Useful Sensors Inc.

I was lucky enough to catch up with Ann Spencer to talk about why speech to intent is going to be so important, here's our conversation.

Useful Sensors CEO Pete Warden recently sat down with Om Malik to discuss the evolution of the Internet of Things (IoT), its shortcomings, and how Useful Sensors is paving the way to create a universe of smart devices that understand how to interact with each other. Dive into the interview and hear Pete’s thoughts on the future of connected devices! 

Useful Sensors is at the Cerebral Valley AI Summit, hosted by Eric Newcomer!  Pete Warden and Jacqueline Cervantes will present a demo of how Torre provides instant, accurate, and private translation powered by Moonshine. If you’re in attendance, check out the “Emergent Behavior: Cutting-Edge AI Product Demos” session! 

Speech recognition models can improve everyone’s experience with language translation. In building Torre, our dedicated translation device, our team created a bespoke family of models that outperform existing offerings. Torre is powered by Moonshine, a family of speech recognition models optimized for live transcription and voice commands, especially for devices with limited resources like low-cost hardware.  Our team recently published a paper demonstrating how Moonshine surpasses OpenAI’s Whisper model in several areas, using fewer computational resources while delivering notable improvements: ✅ 5x reduction in compute requirements compared to Whisper when processing a 10-second speech segment. ✅ Operating at low latency, using fewer computational resources while maintaining the same accuracy level as Whisper.  ✅ Full offline functionality, ensuring privacy and security without cloud dependency. Moonshine is better optimized for low-latency, real-time speech recognition, especially on devices with limited hardware resources. There should be no limits to giving an optimal speech-to-text translation experience across a broad set of devices. Moonshine has over 2,200 stars on Github and counting; try out Moonshine yourself and let us know what you’re building! https://lnkd.in/gfGGMz58

We’ve said farewell to #SpookySeason, yet there’s one haunting story our CEO and founder Pete Warden recently shared with IEEE Spectrum. In 2011, Pete, who had previously worked at Apple, discovered with fellow security researcher Alasdair Allen that iOS 4 was storing his and other iPhone users’ latitude-longitude coordinates in unencrypted files. Wanting to alert the company of the potential privacy challenge, Pete sent Apple an email informing them of the issue. The pair published their findings and built an open-source application that enabled iPhone users to visualize their location data on a map. Suddenly, Pete and his colleague received calls from major news outlets, such as The Guardian, Bloomberg News, and The New York Times. Check out the rest of the tale in IEEE Spectrum:

Children whose parents don’t speak English often serve as de facto translators, helping their families navigate doctor appointments, teacher conferences, trips to the store, and so much more. Telemundo spoke with Jacqueline Cervantes (en español!), our Head of Sales and Business Development, about this challenge and the motivation for creating Torre, our dedicated translation device.  Torre delivers real-time, accurate, and secure translation, which can relieve the burden on those who have to fill this role. Check out the interview to learn more! https://lnkd.in/g3uCKwCq

Congratulations to the team at Useful Sensors Inc. on the release of their Moonshine speech recognition models! Their novel architecture reduces compute needs and increases speed, opening new horizons for real-time and resource-constrained applications. Trained on a Lambda 1-Click Cluster https://bit.ly/3BULvJs ❤️ https://lnkd.in/gztNfDvF