For everyone keeping up with our posts on the latest developments in artificial intelligence, two major themes in how AI is impacting communications are likely already familiar points of interest: how it is changing industries that need communication solutions, and how it is changing communications themselves. On the one hand, the profound changes AI technology is bringing to both new and existing products and services in the market requires a great deal of content to document and market, particularly when entering overseas markets. On the other, AI is rapidly taking on more of the communication tasks that have traditionally been reserved for human writers, linguists, translators, and customer service specialists, thanks to the astonishing natural language processing capabilities now emerging in rapid succession. Beyond either of these trends, though, a third consideration tends to surface in interesting ways: How are AI applications changing people’s own communication needs, preferences, and styles – or, seen another way, language itself?
To some extent, everyone reading this article has already adapted their approach to language on account of the machine learning algorithms they routinely interact with. Taking search engines as an example, even minor variations in search terms and the wording of questions can impact search results. People’s appreciation of this fact tends to influence their entire approach to language when looking for answers to their concerns online. Rather than asking what you would ask another human about a topic, you ask an algorithm what you think it can understand, and hope for progress. This somewhat cynical trial–and–error approach to smart technologies tends to be an accepted compromise with their current limitations, but begs important questions. If we are essentially using machine learning babytalk to engage algorithms, does it even matter if they are becoming smarter? And by the time they can understand us as we would talk to each other, will people have the fundamental confidence in plain language to use them effectively? If people spend years adopting the habit of telling chatbots what they think will get them further along a question–and–answer rotary, how will they navigate actually helpful dialogue systems?
If all of the above describes the all-too-familiar scenario of speaking with AI systems that lack a basic, familiar grasp of what you mean by what you say, signs of coming change include this week’s report from MIT Technology Review on an emerging technique called “vokenization” that is enabling researchers to train AI models with something like common sense. By literally pairing computer vision AI with natural language processing AI, the effort appears to be targeting better linguistic outputs through a broader working knowledge of the world that includes a familiarity with appearances. When algorithms can approach problems comprehensively, ‘hacking’ their shortcomings may not be so likely to yield the kinds of favorable results it has so far. One conceivable extreme is a machine that knows you are mincing words with it, but suspects it knows what you are after. From AI that is too smart to understand you, to AI that you simply cannot fool by disguising yourself in language, a lot is possible, and change is certainly coming.
Communication solutions often beget additional communication solutions as they transition in and out of practical use – on the one hand to introduce them when new, and on the other to replace them when old. As search engines become more like human librarians than Dewey decimal systems, today’s sophistication could soon be tomorrow’s confusion. Training, onboarding, and help resources are likely to play a crucial roll in extending a wide range of astonishing new capabilities to general users and consumers, and for AI-based products and services entering new markets these will need to be multilingual content solutions.
CSOFT International works with leading technology providers to help drive the rollout of their products and services in new markets through targeted localization solutions that we deliver with the help of over 10,000 linguists, subject matter experts, and engineers worldwide. By translating brands, products, and their communications into the languages global consumers demand, CSOFT can help you bridge linguistic and cultural borders and realize growth in the global market for AI-based solutions. Learn more about our services at csoftintl.com!