SRI Ventures Makes Strategic Collaboration in Pienso

Pienso Licenses Speech Processing System to Customize Large Language Models with Voice Data and Voice-Based Prompts for Commercial and Government AI Users

Pienso, the AI software company that provides the building blocks of generative AI to enterprise and public sector users, announces a strategic collaboration with SRI International. For nearly 80 years Menlo Park-based SRI has collaborated across technical and scientific disciplines to discover and develop groundbreaking technology and bring them to the marketplace. 

Pienso software equips would-be generative AI builders to fine-tune Large Language Models with their own data, without coding and while maintaining data privacy standards. Through this investment, Pienso will license SRI’s Open Language for Voice Exploitation speech processing system, OLIVE. OLIVE provides robust speech information extraction from real-world data, delivering accuracy despite high-levels of noise and voice distortion.

Aligned with Pienso’s mission of empowerment, OLIVE equips users to detect meaning from a wide range of languages and dialects. Once embedded in Pienso’s no-code/low-code software, users will be able to use raw voice data to align a Large Language Model to their specific needs, skipping slow and error-prone transcription.

OLIVE is highly versatile and malleable and is especially well-performing with diverse dialects, a benefit Pienso recognized as valuable through the company’s work with commercial call centers which deploy their AI software to categorize and analyze calls from customers throughout the UK, where nearly 40 dialects of English are spoken.

“A Welsh customer calling to cancel service sounds very different from a Northern Irish customer – a conversation with a customer from Liverpool can sound vastly different from a conversation on the same topic with a customer from Newcastle. But they’re all speaking English. The same is true in many languages with diverse dialects; this presents a problem for speech recognition in AI,” says Karthik Dinakar, Pienso’s Chief Technology Officer and co-founder. “In order for voice ingestion to offer value to high-throughput AI use cases like those in commercial business and government, its accuracy must increase. If the conveyed meaning is wrong, the analysis is worthless.”

SRI Ventures works with companies that will apply licensed technology from SRI International to make people healthier, safer and more productive. “We believe that the right IP combined with the right entrepreneur and early product can solve important problems. Empowering a diverse array of end users with the ability to quickly tune AI model outputs to meet their needs is important for both organizational success as well as technology adoption. Pienso is a great example of a company committed to improving both access and efficiency of Generative AI models, specifically LLMs by minimizing the coding as a requirement for users” says Ryan Lewis of SRI.

Pienso’s future plans for OLIVE include building voice prompts into their interactive software, so not only could voice data directly enrich a Large Language Model, but the interaction experience could be speech-based as well.

“Democratization is an overused word, in tech and especially in AI,” explains Birago Jones, Pienso Chief Executive Officer and co-founder. “What we’re committed to is systematically removing barriers that prevent AI use cases from being successful. Often they’re not successful for one of three reasons – AI pilots are too expensive so they never get started; projects require compromising privacy or data sovereignty; or topic experts get divorced from model creation because they don’t have technical skills so models fail to reflect reality.” Jones continues: “Our collaboration with SRI advances efficient, effective use of Large Language Models – the building blocks of generative AI. Imagine talking to your model using human speech, not typing prompts. Imagine your model always speaks your language – in every sense. Your dialect, your organizational vernacular, your historical and present circumstances. Imagine that user experience. Will AI initiatives with that user experience succeed? We bet they will.”