Mercedes-Benz has expanded the capabilities of the MBUX Voice Assistant by harnessing the power of artificial intelligence. It can now provide up-to-date answers to knowledge-based questions by starting a Microsoft Bing search. It then generates natural language responses using ChatGPT via the Microsoft Azure OpenAI service.
Thanks to the addition of the AI-powered Knowledge feature to the MBUX Assistant, users can now ask all kinds of general knowledge questions – from pop culture to geography, science, health, and history – and freely talk about the chosen topic to gain more information. The system can also respond to follow-up requests for additional information, even if the speaker uses ambiguous words. Customers can still use the MBUX Voice Assistant to ask queries about everyday topics like weather, stocks, news reports, and sports scores.
The new functionality is integrated directly into the MBUX Voice Assistant. As before, passengers can activate the assistant by saying, “Hey, Mercedes.” This makes it easy to use and creates a safe driving experience with completely hands-free conversation. Alternatively, they can press the speech button on the steering wheel to activate the assistant. The conversation history is stored for up to one hour to make the most of the function, allowing the user to ask follow-up questions.
Starting today, the new function will be available as a free update for more than three million vehicles with the MBUX infotainment system. It will initially be offered in German, British, and American English. Users from other markets can change their system language to experience the feature.
Mercedes-Benz retains complete control over IT operations in the background. All voice command data collected is stored in the Mercedes-Benz Intelligent Cloud, anonymised, and analysed.
To mitigate the risk of customers accessing toxic, harmful, illegal, and controversial content, Mercedes-Benz has implemented a risk assessment tool to evaluate system responses and identify best practises to reduce the likelihood of malicious responses.
