INTERVIEW ByungHyun An, VP & Regional Business Leader, Korea, Cerence AI
AI technology has opened up infinite possibilities in the in-car voice experience, but ultimately, it is the people who understand cars that will perfect it. It goes beyond just having a conversation with a friend or a partner - it includes personalized convenience that considers entertainment, everyday tasks like EV charging and parking, and future expanded autonomous driving and connected environments, all while maintaining the brand's identity. Cerence AI enables deep personalization, from the voice assistant's persona to the user interface, fostering a strong connection between the driver and the car.
written by Han
한글로보기
Looking back, what have been the key changes in the relationship between Nuance, Cerence, and Microsoft, and how has this affected Cerence AI's strategy?
An Cerence AI was previously the automotive division of Nuance Communications and spun out to become an independent, standalone company in October 2019. Nuance was later acquired by Microsoft in March 2022.
Since establishing as an independent company, Cerence AI has been largely focused on AI innovation for in-car assistants. Most recently, we have been working with the world s leading automakers to deploy our generative AI and LLM-powered solutions, advancing agentic AI interaction in the in-car experience.
As part of that work, we partner with Microsoft on CaLLM Edge, our pioneering embedded small language model (SLM) which harnesses Microsoft's Phi-3 family of SLMs and our own extensive automotive dataset to deliver highly specialized AI capabilities, embedded directly within the vehicle's head unit. CaLLM Edge provides a cost-effective AI solution for OEMs looking to save on cloud costs while delivering a reliable user experience with fast performance, even when no connection is available.
Voice services have become ubiquitous on PCs, smartphones, and in vehicles. What does LLM (Large Language Models) like ChatGPT mean to Cerence AI right now?
An Gen AI and LLMs have the power to transform the in-car experience in a few ways: infusing more reasoning so the system can better understand the user s intent, delivering improved intelligence and expanded capabilities to avoid dead ends, and enabling more contextual knowledge including from the car s sensors, data, etc. At Cerence AI, we make this happen through a strategic combination of proprietary and third-party cloud and embedded language models.
We have developed our own proprietary family of language models, CaLLM , which is the foundation for our next-generation in-car AI platform and is purpose built to deliver highly accurate, contextually relevant, and personalized in-car experiences that generic LLMs simply can't match. We also leverage select third-party LLMs like ChatGPT to deliver added generic intelligence; however, we do so specifically with the automotive use case in mind, prioritizing speed, cost efficiency, and safety within the interactions through guardrailing.
When it comes to voice recognition and AI in automotive, what is Cerence AI s unique position, and what can Cerence AI do that no one else can?
An Cerence AI s deep partnerships with nearly every major automaker, unparalleled automotive dataset, and extensive patent portfolio set us apart from general voice assistant solutions. We have a 25-year history of delivering automotive-grade solutions that are deeply integrated within vehicle systems and available across many languages. This unique combination of industry expertise and global reach allows us to understand automakers needs and develop and deliver truly personalized experiences. Plus, we are truly automotive native, meaning we speak the same language as our customers and have a full understanding of how to execute against specific automotive development timelines and processes.
In addition, we are unique in that we offer an end-to-end product portfolio, including advanced sound processing technologies that remove noise from microphone inputs for crystal clear recognition, as well as industry-leading text-to-speech that enables our customers to bring a unique brand voice to their in-car assistant. We ve even extended our innovation to outside the car, like Cerence Exterior Vehicle Interaction, which leverages external microphones to enable voice-activated commands from outside the car, like asking to close the trunk, or turn the lights on. We also developed Cerence Emergency Vehicle Detection which recognizes emergency siren signals and the direction they are coming from, then minimizes distractions like loud music and alerts the driver that an emergency vehicle is nearby, which will be critical as cars become increasingly autonomous.
To communicate more like a person, emotional recognition, emotional expression, and other interactions need to be added. In this era of electric and autonomous vehicles, SDVs, and an emphasis on experiences, how is Cerence AI s solution making cars our friends?
An We are elevating the future of the in-car user experience by enabling vehicles to provide greater levels of personalized, intelligent interaction. By creating a true companion in the car, we are supporting automakers in forming more meaningful relationships with drivers.
For example, Cerence Assistant, our turnkey in-car assistant platform, features AI-powered personalization so that over time, the car gets to know the driver's preferences and emotional "fingerprint" and can personalize its responses accordingly. Our AI isn't just reacting to words and signals - it's understanding the context of the conversation and the driving situation. For example, if a driver says they're hungry, the car will consider the driver's preferences, location, and even the time of day to make relevant suggestions.
Looking at Cerence AI's portfolio, it seems to be divided into Audio & Communication, Conversational & GenAI. What s the difference between them? What are the other key solutions?
An All of our solutions work together to create seamless, elevated experiences that are intuitive and personalized for every driver.
Conversational AI & GenAI: Cerence AI is leading a new era of vehicle intelligence, integrating these technologies into infotainment systems to entirely transform the way drivers interact with them. Our AI solutions give automakers a competitive edge with customizable solutions that champion their brand while providing their drivers with intelligent assistance that can perform tasks for more personalized, fun, and productive experiences. At the core of this is our CaLLM family of language models, which are the foundation of our next-generation platform that will make in-car assistants more conversational, contextual, and capable than ever before.
Audio & Communication: This is the foundation for all interactions, focusing on ensuring clear, high-quality audio within the vehicle, regardless of external noise or other disruptions. This includes Speech Signal Enhancement, a suite of sound processing technologies that remove noise from microphone inputs, providing our world-class voice recognition and natural language understanding engines with clear input from both drivers and passengers.
What criteria do you use to determine domains and commercialization? What factors do you consider in collaboration with customers?
An At Cerence AI, we believe that the in-car experience should be an extension of the automaker's brand. We work closely with OEMs to understand their unique vision and translate that into a customized solution. Our adaptable products allow for deep personalization, ensuring everything from the AI assistant's voice and persona, to the user interface seamlessly align with the OEM s brand identity and foster connection with their drivers.
AI in particular has unlocked a world of possible features, however, we are keenly focused on innovations that elevate the automotive experience. We re tuned into the user experience and are developing solutions targeted to making everyday tasks like EV charging or parking more seamless for drivers.
Some say one of the key topics in HMI is its expansion beyond the car. What does this mean for Cerence AI? What are the future Cerence AI solutions?
An We believe that in the future, voice-activated and context-aware AI solutions have a critical role to play in the way that people use and interact with the technology around them. Our next-generation platform blends voice and touch modalities to provide a truly holistic and intuitive experience. Drivers can effortlessly switch between voice commands and touch interactions depending on their preference and the situation, creating streamlined and efficient interactions that enhance both convenience and safety on the road.
In addition, as automakers look to offer enhanced multimedia options - bringing the living room to drivers cars - we are supporting a seamless integration of automotive and home entertainment experiences.
Cars are cost-sensitive, and they need to provide the same level of service while being customized to different regions. How does Cerence AI address this?
An We work with automakers across global markets, each with unique needs and budgets. Our technology is designed to be flexible and accommodates local languages, dialects, and accents, ensuring clear communication and accurate voice recognition. We also provide the tools for automakers to incorporate region-specific content, services, and preferences, creating a truly personalized experience for drivers wherever they are.
Which car models currently offer the best voice services, or what would Cerence AI like to highlight?
An We work with many automakers to create solutions that are tailored to their needs. Each application is therefore unique and brings its own benefits to each customer base.
Finally, is there any message you d like to share with Korean readers?
An Cerence AI is committed to partnering with Korean automakers and drivers. We are offering language support for the Korean market as diligently as we do for other major languages. Additionally, we are collaborating with leading OEMs in Korea and supporting their success in the market. We will continue our efforts to provide features that meet the needs of all customers in Korea.
<저작권자 © AEM. 무단전재 및 재배포, AI학습 이용 금지>