'Hi, Renault. Warm my butt!' New Filante features advanced conversational AI

A JoongAng Ilbo reporter slid into the driver’s seat of Renault Korea’s new car, the Filante, and decided to skip the technical jargon for a very important request on Monday.

Display of Renault Korea’s new model Filante, featuring SK Telecom’s in-vehicle AI assistant service, A.Dot Auto. [SK TELECOM]

“Hi, Renault. Warm my butt!”

A JoongAng Ilbo reporter slid into the driver’s seat of Renault Korea’s new car, the Filante, and decided to skip the technical jargon for a very important request on Monday.

Without missing a beat, the car cheerfully replied, “I’ll turn on the heated seats,” and the seat warmer started immediately.

The reporter used the everyday Korean slang eongtta for his request — meaning “warm my butt” — and the AI obediently got to work, proving it can understand drivers at their most conversational moments. No “seat heater,” no “heated seats,” no "please warm up my seat," just eongtta.

The flawless execution of the seat-heating was possible because SK Telecom’s large language model, A.Dot X 4.0, has been embedded in the vehicle’s infotainment system. Conventional in-car voice recognition tends to understand only preset commands and often misfires, but A.Dot Auto carried on a natural conversation without much inconvenience.

Beyond controlling in-car functions such as climate settings and opening and closing windows while driving, A.Dot Auto can also search for information.

“Who is the newly appointed Fed chair?” asked the reporter, to which the machine responded, “Former Fed governor Kevin Warsh has been nominated. However, it is uncertain how the confirmation process will proceed in the Senate.”

Renault Filante [RENAULT KOREA]

When the reporter followed up by asking whether Warsh is a hawk or a dove, the machine replied with an informed and balanced answer.

“He showed hawkish tendencies in the past by opposing quantitative easing, but he has recently made remarks supporting rate cuts, which could also be interpreted as dovish,” said the machine.

The machine can also handle more emotional, friendly exchanges. “I just had a breakup,” confessed the reporter. “What do I do?”

“That must hurt a lot,” assessed the machine. “It’s important to give yourself time and slowly sort through your feelings.”

There were limits. When someone spoke from the back seat, voice recognition sometimes struggled because of noise. It also failed to give satisfactory answers when more detail was needed.

Renault Filante [RENAULT KOREA]

For example, when asked to “tell me about good restaurants nearby,” it responded based on information from T-map, SK Telecom’s navigation app. But when asked to recommend nearby restaurants suitable for going with a child, it did not respond properly.

Still, A.Dot Auto also includes functions and technologies tailored for the in-car environment.

“We set the AI assistant’s answers to be short — around 80 characters — so it won’t distract from driving,” said Choi Byung-hwi, product manager for A.Dot Auto at SK Telecom. “The service applies automatic speech recognition technology designed to accurately capture voices inside a vehicle. It also features proactive AI that recommends actions first, such as suggesting that drivers close the windows when fine dust levels are high.”

In the global auto industry, competition has intensified to build “cars that talk well,” adding AI assistants to in-car infotainment systems since the emergence of ChatGPT.

Overseas, Mercedes-Benz has moved first with assistants based on ChatGPT and Gemini, while Tesla has introduced an AI assistant built on Grok. In Korea, Hyundai Motor and Kia have applied ChatGPT-based assistant functions to some models since 2024.

Renault Filante [RENAULT KOREA]

Renault Korea’s Filante is the first case of commercializing an in-car AI assistant in Korea based on a homegrown generative AI model.

“There is a need to implement specialized in-car services that allow conversations to be as natural as possible without interfering with driving,” said Jang Hong-chang, a principal researcher at the Korea Automotive Technology Institute. “Using overseas models can reduce development costs and feel familiar to consumers, but it is difficult to maintain control over data and security. We need to further improve the performance of domestically developed AI models.”

This article was originally written in Korean and translated by a bilingual reporter with the help of generative AI tools. It was then edited by a native English-speaking editor. All AI-assisted translations are reviewed and refined by our newsroom. BY KANG KWANG-WOO [kim.minyoung5@joongang.co.kr]