The Japan Lens

In the last post, we discussed how high-utility A.I. can be useful for translating or interpreting in low-risk situations. We also looked at how even the latest A.I.-powered language tools* are not up to the job when it comes to diplomacy since too much is at stake. In this post, we will delve a little deeper into intercultural communication. Hopefully, this will help readers understand why diplomatic interpreting isn’t just about mechanically translating words or phrases.

Please note, this is written from the perspective of U.S. diplomacy since that is my area of expertise, but many of these points apply to all diplomatic interpreting, regardless of nation.

With that, here are three more points U.S. officials should keep in mind when contemplating use of A.I.-powered tools in diplomacy:

(4) Figure out the OPSEC

Technologists in Silicon Valley often lampoon American officials for being tech slowpokes. It’s true, the government is often slower to adopt technology than the private sector. Part of the reason is U.S. officials, especially diplomats, often have to handle sensitive information in restricted environments. Maintaining operational security, or OPSEC, is a daily part of the job. So while technologists focus on functionality, U.S. officials don’t have the luxury of viewing tools so narrowly. In order for it to be useful in public service, it has to be both functional and secure.

But one of the initial hurdles to using Zoom was concern about the company’s servers, many of which were located in mainland China. Similar geopolitical concerns can come into play when looking at A.I.-powered tools that rely on massive language lists called corpora, since these are usually stored in the cloud.

Some may argue that there are OPSEC vulnerabilities to be exploited with human interpreters as well, and that can certainly be true. But, speaking from my experience in the U.S. government, diplomatic interpreters undergo rigorous testing and interviewing, not to mention a battery of background checks. And, in addition to being apolitical public servants who swear to uphold and protect the Constitution, they are also professional interpreters bound by standards of ethics and confidentiality. In short, they are trained and motivated to take OPSEC seriously, and they have proven themselves excellent stewards of the nation’s secrets. Any use of A.I. will have to be conducted in a way that provides at least the same level of security as the current system.

(5) Know where A.I. should never be used

As seen in the A.I. vs. human showdown, computers can spit out words and phrases faster than your average person. As long as the machine’s voice recognition and dictation tools are operating without a hitch – by no means a given, even with all the recent technological progress – then it’s natural for computers to perform better in this arena. But this is just one metric and, as it turns out, not even the most important one. What is truly important in high-quality interpreting isn’t mechanically firing off words at machine gun speed; it’s the uniquely human ability to leverage intercultural competence against the backdrop of a shared context. Here is an example that illustrates this point.

I once had to interpret between two conversants, a Japanese education official and a working mother from the Philippines. He spoke Japanese and she spoke English. She came to the board of education and asked if she could enroll her child in the Japanese school system. The direct translation of the official’s response was, “It would be difficult.” This mother knew a few words in Japanese, “difficult” being one of them, so her face lit up because she thought that, with some hard work, they would be able to pull it off. However, as the interpreter straddling both cultures, I suspected the official really meant, “It’s impossible.” Seeing the miscommunication about to unfold, I gently interjected and confirmed my suspicion with the official, and obtained his permission to rephrase in a way she would clearly understand. I interpreted his utterance into English as, “I’m sorry but it’s impossible.” Needless to say, she was crestfallen. But at least she left the meeting knowing exactly where she stood.

A.I. should never be used to to interpret important discussions in multilingual diplomacy.

In the above scenario, if an A.I. program had been used to robotically interpret the words rather than the meaning, the potential miscommunication would have affected the lives of a few people. Now imagine the discussion was not about one child’s education but rather an international agreement. How many thousands of children would be affected such a miscommunication? Or, if it was about a defense treaty or a trade agreement between two nations, how many thousands and perhaps millions of people would be affected? That is why, even though machines are better than humans at mechanically grabbing each utterance, A.I. should never be used to interpret important discussions in multilingual diplomacy.

Human interpreters interpret meaning, not words. They also have other abilities that are difficult to quantify, such as switching between communication styles that differ from culture to culture, all while leveraging their knowledge of context and history. Such abilities are the purview of natural intelligence**, and the artificial variety has a long way to go before it can even get close. Which brings us to the final point below.

(6) Remember diplomacy is a human endeavor

Many U.S. diplomats will remember that during the worst periods of the Covid-19 pandemic, it was easier and cheaper to conduct non-sensitive diplomatic meetings over remote meeting platforms such as Zoom or MS Teams. After all, there were no advance parties to send overseas, no security arrangements to make for high-level principals, and no interpreters to fly out and house in hotels with other support staff. Digital diplomacy made things cheaper and easier. But, as most diplomats know, the person-to-person connections so crucial to effective diplomacy were found lacking. That is because a lot of human communication takes place between the lines, and sometimes an interpreter’s output will shift in subtle but important ways depending on a nonverbal cue. A president may raise an eyebrow, or a minister may pause a little too long before clearing their throat. Research shows that as much as 55% of communication is nonverbal. Interpreters are incredibly tuned into these nonverbal cues.

In bilateral meetings, leaders are always observing each other’s body language, whether they are aware of it or not. And when they are really concerned about a certain topic, they want to ensure they are heard and seen. A leader may lay their hand reassuringly on the other’s shoulder to express sincerity, or look them in the eye to know their point has been made and understood. These are crucial elements of diplomacy that cannot be outsourced to some app, no matter how artificially intelligent it may seem.

Conclusion

Technologists working to solve the interpreting puzzle are motivated, and they are making a lot of headway. Those who are deep into the process are just starting to wake up to how difficult interpreting really is. Challenges include hearing in noisy settings, dealing with false cognates, comprehending speakers with heavy accents, and producing appropriate intonation and emotion. In order to even make it through the front door of viability, A.I. will have to consistently and reliably clear these hurdles, which professional interpreters deal with every day. After that will come the harder stuff, including how to interpret jokes, slang, jargon, sarcasm, irony, cultural/historical references, euphemisms, acronyms, regionalisms, run-on sentences, and half-formed thoughts, just to name a few. And after that will come the really difficult stuff, such as intercultural competence and the ability to read between the lines.

Doubtlessly, technologists will score some successes in dealing with these challenges, and the world can expect even better A.I.-powered language tools going forward. But, for the foreseeable future, diplomacy that involves cultural context – which is to say, all diplomacy – will need to be handled by trained and experienced diplomatic interpreters, professionals endowed not only with language ability but also highly developed intercultural competence.

Despite all this, some officials will inevitably be wowed by the performance of the latest A.I. tools, not to mention the potential cost savings. Some may insist on asking, “When can we safely replace our human interpreters with A.I.?” To which we can honestly answer, “When you and your boss can safely be replaced. And not a day earlier.”

Key takeaways

The following is a summary of the main points officials might want to keep in mind when considering whether to use A.I.-powered language tools in diplomacy:

(1) Know what you don’t know

First, familiarize yourself with the basics of what interpreters and translators do. You can’t make sound decisions about replacing them without knowing what their work entails.

(2) Separate the hype from the hypothetical

A.I. is useful and promising. But when it comes to highly accurate and effective intercultural communication, it is still lacking.

(3) Discern where A.I. can be used

A.I. might be safely used to interpret or translate mundane, non-sensitive interactions in low-risk settings, provided the process is supervised and controlled by highly trained professionals.

(4) Figure out the OPSEC

Not even the neatest A.I. tools are useful if they cannot be used safely and securely.

(5) Know where A.I. should never be used

Interpreters interpret meaning, not words. A.I. should never be used to interpret sensitive content requiring intercultural competence and a deep understanding of context, which includes knowledge of history, diplomacy, and national security.

(6) Remember diplomacy is a human endeavor

People prefer talking about important issues in-person, and a lot of communication happens between the lines. Now more than ever, the U.S. needs highly trained bilingual and bicultural linguists to facilitate the nation’s diplomacy.

– – –

* The term “A.I.-powered tools” is used here to refer to both language translation and interpretation platforms, since the line between these has blurred in the tech realm due to rapid development of voice recognition and audio functions.

** Despite their intelligence, human interpreters can sometimes get it wrong in diplomatic settings. In many East Asian nations, this fear is so prevalent that governments go to great lengths to mitigate risks. Their principals often read from carefully curated scripts, and their interpreters dutifully follow along while reading from pre-translated versions of said scripts. For Americans who have been trained from adolescence to communicate in a dynamic and engaging fashion, this style of discourse can seem boring and monotonous. But, unfortunately, many U.S. officials take it to the opposite extreme. They make jokes to break the ice, or ad lib to show their comfort and familiarity with a particular topic. They launch into lengthy speeches without briefing their interpreters, forcing them to shoot from the hip. Part of this is because, ironically, many American officials think interpreters are basically walking/talking versions of Google Translate. As a result, U.S. interpreters often have to do their jobs with one arm tied behind their backs, and the diplomatic team fails to capitalize on their deep well of linguistic and cultural knowledge. Such self-inflicted wounds are entirely preventable, and U.S. leadership should make concerted efforts to institutionalize best practices.

Photo 1: A widely available A.I.-powered language tool.

Photo 2: Interpreters on the sidelines facilitate a personal moment between U.S. President Biden and South Korean President Yoon at Camp David, Aug. 2023. (White House photo)

Note: No artificial intelligence or machine translation programs were used in the creation of this post.