Language I/O is thrilled to announce the addition of Diego Bartolome, Chief Technology Officer, to our executive team. With more than 16 years of experience as a leading figure at the intersection of languages and technology, he has extensive expertise in multilingual technology and artificial intelligence. As part of coming on board, Diego is sharing his vision for the future of responsible AI at Language I/O.
Though perspectives vary on the benefits and drawbacks to using artificial intelligence (AI), the fact that it is changing the way we live is indisputable. From self-driving cars to voice assistants to predictive text, AI is woven into the fabric of our everyday lives—and this doesn’t stop when we go to work. In 2019, Gartner research found that 37 percent of organizations had implemented AI in some capacity; most recently, Gartner has forecasted that the global AI software market will be worth a total of $62.5 billion in 2022, demonstrating an increase of 21.3 percent from 2021.
AI also plays a critical role in the customer experience as we know it. With customer expectations for support higher than ever, brands must implement the latest technology in order to meet these demands and retain customers. In fact, some predictions believe that AI will power 95 percent of brand interactions with customers as soon as 2025.
But though AI continues to grow and evolve at an exponential rate, the question still remains: how can businesses ensure, to both customers and employees alike, the highest level of data privacy, confidentiality, and security?
What is Responsible Artificial Intelligence?
What responsible AI means to each organization may vary, but most would agree that responsible AI involves a concerted effort to reduce the biases of the models and preserve the confidentiality and privacy of the data being used.
In the past several years, many of the world’s largest global corporations have started to build frameworks and share knowledge related to their policy on responsible AI. Google, for example, has an entire section of its AI-focused subdomain dedicated to its practices and policies for responsible AI. Microsoft, meanwhile, has recently released a comprehensive article covering its recent advancements in its responsible AI research. Additionally, the European Commision has two main principles in its implementation of AI: excellence and trust.
As more brands adopt language related to responsible AI, customer expectations are going to increasingly demand that the organizations they work with are proactively incorporating responsible AI frameworks into their business practices.
What Responsible Artificial Intelligence Means to Language I/O
Language I/O uses AI and machine learning (ML) technologies to power its real-time multilingual solution for customer service teams. As a result, our systems process a massive quantity of data every single day, and we are committed to ensure the highest level of data security, confidentiality, and control by the owner of that data. We are therefore ISO 27001 certified and GDPR compliant. We are currently working towards additional security certifications like PCI, HIPAA, and SOC.
However, we go far beyond those requirements. Customers might share various types of Personal Identifiable Information (PII) with support teams via live chats, emails, and tickets submitted for translation. Before we query any of the machine translation (MT) engines we utilize (Google, Microsoft, DeepL, Systran, and Amazon), we protect that PII and encrypt data in transit, ensuring that none of that sensitive information is sent beyond our services. We only integrate with MT engines that agree in writing not to hold onto the data we send them for translation in case any PII sneaks through our detection/encryption process. Furthermore, none of that information is stored in any of Language I/O’s log files or databases.
The only instances where we keep translation data is when customers want to collaborate on further increasing the quality of their machine translations through big data analysis, or in the case of customers who want to improve the consistency of their terminology via our Self-Improving GLOssary (SIGLO) solution. In the case of big data analysis, once the customer approves the conditions of a temporary data storage with encryption at rest, only a dedicated member of our R&D team will have access to the data and process it according to the conditions established in the customer agreement. As for SIGLO, that data is segregated and encrypted at rest, and our customers have full control over it. Most of the work is handled automatically by our continuous active learning processes that have a deep impact on translation quality.
Beyond Data Security and Privacy: Eliminating Bias in Our Models
Data security is of the utmost importance, but we believe that responsible AI goes far beyond it. Namely, eliminating any bias embedded into models powering our MT solutions. In our regular MT benchmarking of the vendors we connect to, we evaluate the bias that may be present in their models so that we can advise our customers on the best engine, while also correcting some of those biases with our pre- and post-processing rules as well as with SIGLO.
Any feedback we receive from our customers is also treated in the same manner, ensuring that Language I/O’s MT output is always the best quality our customers can achieve. While we cannot remove the biases inherent to AI systems trained on large quantities of data, we can certainly minimize their negative effects. Furthermore, we make the decisions of the MT engines easier to understand (interpretable AI) thanks to our rating systems and other business metrics that we capture during the translation process.
Finally, we all know that AI is evolving at an extremely fast pace. Language I/O is committed to providing technology that makes customer support efforts more efficient for both brands and their customers. We use AI to empower customer support teams to communicate with customers in any language as if it is their own, with transparency to all the different stakeholders in the process.
Looking for a responsible customer engagement solution to power multilingual conversations? Reach out to Language I/O for a demo or to speak with a specialist.