AI Voice Control's Supreme Command Poses Potential Hazards, According to Kardome CEO
In a world where AI-powered voice systems are becoming increasingly common, concerns about privacy have taken centre stage. Dani Cherkassky, CEO of Kardome, a human-machine interaction startup, has warned that the advance of AI within voice-activated systems could accelerate the leaking of private data unless it is ring-fenced from third-party data-gathering operations.
Cherkassky's concerns stem from the fact that the effectiveness of AI-powered voice systems requires them to be on all the time to learn and interact with the user and their needs. This constant connection raises questions about user privacy, as personal data can be mined without the user's knowledge.
To address these concerns, Cherkassky suggests a different approach. Instead of large language models used by several mainstream AI providers, he proposes the use of bespoke tiny language models for interactions with specific devices, which should happen privately on the Edge. This approach, as proposed by Kardome, would mean a voice assistant that works entirely on Edge compute, even without internet connectivity.
This solution is not just a theoretical concept. Kardome's voice assistant system, born out of frustration with modern voice assistance that frequently misunderstands instructions, particularly in noisy background environments, is set to be the first of its kind. The automotive brand BMW will be the first to equip its interior system with Kardome's voice-activated system this year.
Hyundai Motor backed Kardome's system in 2020 to develop a system that can handle instructions from several users in noisy conditions. The rollout of Kardome's in-cabin system by a mass-market global automaker indicates the growing trend towards AI-powered voice systems in the automotive industry.
Cherkassky's approach to privacy-focused AI is not just beneficial for the automotive industry. He argues that systems should be fast, private, and able to work even if the network fails. This approach, he believes, could help alleviate the concerns people have about AI-powered voice assistance, as it addresses the privacy concerns that have been a barrier to widespread adoption.
However, concerns about the constant listening and data uploading of voice systems like Alexa remain. Amazon's practice of automatically sending all voice recordings by users to the cloud for training purposes is a concern, as it deprives users of the ability to keep instructions and conversations private.
In a future dominated by AI-powered voice systems for cars and other devices, according to Cherkassky, systems have to be tailor-made for specific uses to protect privacy, and should not automatically send data to the cloud. This approach, he believes, could pave the way for a more trustworthy and widely adopted AI-powered voice system.
