During the pandemic, we’ve seen several companies do extremely well adapting to the revolution in the workplace. With all that’s happened to linguists over the last year and a half, you might think the worst is over. However, with businesses and governments trying to re-start after lockdown protocols many of us have seen the ‘hiring’ signs on business all over town. This trend is particularly worrisome to me because the labor shortage is being mitigated by commercially available artificial intelligence. Most of the time we are accustomed to automation taking over manual labor jobs or eliminating repetitive administrative tasks, but the pandemic inflamed sore spots more broadly throughout our labor markets. The most obvious example: the shortage of medical care workers and the general need for employees to isolate themselves from one another in meatpacking, manufacturing, and shipping. Now, commercial artificial intelligence (Ai) is moving into a much more sophisticated area of production: the knowledge economy. Content production for news, accounting at small to medium-sized businesses, and graphic design have already experienced the impact from low-cost Ai solutions. Interpreters and translators are the latest professionals to confront a significant challenge from a commercial Ai solution.
A few days ago Zoom announced its acquisition of a real-time translation start-up called Kites. It’s a big deal because of Zoom’s rise to dominance as a result of the remote work revolution. Social media is buzzing with interpreters and translators opining on the development and rightly concerned for their careers. I, for one, am not quite shaking in my boots. There are a few advantages legal interpreters have over machines and that will likely not change anytime soon. However, the real question is: in what capacity will human interpreters work alongside Ai?
My thinking is Zoom’s Ai makes it easy to keep costs down for certain types of litigation. Family law comes to mind. In Texas, there are no court-appointed lawyers for family law cases. In general, family law cases are very expensive for families when it comes to taking time off from work, hiring child care, and paying attorney’s fees. Courts can leverage Kites to keep their costs down and make it more convenient for families, and especially single parents. However, keep in mind Kites must be used in conjunction with Zoom. I don’t think courts will find it very useful for contentious litigation. Using Zoom also makes sense for family law cases rather than meeting at the courthouse. In my recent experience Kites can eliminate conflicts regarding restraining orders and general privacy concerns. This makes Kites a no-brainer for AG courts who need interpretation. If you’re working in the AG or areas of family law, you might want to consider exploring another niche. Final divorce decrees will likely remain in person but that is probably it.
For personal injury litigation, I believe we will see the rise of post-edited consecutive interpretations. It is easy to control the sound when there are a limited number of parties in a room. If each person gets a computer screen and a microphone, the AI can simply create a voice-to-text scenario. It would go like this: The lawyer asks, “Where were you on the night in question?” The Spanish speaking witness reads on her screen, “¿Dónde estuvo durante la noche en cuestión?” She responds in Spanish, “Estuve en la casa de mi novio.” But the AI makes an error and prints, “Estufa en la casa de mi novio.” The lawyer also sees the error on his screen, “In the stove at my boyfriend’s house.” At this point, the interpreter steps in corrects the error and the deposition continues.
We are all going to need to find new and more well-defined, specialized niches. Building a niche is going to be incredibly important to continue to earn well in the AI economy. We won’t be able to earn as easily. Getting paid to sit around the courthouse may become a thing of the past. But there are a few advantages humans have over machines, especially in a legal context.
Machines cannot be sworn in as witnesses
It will probably be hundred years before a computer can be dutifully be sworn in as an interpreter. There needs to be a lot of money and human effort to support laws making it legal to use AI translators, especially in criminal cases (not that they won’t try!).
Humans are masters at variation and style
One of the great accomplishments of human cognition is style. Using fresh ideas to describe events, places, people, and things. Although AI is improving it still comes out stilted and repetitive.
Humans know when they don’t know something
When I was in school studying computer science, we had a regular discussion about how much a piece of software can know, and know when it doesn’t know it. The short answer is: a machine’s lack of awareness means it cannot know. Humans easily recognize when they don’t know an answer but can rely on what they do know to create a meaningful and relevant interpretation.
Humans are creative
AI is great at describing mundane details but struggles to reframe a statement into something culturally and contextually appropriate.