10 The Linguist Vol/63 No/1
thelinguist.uberflip.com
ARTIFICIAL INTERPRETING
Is it me, or does every article about the advent of Large
Language Models (LLMs) and AI published lately seem
designed to terrify human interpreters? Cue Paul Revere
1
riding around yelling "The robots are coming…". Now, I
understand the reason for the hype. People fear what
they don't know. So please bear with me through the
irony, but I recently asked ChatGPT3.5
2
"Will AI replace
human interpreters?" in an effort to get some clarity on
the subject. Not surprisingly, the response adequately
addressed the topic, but (and this was a comfort) was
also incomplete.
While some might attribute the incomplete answer to
the fact that ChatGPT has no knowledge base past 2021
and no internet access, or to my newbie-level prompting
ability, I prefer to think of it as indicative of the fact that
AI does not yet have General Intelligence,
3
and is a long
way from gaining it. It has no self-awareness and is
unable to determine context or draw on knowledge of
the world (what some deem 'common sense') to make
decisions. In short, it exists but it cannot 'think'. That
said, the answer it gave was not bad.
ChatGPT deemed the human interpreter superior to
AI-generated interpreting for several reasons. First, AI
has no contextual understanding of what is being said
and has difficulty with cultural nuance, and therefore with
idiomatic expressions, which are culturally bound. As a
Holly Silvestri considers the big questions around artificial
interpreting and the work to ensure it can be used ethically
THE ETHICS OF AI
ON BALANCE
A central issue is not
only whether AI could
ever be more effective
than the human brain,
but also who could
be held responsible
when it fails
©
SHUTTERSTOCK
