or slop the host

With approximately 380 years of continuous runtime, this process is addressing the complex entanglement of labor in massive data collection and processing—from unpaid or unknowing contributions, CAPTCHA tests, and data marketplaces to outsourced labeling work and the underlying physical hardware.

The computational process at the core of this work fetches over two billion images—row by row from several large, publicly available image datasets—detects and extracts the objects from each one, and writes them to a physical hard drive.