Ariel Seidman - Taking on Google Maps, Crowdsourced mapping & Crypto - MBM#58
Jan 1, 2024
auto_awesome
Ariel Seidman, co-founder of Hivemapper, discusses competing with Google Maps, crowdsourced mapping, and crypto. They explain the process of collecting data for maps using dash cams and how contributors are paid. The role of cryptocurrency in building trust is explored, as well as privacy measures in data collection. The podcast also covers mapping objects in a map AI system and the benefits of exterior-mounted dashcams. Book and podcast recommendations are shared, and the changing landscape of journalism and technology is discussed.
Hivemapper aims to compete with Google Maps by crowd sourcing mapping through dashcam data collection.
Hivemapper rewards contributors with crypto-based tokens called Honey, providing transparency and trust in the system.
Hivemapper prioritizes user privacy by masking faces and license plates in collected imagery.
Hivemapper is developing a cellular dashcam option and refining object classification to improve data collection for contributors and customer experience.
Deep dives
Creating a Competitor to Google Maps: The Quest of Hivemapper
Ariel Sedeman, a former Yahoo employee, started Hivemapper in 2015 to provide a competitor to Google Maps. Hivemapper aims to crowdsource mapping by collecting data from dashcams. Contributors upload dashcam footage to Hivemapper's servers and are rewarded with crypto-based tokens called Honey. The use of cryptocurrency provides transparency and immutability, ensuring trust in the system. Hivemapper also prioritizes user privacy by masking faces and license plates in the collected imagery. They have standardized dashcams to ensure accurate data collection and are developing a cellular dashcam option for increased ease of use.
The Data Collection Process and Map Creation
Hivemapper's contributors use dashcams while driving to collect images and footage. The frame rate of the dashcam adjusts dynamically based on the driving speed, optimizing data consumption. Once the data is collected, it is processed and mapped using Hivemapper's image API. The map AI pipeline extracts relevant objects such as speed limit signs, stop signs, and traffic lights. Accuracy is achieved through a combination of 3D reconstruction, stereo camera technology, and user feedback. Privacy is prioritized, with contributors remaining anonymous and privacy zones being respected.
Privacy and Trust in the Crypto-Ecosystem
Hivemapper strives to protect user privacy and build trust within the crypto-ecosystem. Contributors are rewarded with Honey tokens that hold real financial value. The transparency and immutability of the cryptocurrency blockchain ensure trust and prevent fraudulent behavior. The crypto enthusiasts within the community embrace the financial incentives, while others prefer to receive a cash payout in USDC, a token pegged to the US dollar.
Future Developments and Focus on User Experience
Hivemapper is continually improving its technology and user experience. They are developing a cellular dashcam option to eliminate reliance on users' smartphones. They are also focusing on refining the classification and positioning processes for objects in the collected data. Hivemapper's goal is to provide an easy-to-use data collection experience that benefits both contributors and customers.
Ridding the Scene of Distractions
The podcast episode discusses the process of using dashcams to capture imagery for creating maps. The speaker explains how they remove unnecessary elements from the scene, such as the sky and other vehicles, in order to focus solely on the objects of interest, like utility poles and traffic lights. By simplifying the 3D scene, they can accurately position and differentiate these objects in the real world. The advantage of having multiple shots on goal, as opposed to just one, like Google's cars, allows for continuous refinement of object positioning over time.
Map AI Trainers and Auditing
In the podcast episode, the speaker describes how the collected imagery is presented to map AI trainers who analyze it for accuracy. They use satellite views to verify the correctness of objects like speed limit signs. However, they primarily focus on identifying problematic and poorly positioned objects, which are then reviewed by map AI trainers to improve the positioning. The process also involves auditing before finalizing map features, ensuring that all exceptions and errors are identified and corrected.
Customer Categories and Future Vision
The podcast episode categorizes potential customers into automotive and ADAS, robo taxis, and logistics and mobility companies. The speaker highlights the importance of precise ETA and location monitoring for these customers, emphasizing the value of accurate and up-to-date map data. Looking into the future, the speaker envisions the camera technology becoming open source and integrated into more vehicles, enabling owners to contribute to map creation and maintenance on a global level, resulting in a higher quality and fresher map data. Collaboration with auto manufacturers is also considered for wider adoption of the technology.
Ariel Seidman is one of the co-founders of Hivemapper, a company building a map through selling dashcams & paying contributing drivers with the aim of competing with Google Maps. Ariel has a long history of mapping, working on Map & Search at Yahoo in the mid 2000s.