For humanoid robots to operate in the real world, they’ll need more than a sense of sight. | Source: Adobe Stock
At recent major tech events, humanoid robots were everywhere. Some of them walked, navigated, and manipulated objects with a level of dexterity that would have seemed unrealistic just a few years ago. And yet, despite all of this progress, I find myself consistently underwhelmed when I try to communicate with them at Treble.
Supporting the development of advanced audio and voice technologies has been Treble’s bread and butter for a long time. We’ve already grown used to building powerful audioAIfor world-leading products. So, naturally, my perspective is shaped by that. Still, I find it striking how underdeveloped communication and auditoryperceptionremain in robotics.
We’re building astonishing machines that can move like us and, in many cases, see better than us. But these machines can’t interact with us in a natural way yet. In my opinion, this is a central challenge that can determine whether this technology is truly adopted.
In my view,human-machine interactionwill become the defining hurdle for widespread acceptance of robots by humans because it heavily impacts trust, safety, efficiency, and ease of use.
Humanoids, and more broadly mobile physical AI, are unlikely to be accepted by people until they can communicate effectively through voice. And not just in controlled, quiet environments, but in the places where real life unfolds.
This includes crowded trade show floors, industrial settings, city streets, and homes filled with noise, movement of sound sources, reverberation, and unpredictability. These are the environments in which humans operate, and they’re precisely the environments where current embodied AI audio and voice systems tend to break down.
At the same time, it’s completely understandable why sound has not been a primary focus. The current competitive landscape in robotics rewards visible, measurable progress.
Locomotion is a clear signal of advancement.Vision-based perception has a mature and powerful ecosystem behind it, with abundant data, well-established models, and scalable training pipelines.
The entire stack, from data collection to simulation, has evolved to support these modalities. Data is abundant, benchmarks are clear, and improvements are easy to demonstrate.
Simulation, in particular, has become a cornerstone of progress. Platforms likeNVIDIAIsaac Simhave enabled rapid iteration and large-scale training in ways that were previously impossible. These systems are powerful, well-designed, and aligned with the broader economics of the industry.
They also reveal something important. The environments we use to train intelligent machines are overwhelmingly visual. But, they are, for the most part, silent.