrong with weird translations like "He said 'eagles like giant breasts'." instead of "He said 'I love your huge tits'."
I am not sure how or why "ika" could be corrupted into "iku" or vice versa, however, "ika" is indeed squid but can also be used as a sound effect implying some moist repetitive motion. It is true that "iku" means "go" or "to go" and the Japanese do use it in the context of climaxing much as we use "come" so it probably should be translated as something like "cumming" in the right context. It is a bit comical to hear Japanese people involved in sexual activities to say "iku iku iku" (notice the lack of pronouns) meaning "I am going; I am going; I am going" when we might say "I am coming; I am coming; I am coming". Japanese verbs also conjugate but not much like other languages. For example, Japanese has a present and a past tense but no future tense so it is very normal to say something like "I go to the store tomorrow".
One of the more problematic things with Japanese translation is the incredibly large number of homonyms in the language. Here are just a very few examples:
You must be registered to see the links
,
You must be registered to see the links
and
You must be registered to see the links
. This forces Japanese readers and translation to depend on complex Chinese "kanji" characters and/or context. Machine translators can often get the Chinese character meaning right (although they can have multiple readings and meaning too making things nontrivial), but raw kana (either hiragana or katakana) almost entirely depends on context. And they also use a large number of onomatopoeia (even for actions that we would not consider like mostly silent actions).
The main language grammar itself is actually considerably simpler than English but the complexity exists in the many subtleties, noted earlier.
