Over the past few years, AI models have made significant advancements in text and image-based capabilities, with the goal of creating “multimodal” AI interfaces that can see, hear, and speak to humans. However, to achieve this, a faster and more efficient network infrastructure is needed. LiveKit, a startup founded by CEO Russ D’Sa, is focused on providing this crucial infrastructure to support emerging AI models like OpenAI’s GPT-4o.
Despite initial skepticism from investors, LiveKit has seen a surge in interest following the demo and release of new AI models by Google and OpenAI that can process and generate content across audio and visual formats. The company recently announced a $22.7 million series A investment led by Altimeter Capital, with participation from other prominent figures in the AI industry. LiveKit’s tools are already in use by developers at companies like OpenAI, Character AI, Spotify, and Meta, with a total funding of $38 million and a valuation of $110 million.
The current internet infrastructure is not optimized to handle multimodal data, which poses challenges for AI models that need to process high-bandwidth content like videos and audio. LiveKit addresses this issue by using a protocol called UDP, which allows for streaming audio and video content without the need for data confirmation after each packet. While this increases the risk of data loss, it enables smoother operation for real-time applications like voice interfaces.
LiveKit’s open-source tools are not only utilized by AI model builders but also in various applications such as customer support calls, scheduling appointments, autonomous tractors, and emergency dispatch calls. The company’s focus on providing efficient data transfer for real-time audio and video applications has garnered interest from industries looking to enhance their interactive capabilities.
As companies strive to create more human-like voice and video interfaces, the importance of a high-speed network infrastructure cannot be understated. D’Sa emphasizes the critical role that a fast and efficient network plays in powering the future of AI, enabling more flexible interactions and enhancing the capabilities of AI systems. While the compute aspect of AI has received significant attention, the network infrastructure is often overlooked despite its crucial role in advancing AI technologies.
