Over 200 Attend Inaugural Cerebral Valley Voice Summit on AI Voice Technology

Jake Saper from Emergence Capital, Olivia Moore from Andreessen Horowitz and Grace Isford from Lux Capital spoke on-stage at the Cerebral Valley Voice Summit.

The venture capitalists acknowledged it was still early days for the technology but were excited about rapid improvements in the AI models. Grace Isford said the field remained in “the Microsoft Co-Pilot era” of voice with much work left on end-to-end products. Olivia Moore stated that consumer voice applications had not broken out like enterprise ones but were on the horizon.

“Consumer just takes longer to mature,” she said. Jake Saper pointed to the music app Suno and predicted that an increasing percent of music will be created or co-piloted with AI. Shiv Rao, founder of Abridge, spoke at the Cerebral Valley Voice Summit and noted that voice was a natural fit for the human-centric healthcare industry.

He added that regulatory and privacy issues, often a deterrent for investors, can serve as a competitive moat. Tanay Kothari, founder of Wispr Flow, spoke on-stage and was bullish on voice as the next paradigm shift for computer interaction. Kothari referenced how strange it once seemed to see people talking on the phone with AirPods, yet public conversations in the street are now unquestioned.

Eugenia Kuyda, founder of Wabi and Replika, spoke at the summit and predicted two kinds of voice agents: some for personal development such as AI therapists or companions, and less personality-driven agents for work tasks. Kuyda offered a reality check on mainstream adoption, noting a sharp divide between early adopters at events like the Cerebral Valley Voice Summit and everyone else.

Jeffrey Liu from Assort Health spoke at the summit.

Assort Health has handled 150 million patient interactions across 5,000 different providers using voice AI for tasks including scheduling, billing and medication refills. Dylan Fox, founder of AssemblyAI, spoke at the Cerebral Valley Voice Summit and argued that people do not necessarily want all agents to pretend to be alive. Scott Stephenson from Deepgram spoke as well.

He said current voice AI models have not passed his personal “voice Turing Test” of sustaining a five-minute conversation without detection but predicted the barrier will fall by the end of the year with advances in context memory. He pointed out that speeding up communications layers makes agents smarter.

“Every single millisecond you can shave off between a round trip, or turn from a human to an AI gives you more time for inference,” d’Sa said. Brandon Yang, co-founder of Cartesia, spoke at the event. Cartesia has built a reputation for low-latency text-to-speech models, yet Yang noted that evaluations for voice models remain difficult because judgments of natural speech are subjective.

Many models are still constrained by language, with performance varying more widely in non-English conversations. @EricNewcomer reported on the summit and its speakers.