Looking at examples I have encountered of what these new developments, more technically called "large language models (LLMs)," have achieved so far, I must say that they are still far from replacing us humans, although they may be of some help.
As many experts on artificial intelligence have warned us, they are very impressive in turning out what looks like intelligently written prose and poetry, but they are equally adept at spitting out quite unintelligent and even very weird stuff, what are called "hallucinations" — that is, strings of words, sentences, and paragraphs that are simply products of their "imaginations," if computers can be said to imagine things. They seem quite often to disregard simple realities. One of my favorite examples is the reply someone got when they asked a chatbot which was heavier, a kilogram of lead or a kilogram of cotton; it answered that, since lead is heavier than cotton, the kilo of lead was heavier. But I saw this a few months ago; the field is changing so rapidly that the same bot might not make that mistake today.
When asked to translate a source text, they can come up with a response that seems quite impressive — and sometimes it is. A fellow Japanese translator who is experimenting with this technology asked it to translate the start of a famous Japanese novel. What he received he put side by side with a recently published translation by a human translator, and much of it did look quite good in comparison in its style and fluidity. It was also quite faithful to the meaning of the source text.
But was that just a rare stroke of luck? Could any person who wanted an instant translation from any language to any other one be confident of getting such a good product? Given the rather wild hallucinations these bots are still coming up with, despite the improvements that have been made in recent months, I would strongly doubt it.
Personally, I will continue to work as I always have, using what the industry calls "computer-aided translation" programs when I find them helpful, but always keeping to the principle that my own brain must be responsible for the quality of what I turn out.
Comments