Harry Dewhirst: What If the Physical World Had an API? — 375ai, DePIN, and Real-Time Data for AI

Harry Dewhurst, co-founder and CEO of 375 AI, discusses the company's innovative approach to capturing and monetizing real-world data through a decentralized network of sensors. 375 AI utilizes devices on billboards and storefronts to gather data on traffic flows, vehicle counts, and weather conditions, processing this information with edge AI and incentivizing participation through on-chain tokens.

Dewhurst's entrepreneurial background includes significant experience in data analytics and a focus on decentralized physical infrastructure networks (D-PIN). Initially deploying D-PIN devices from other companies, 375 AI pivoted to create its own network, holding an exclusive contract with Outfront for access to 50,000 billboard locations. Their advanced sensors, including NVIDIA GPU cores, capture detailed vehicular data, addressing the challenge of processing noisy, unstructured data while ensuring privacy.

The company’s granular data collection method allows for accurate analysis of reach and frequency, continuously enriching its dataset to provide real-time, high-fidelity data. This approach has attracted paying customers early on, generating immediate on-chain revenue. The devices, priced at $50,000 each, are strategically placed on major routes, enhancing data monetization opportunities, particularly in urban areas.

A key technical decision involves running AI at the edge for efficiency and privacy, processing raw footage on-site and deleting the original video. The modular system architecture allows for easy upgrades, and the smaller 375 Street device is designed for self-deployment, enhancing data collection in diverse locations. This device contrasts with HiveMapper's mobile dash cam approach by gathering comprehensive information about all vehicles in view.

The introduction of 375 GO enables users to contribute data via smartphones, collecting anonymized location and RF signature data valuable for ISPs and network operators. Future plans include encouraging user participation to report real-time gas prices for rewards, increasing network utility.

The demand for real-time data is emphasized, particularly for Large Language Models (LLMs) that require live context. The conversation touches on the value of proprietary data for businesses and the intention to partner with AI companies to become a leading source of physical information. Challenges in manufacturing, such as chip shortages, are acknowledged, alongside aspirations to deploy nodes globally, humorously suggesting the potential for spotting aliens.