Language AI Learns the Art of Smooth Talking

Last week, we looked at how research on a multilingual knowledge base at Google could be forging the beginnings of truly multilingual language AI, marking one of the more interesting recent technological developments for the localization industry. Now, days later, new reports on a study at Facebook AI Research appears to signal even more compelling developments in NLP (natural language processing) than systems that can cross-reference world idioms. Challenging criticisms that NLP agents tend to be flat approximations of human speakers, it appears that by inserting a chatbot into a video game where it can interact with other characters, researchers have taught the bot to use language not only do its own bidding but also get others to do things for it.  

Imagining that all of the above could be combined in a single AI, what should we make of a system that can wield persuasive rhetoric in any of the world’s languages? Considering the nature of machine learning, one possible conclusion is that such a system could steadily learn what works and what doesn’t when it comes to engaging people to get results, country by country and region by region. These types of insights could be a powerful tool in cross-border business and cross-cultural communication, whether the engaged parties opt to speak in an agreed upon common language or instead to engage interpreters. Even native or professional experts in a language may fail to grasp subtle but decisive nuances that can make or break an exchange, correlating with real-world gains and losses in business scenarios. Language AI with an ability to improve on these margins could potentially furnish a basis for new training and coaching programs for business professionals, evaluation metrics for interpreters, and software tools that can be used when planning for crucial meetings. Even in monolingual settings, such a system could provide metrics for assessing communications – and communicators – in the workplace.  

A look at AI’s potential is always also a look at its unknown threats. A software that knows too well how to get what it wants from people sounds eerily sociopathic, and it is troubling to imagine people dispatching such a tool to do their bidding even without considering what it might manage on its own. Nevertheless, commentary this week that AI could be approaching a need for not just regulation but accreditation – for instance, certification as ‘professional’ in whatever practice it is conceived for – suggests that a multilingual AI business consultant that can help solve communication challenges is exactly the kind of thing we need to begin to account for as AI’s applications expand.  

Regarding whether these capabilities will replace or rather augment and assist human efforts, the latter seems quite promising. Professional athletes use technology to analyze their own and their opponents’ performance; professional chess players may agree to adjourn a match when both players want to consult outside perspective on a position, which often means practicing possible lines against computers; with AI that understands the rhetorical power of specific words, phrases, and patterns better than people do, businesses might soon have analogous resources and practices that empower their efforts when preparing to compete in global markets. 

CSOFT International works with leading technology firms and enterprises to help ease barriers to entry for their products in new markets through targeted, customized localization services. Our global network of linguists, engineers, and subject matter experts excels at documenting and delivering new technologies across borders in over 250 languages, ensuring both the accuracy and cultural appropriateness of your brand and content for consumers in world markets. Learn more at csoftintl.com!