in All Things Localization, Technology

When bombarded almost weekly with reports of the latest hair-raising machine learning or facial recognition capabilities, it is easy to tune out the rising clamor over advancing AI while a human-obsolete world fails to materialize in our midst. Even so, headlines this week like Microsoft Gets Exclusive Access to AI Deemed ‘Too Dangerous To Release’ may not be all that hyperbolic following news that the preeminent tech giant has gained full rights to begin using the astonishing GPT-3 natural language processing algorithm, a powerful language generation tool that has already made news simply for existing, grasping the attention once again in the AI translation space.

Now potentially on its way toward real world applications, GPT-3 is the latest iteration of OpenAI’s experimental GPT-2 algorithm, known, among other things, for writing entire articles indistinguishable from a human author’s work, and even poetry with metaphorical depth. Simply put, GPT-3 is the kind of AI that people are talking about when they say “AI” with fear and awe, and the thought of its unsupervised release into the world does appear to inspire caution in its creators as well. While OpenAI has previously released an API enabling developers to explore ways of plugging the algorithm’s natural language processing (NLP) capabilities into their applications, teaming up with Microsoft to do the same represents a new departure, as well as a true shift in scale and possibility. Partnering with Microsoft also echoes the need for this technology to undergo development under the close watch of experts with abundant computing resources at their disposal, offering some assurance that we will not be seeing an ‘open season’ on unsuspecting web users at the hands of unaccountable developers.

Related:  The Top 5 M&A Localization Documents for Successful International Integration

What could be in store? At the highest level, activating GPT-3 for development seemingly dispels the ‘slow and gradual’ version of AI’s rollout that some have expressed hope for. While GPT-3 is not synonymous with the idea of an AI singularity, it is strikingly close to being autonomous in language-related tasks. In the past, perfecting multilingual translation has been the primary frontier for powerful NLP tools, and original content creation represents a massive hurdle forward. In terms of applications, this bodes the potential to automate customer service, replace human writers in marketing content creation, and – more concerning to people outside of the language services industry – generate the enormous volumes of media consumed every day by public audiences around the world.

The good news is that thought leadership in technological circles seems to have anticipated these developments fairly well. As TNW reports, influential companies in the AI sector are increasingly looking at how AI can be used to serve human knowledge and decision-making, rather than to amass and process all of our data. While the automation of linguistic tasks may not fit either description, and is heavily demanded by crucial industries like financial services, there is reason to hope that the true value of an algorithm that can communicate credibly in natural language will be in the service of human purposes. In content services, for instance, a collaborative human-AI model has already gained traction as an approach to translation tasks that is often optimal over other solutions, enabling human linguists to focus on adding cultural nuance to machine translations. A tool that can itself work with nuance may not replace this paradigm, but rather drive it to evolve new efficiencies, or provide human linguists with valuable feedback. The opposite may also be true, whereby a person is needed to verify the information that an AI tool has become adept at persuading people of, with analogies to the practice of ensuring in-country linguistic reviews from subject-matter experts to ensure the quality of initial translations.

Related:  New Contenders in Multimedia Translation: Can Automated Dubbing and Subtitling Help Scale Production?

With a full range of technology-driven translation and content creation services that we deliver with the help of more than 10,000 linguists worldwide, CSOFT International is well-versed in best practices for the application of AI translation tools to optimize the efficiency, quality, and ROI of multilingual localization projects. As NLP enters more of the mainstream, day-to-day content and communication needs of more people and businesses worldwide, this is sure to be an exciting period of innovation for the communications sector, and one that will bring new focus to the crucial role of both human linguists and advancing technologies in driving communications for a positive global future.