Industry executives and experts share their predictions for 2024.  Read them in this 16th annual VMblog.com series exclusive.

Translation tech in 2024 - AI's impact and ethics

By Heather Shoemaker, founder and CEO of Language I/O

The rapid advancement of AI technology in the last year has resulted in a seismic shift in what is technologically possible to enable seamless, scalable communication across languages. Powerful machine learning models can now translate content with near human-level fluency, nuance and accuracy almost instantly.

This remarkable leap forward unlocks game-changing potential for global business. However, the appropriate adaptation and application of such influential technology remains imperative as well. As generative AI transforms the translation industry in 2024, businesses must balance speed and innovation with business-specific customization, privacy protection and cultural sensitivity in order to provide quality experiences that foster customer trust. It will be critical for organizations to figure out how to maximize the benefits of AI-enabled global communication while proactively addressing risks.

Breaking down the (language) barriers to exceptional customer service

Large language models (LLMs) and generative AI (GenAI) stand poised to revolutionize real-time translation and bring us closer than ever to the vision of seamless global connection.

Gone are the days of clunky phrase-based systems and rigid quality metrics. Now, cutting-edge LLMs can make conversations as smooth and lively as chatting with a fluent multilingual friend. The best translations grasp subtle cultural cues conveying precisely the right sentiment.

Of course, engineering AI is both an art and a science. Future winning platforms will be those focusing first on human needs, not just efficient coding. That focus means deep respect for nuance, dialect, context and emotional resonance no matter what languages are involved. Further, an AI platform is no good to a business if it hasn't been adapted to the context and domain of the business. Businesses will struggle with the right approach to adapt generative AI to be more cultural and domain-sensitive. Few businesses will go all in and pre-train a generative AI model from scratch, given the enormous resources this requires. The next best will be fine-tuning approaches, which start with an existing Gen AI platform but still require immense amounts of tagged data for each language the company requires domain adaptation.

Easier, but less effective methods, in order of complexity, include RAG (Retrieval-Augmented Generation) and simple prompt engineering. RAG is coming to the forefront as the favored method as it provides a GenAI platform with access to missing context, but does not require engineers to tag massive amounts of data or actually alter the generative AI models. Prompt engineering is the least invasive as it involves coming up with clever ways to prompt the GenAI platform so it returns the desired answer. While not as effective as the other methods, it is often coupled with the other methods and will continue to be used well into the ‘20s.

 A domain-adapted generative AI platform is useful across the business but plays a special role in the multilingual customer service approach. Given that nearly 80% of people want to buy products with information in their own language, and existing customers are the fastest path to increased revenue, customer service will likely become one of the early adopters of GenAI within an organization Customer communications in just one language won't suffice.

By harnessing algorithms as allies in understanding, businesses can overcome interactions stifled by language barriers and open new pathways to collective growth fueled by our diversity of thought and experience.

An emphasis on ethical AI

As AI permeates business operations, the data fueling its development deserves thoughtful, holistic protections. With users scrutinizing terms of service, clarity around data collection and application is key.

Robust safeguards governing data collection, storage and processing should become standard elements of corporate tech stacks in 2024. These protections will not only ensure adherence to emerging legal standards and regulations - such as the Biden Administration's recent Executive Order focused on mitigating AI-related risks - but also earn customers' trust. Customer trust drives brand loyalty and organic recommendations; 62% of customers who trust a brand remain loyal, and nearly 90% of loyal customers will suggest the brand to others.

Institutions prioritizing transparency and consent reduce vulnerabilities while future-proofing partnerships. Explicit data permissions allow users to opt out of training datasets. Restricting access minimizes threats and compartmentalizing functions prevents data leakage across departments.

Air-tight data governance communicates a commitment to data protection and addresses ethical considerations. In the new world of generative AI, a solid governance strategy involves a laser focus on which third-party platforms are holding on to your company's and your customer's sensitive data. The best possible governance involves using only AI platforms that will commit to zero data retention policies. Companies hoping to responsibly reap returns have a small window to shore up data integrity foundations. By intentionally self-regulating now, the stage is set for an equitable AI-powered tomorrow.

The importance of thoughtful AI adoption

GenAI's explosion in popularity has escalated already soaring customer experience expectations. Instant service has become table stakes as users grow accustomed to lightning-fast results. Yet racing to plug the latest experimental tech into production may not always be the best idea.

While GenAI's capabilities are impressive, organizations that fail to carefully strategize risk their AI falling flat, or worse, diluting the brand. Responsible adopters know that plug-and-play AI solutions won't necessarily solve every business problem, so they will take the time to slow down their implementation plans and determine what enhancements will solve real pain points. Thoughtful implementation also requires asking:

• What are the most impactful use cases for this technology?
• What is the best strategy to domain-adapt AI technology for your organization?
• How are you managing customer expectations?
• Do you have the security measures in place to ensure data is protected?
  

Answering these questions is critical before launching an AI tool.

The choice between accelerated disruption or sustainable improvement is clear. When expectations keep changing, lasting loyalty derives from relationships not features, and relationships flow from understanding people's needs - not algorithms alone.

As AI capabilities continue advancing at a remarkable pace, nearly every industry, including the translation technology space, is poised for positive disruption through enhanced automation, personalization and accessibility. To ensure these powerful innovations serve customers responsibly and drive business success, businesses must prioritize governance and ethics.

##

ABOUT THE AUTHOR

Heather is the CEO of Language I/O, an AI-powered translation platform that provides real-time language translations in over 150 languages. With integrations with Salesforce, ServiceNow, ZenDesk and Oracle, Language I/O's tech can be up and running for companies in less than a day. Language I/O also has best-in-class security that encrypts data in transit and retains zero data.

Prior to co-founding Language I/O, Heather was well-known for globalizing code for Fortune 500s. She was also the senior director of Product Management and Globalization for eCollege, which was acquired by Pearson Education during her tenure. While at Pearson/eCollege, Heather and her team built a next-generation, online college education platform, which was launched globally.

Heather holds a Master of Science from the University of Colorado at Boulder College of Engineering as well as a Bachelor of Arts in Latin American Studies from the University of Washington in Seattle. She has lived in various parts of the United States and Mexico and speaks English and Spanish.