Let’s pause for a moment and actually consider what AI technology is actually doing in order to translate or to give texts.

At a basic level, AI works by taking tons of existing data and identifying patterns. Those patterns are then used to give an answer for a question (ie. translate this…). These patterns are generally pretty basic (like, ChatGPT seems to be pretty advanced, but all it is doing is predicting what the next word/phrase is based on a massive database of writing.) AI has absolutely no conception of what it is saying; it is just spitting out a series of patterns as generated by its algorithm and background data.

Working backwards, this tells us that an AI has to have developed a pattern before it can give an answer, and my extension, it has to have already seen a lot of data that supports the answer. By extension, somebody would have had to ALREADY done a whole lot of translating and then given those translations to the AI before it can be useful to you. And, because languages change over time (through slang and local phrases, among other methods), this inherently means that there is a perpetual gap in the knowledge of any translation AI that makes having SOME people available to translate absolutely necessary, at least barring a complete reinvention of how AI is currently done.