Three threads, twisting and turning.
A man who doesn't know Chinese sees some bizarre 字 and asks what they mean.
A man who doesn't know :guild_sign_1: sees a man yelling about the distinction between purple and violet and doesn't know the point of it.
A man looking at the machine, asking it a question in English about color-words in Chinese.
💬 "Hold your horses"
🔥 "Can the machine think?" A question any sane 人 wants to avoid, as any answer of "no" is likely to become obsolete in the next five years. Currently, the machine cannot "think". And, most of the time, when interacting with a computer, this is a good thing.
⚙️ as far as the question of "What is enough Chinese": we will explain tone sandhi at some point. 是为什么《你好》的《你》和《你》不听一样。
💡 if you ask the machine something it does not know from the training data, it will not figure it out. the machine does not think in that way. yet.
🌎 of course, this is the year of "Attention is All You Need". it would be heresy to use a more effective architecture.