A few more comments from a long-time professional translator:
The unmet demand for communication across language barriers is immense, and, as AI quality continues to improve, translators and interpreters working in more faceless and generic fields will gradually be replaced by the much cheaper AI. That is already happening in translation: Some experienced patent translators I know have seen their workflows dry up in the past couple of years and are trying to find new careers.
This past August, I gave a talk on AI developments to a group of professional interpreters in Tokyo. While they hadn’t heard of any interpreters losing work to AI yet, they thought that some types of work—online interpreting for call centers, for example—would be soon replaced by AI. The safest type of human interpreting work, I suspect, is in-person interpreting at business meetings and the like. The interpreters’ physical presence and human identity should encourage the participants to value and trust them more than they would AI.
Real-time simultaneous translators are an interesting case. On the one hand, they are the rarest and most highly skilled, and national governments and international organizations have a long track record of relying on and trusting them. On the other hand, they usually work in soundproof booths, barely visible (if at all) to the participants. The work is very demanding, so interpreters usually work in twenty- to thirty-minute shifts; the voice heard by meeting participants therefore changes periodically. The result is less awareness that the interpreting is being done by real people, so users might feel less hesitation about replacing them with AI.
When I’ve organized or participated in conferences that had simultaneous human interpreting, the results were mixed. While sometimes it worked well, people often reported having a hard time following the interpretion on headphones. Question-and-answer sessions were often disjointed, with an audience member asking about some point that they seem to have misunderstood in the interpretation and then their interpreted question not being understood by the speaker. The interpreters were well-paid professionals, though not perhaps UN-level.
A few more comments from a long-time professional translator:
The unmet demand for communication across language barriers is immense, and, as AI quality continues to improve, translators and interpreters working in more faceless and generic fields will gradually be replaced by the much cheaper AI. That is already happening in translation: Some experienced patent translators I know have seen their workflows dry up in the past couple of years and are trying to find new careers.
This past August, I gave a talk on AI developments to a group of professional interpreters in Tokyo. While they hadn’t heard of any interpreters losing work to AI yet, they thought that some types of work—online interpreting for call centers, for example—would be soon replaced by AI. The safest type of human interpreting work, I suspect, is in-person interpreting at business meetings and the like. The interpreters’ physical presence and human identity should encourage the participants to value and trust them more than they would AI.
Real-time simultaneous translators are an interesting case. On the one hand, they are the rarest and most highly skilled, and national governments and international organizations have a long track record of relying on and trusting them. On the other hand, they usually work in soundproof booths, barely visible (if at all) to the participants. The work is very demanding, so interpreters usually work in twenty- to thirty-minute shifts; the voice heard by meeting participants therefore changes periodically. The result is less awareness that the interpreting is being done by real people, so users might feel less hesitation about replacing them with AI.
When I’ve organized or participated in conferences that had simultaneous human interpreting, the results were mixed. While sometimes it worked well, people often reported having a hard time following the interpretion on headphones. Question-and-answer sessions were often disjointed, with an audience member asking about some point that they seem to have misunderstood in the interpretation and then their interpreted question not being understood by the speaker. The interpreters were well-paid professionals, though not perhaps UN-level.